[X86] Combine bitcast(v1Ty insert_vector_elt(X, Y, 0)) to Y #130475
Conversation
|
@llvm/pr-subscribers-backend-x86 Author: Phoebe Wang (phoebewang) ChangesThough it only happens in v1i1 when we generate llvm.masked.load/store intrinsics for APX cload/cstore. https://godbolt.org/z/vjsrofsqx Full diff: https://github.com/llvm/llvm-project/pull/130475.diff 2 Files Affected:
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 1992ef67164d8..be45a678bbcfe 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -45486,6 +45486,11 @@ static SDValue combineBitcast(SDNode *N, SelectionDAG &DAG,
if (SDValue V = combineCastedMaskArithmetic(N, DAG, DCI, Subtarget))
return V;
+ // bitcast(v1Ty insert_vector_elt(X, Y, 0)) --> Y
+ if (N0.getOpcode() == ISD::INSERT_VECTOR_ELT && SrcVT.getScalarType() == VT &&
+ SrcVT.getVectorNumElements() == 1)
+ return N0.getOperand(1);
+
// Convert a bitcasted integer logic operation that has one bitcasted
// floating-point operand into a floating-point logic operation. This may
// create a load of a constant, but that is cheaper than materializing the
diff --git a/llvm/test/CodeGen/X86/apx/cf.ll b/llvm/test/CodeGen/X86/apx/cf.ll
index c71d7768834f3..216f187d986d6 100644
--- a/llvm/test/CodeGen/X86/apx/cf.ll
+++ b/llvm/test/CodeGen/X86/apx/cf.ll
@@ -124,3 +124,18 @@ entry:
call void @llvm.masked.store.v4i64.p0(<4 x i64> %0, ptr %p, i32 8, <4 x i1> %cond2)
ret void
}
+
+define void @no_xor(i32 %a, i32 %b, ptr %c, ptr %d) #2 {
+; CHECK-LABEL: no_xor:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: cmpl %esi, %edi
+; CHECK-NEXT: cfcmovnew (%rdx), %ax
+; CHECK-NEXT: cfcmovnew %ax, (%rcx)
+; CHECK-NEXT: retq
+entry:
+ %0 = icmp ne i32 %a, %b
+ %1 = insertelement <1 x i1> poison, i1 %0, i64 0
+ %2 = tail call <1 x i16> @llvm.masked.load.v1i16.p0(ptr %c, i32 2, <1 x i1> %1, <1 x i16> poison)
+ tail call void @llvm.masked.store.v1i16.p0(<1 x i16> %2, ptr %d, i32 2, <1 x i1> %1)
+ ret void
+}
|
Though it only happens in v1i1 when we generate llvm.masked.load/store intrinsics for APX cload/cstore. https://godbolt.org/z/vjsrofsqx
It does, but is not necessary for now, because only |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/205/builds/2838 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/203/builds/4047 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/204/builds/2859 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/123/builds/15111 Here is the relevant piece of the build log for the reference |
Though it only happens in v1i1 when we generate llvm.masked.load/store intrinsics for APX cload/cstore.
https://godbolt.org/z/vjsrofsqx