[PATCH] D158600: [InferAlignment] Enable InferAlignment pass by default

[Dhruv Chawla via Phabricator via llvm-commits]

0xdc03 added inline comments.


================
Comment at: llvm/test/Transforms/PhaseOrdering/X86/excessive-unrolling.ll:135
 ; CHECK-NEXT:    store <2 x double> [[TMP29]], ptr getelementptr inbounds ([58 x double], ptr @a, i64 1, i64 0), align 16
-; CHECK-NEXT:    [[TMP30:%.*]] = load double, ptr getelementptr inbounds ([58 x double], ptr @b, i64 1, i64 2), align 16
-; CHECK-NEXT:    [[TMP31:%.*]] = load double, ptr getelementptr inbounds ([58 x double], ptr @c, i64 1, i64 2), align 16
+; CHECK-NEXT:    [[TMP30:%.*]] = load double, ptr getelementptr inbounds ([58 x double], ptr @b, i64 1, i64 2), align 8
+; CHECK-NEXT:    [[TMP31:%.*]] = load double, ptr getelementptr inbounds ([58 x double], ptr @c, i64 1, i64 2), align 8
----------------
aeubanks wrote:
> any idea why this regressed?
Loop unrolling causes generation of instructions after InferAlignment runs, so this is not caught by it.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D158600/new/

https://reviews.llvm.org/D158600