[nfc] Make function-specialization3.ll test compatible with profile propagation unittests

[alanzhao1]

The build configuration described in
https://discourse.llvm.org/t/rfc-profile-information-propagation-unittesting/73595/11 causes this test to fail because that build configuration injects profile metadata which show up in the output. To fix this we just change the assertion to tolerate the presence of profile metadata.

Tracking issue: #147390

[llvmbot]

@llvm/pr-subscribers-function-specialization

@llvm/pr-subscribers-llvm-transforms

Author: Alan Zhao (alanzhao1)

Changes

The build configuration described in
https://discourse.llvm.org/t/rfc-profile-information-propagation-unittesting/73595/11 causes this test to fail because that build configuration injects profile metadata which show up in the output. To fix this we just change the assertion to tolerate the presence of profile metadata.

Tracking issue: #147390

---

Full diff: https://github.com/llvm/llvm-project/pull/156942.diff

1 Files Affected:

• (modified) llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll (+2-2)

diff --git a/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll b/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
index 8e075edaa6844..a26f07bfeb878 100644
--- a/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
+++ b/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
@@ -34,14 +34,14 @@ return:
 
 ; FORCE-NOT: define internal i32 @foo(
 ;
-; FORCE:      define internal i32 @foo.specialized.1(i32 %x, ptr %b) {
+; FORCE:      define internal i32 @foo.specialized.1(i32 %x, ptr %b) {{(!prof !0 )?}}{
 ; FORCE-NEXT: entry:
 ; FORCE-NEXT:   %0 = load i32, ptr @A, align 4
 ; FORCE-NEXT:   %add = add nsw i32 %x, %0
 ; FORCE-NEXT:   ret i32 %add
 ; FORCE-NEXT: }
 ;
-; FORCE:      define internal i32 @foo.specialized.2(i32 %x, ptr %b) {
+; FORCE:      define internal i32 @foo.specialized.2(i32 %x, ptr %b) {{(!prof !0 )?}}{
 ; FORCE-NEXT: entry:
 ; FORCE-NEXT:   %0 = load i32, ptr @B, align 4
 ; FORCE-NEXT:   %add = add nsw i32 %x, %0

[alanzhao1]

This test does indeed to be an issue with the build config not replacing FileCheck if FileCheck is not in the same line; this test passes with the following diff:

diff --git a/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll b/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
index 8e075edaa684..6720a2f11219 100644
--- a/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
+++ b/llvm/test/Transforms/FunctionSpecialization/function-specialization3.ll
@@ -1,7 +1,5 @@
-; RUN: opt -passes="ipsccp<func-spec>" -S < %s | \
-; RUN:   FileCheck %s --check-prefixes=COMMON,DISABLED
-; RUN: opt -passes="ipsccp<func-spec>" -force-specialization -S < %s | \
-; RUN:   FileCheck %s --check-prefixes=COMMON,FORCE
+; RUN: opt -passes="ipsccp<func-spec>" -S < %s | FileCheck %s --check-prefixes=COMMON,DISABLED
+; RUN: opt -passes="ipsccp<func-spec>" -force-specialization -S < %s | FileCheck %s --check-prefixes=COMMON,FORCE

 ; Test for specializing a constant global

[mtrofin]

Could we, alternatively:

• change the test to actually have function entry counts (and branch weights)
• verify that the entry counts of the specializations are accordingly adjusted (i.e. if the original was called 1000 times, and the branch probs are, say, 20% - 80%, then the entry counts of the specializations should be 200 and 800)

[alanzhao1]

change the test to actually have function entry counts (and branch weights)

I'll create a separate PR for that as the original issue (the test failing with profile propagation unittests) is fixed with #156985