[NFC][SelectionDAG] Replace generic @llvm.expect.with.probability codegen test with X86 test

[antangelo]

Adds test case for X86 to check that the output of @llvm.expect.with.probability's generic lowering is reasonable. This replaces a generic test which only asserts that llc does not crash.

[llvmbot]

@llvm/pr-subscribers-backend-x86

Author: None (antangelo)

Changes

Adds test case for X86 to check that the output of @llvm.expect.with.probability's generic lowering is reasonable. This replaces a generic test which only asserts that llc does not crash.

---

Full diff: https://github.com/llvm/llvm-project/pull/117848.diff

2 Files Affected:

• (removed) llvm/test/CodeGen/Generic/builtin-expect-with-probability.ll (-8)
• (modified) llvm/test/CodeGen/X86/fast-isel-expect.ll (+33-2)

diff --git a/llvm/test/CodeGen/Generic/builtin-expect-with-probability.ll b/llvm/test/CodeGen/Generic/builtin-expect-with-probability.ll
deleted file mode 100644
index aef134b636d5a7..00000000000000
--- a/llvm/test/CodeGen/Generic/builtin-expect-with-probability.ll
+++ /dev/null
@@ -1,8 +0,0 @@
-; RUN: llc < %s
-
-declare i32 @llvm.expect.with.probability(i32, i32, double)
-
-define i32 @test1(i32 %val) nounwind {
-    %expected = call i32 @llvm.expect.with.probability(i32 %val, i32 1, double 0.5)
-    ret i32 %expected
-}
diff --git a/llvm/test/CodeGen/X86/fast-isel-expect.ll b/llvm/test/CodeGen/X86/fast-isel-expect.ll
index 9c2d6488db3c91..e804076304e309 100644
--- a/llvm/test/CodeGen/X86/fast-isel-expect.ll
+++ b/llvm/test/CodeGen/X86/fast-isel-expect.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc < %s -O0 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
@@ -6,9 +7,17 @@ target triple = "i686-unknown-linux-gnu"
 @glbl = extern_weak constant i8
 
 declare i64 @llvm.expect.i64(i64, i64)
+declare i64 @llvm.expect.with.probability.i64(i64, i64, double)
 
-define void @test() {
-; CHECK: movl $glbl
+define void @test_expect() {
+; CHECK-LABEL: test_expect:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    movl $glbl, %eax
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    jne .LBB0_2
+; CHECK-NEXT:    jmp .LBB0_1
+; CHECK-NEXT:  .LBB0_1: # %bb1
+; CHECK-NEXT:  .LBB0_2: # %bb2
   %cmp2 = icmp eq ptr @glbl, null
   %ext = zext i1 %cmp2 to i64
   %tmp = call i64 @llvm.expect.i64(i64 %ext, i64 0)
@@ -21,3 +30,25 @@ bb1:
 bb2:
   unreachable
 }
+
+define void @test_expect_with_probability() {
+; CHECK-LABEL: test_expect_with_probability:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    movl $glbl, %eax
+; CHECK-NEXT:    testl %eax, %eax
+; CHECK-NEXT:    jne .LBB1_2
+; CHECK-NEXT:    jmp .LBB1_1
+; CHECK-NEXT:  .LBB1_1: # %bb1
+; CHECK-NEXT:  .LBB1_2: # %bb2
+  %cmp2 = icmp eq ptr @glbl, null
+  %ext = zext i1 %cmp2 to i64
+  %tmp = call i64 @llvm.expect.with.probability.i64(i64 %ext, i64 0, double 0.5)
+  %tmp2 = icmp ne i64 %tmp, 0
+  br i1 %tmp2, label %bb1, label %bb2
+
+bb1:
+  unreachable
+
+bb2:
+  unreachable
+}

[RKSimon]

Any reason not to keep the generic test as well to help sniff out regression on other targets?

[antangelo]

Any reason not to keep the generic test as well to help sniff out regression on other targets?

I was told to remove it in the GISel patch to fix the same issue: #117835 (comment)

This is my first set of patches to the generic backend code, so I'm not sure what the best testing methodology is or degree to which tests should be generic vs specific to various backends.

[RKSimon]

LGTM - I expect other targets will hit issues at some point with this intrinsic, but they can hopefully just copy what x86 does.