diff options
| author | Mehdi Amini <mehdi.amini@apple.com> | 2017-01-06 21:06:51 +0000 |
|---|---|---|
| committer | Mehdi Amini <mehdi.amini@apple.com> | 2017-01-06 21:06:51 +0000 |
| commit | 27d224fbbb5185bb84df5535ff8682560796ee18 (patch) | |
| tree | b15f5c6dc2dd9ef0917306bf26fa019fd55532d8 /llvm/test/Transforms/LoopLoadElim/forward.ll | |
| parent | 218ecacbf10a238ce2f71d620fa510c9aada0e73 (diff) | |
| download | bcm5719-llvm-27d224fbbb5185bb84df5535ff8682560796ee18.tar.gz bcm5719-llvm-27d224fbbb5185bb84df5535ff8682560796ee18.zip | |
Fix LoopLoadElimination to keep original alignment on the inital hoisted store
This is fixing a bug where Loop Vectorization is widening a load but
with a lower alignment. Hoisting the load without propagating the alignment
will allow inst-combine to later deduce a higher alignment that what the pointer
actually is.
Differential Revision: https://reviews.llvm.org/D28408
llvm-svn: 291281
Diffstat (limited to 'llvm/test/Transforms/LoopLoadElim/forward.ll')
| -rw-r--r-- | llvm/test/Transforms/LoopLoadElim/forward.ll | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/llvm/test/Transforms/LoopLoadElim/forward.ll b/llvm/test/Transforms/LoopLoadElim/forward.ll index ed0d162ab7e..9a0e03a317c 100644 --- a/llvm/test/Transforms/LoopLoadElim/forward.ll +++ b/llvm/test/Transforms/LoopLoadElim/forward.ll @@ -16,8 +16,8 @@ define void @f(i32* %A, i32* %B, i32* %C, i64 %N) { ; CHECK-NOT: %found.conflict{{.*}} = entry: -; for.body.ph: -; CHECK: %load_initial = load i32, i32* %A +; Make sure the hoisted load keeps the alignment +; CHECK: %load_initial = load i32, i32* %A, align 1 br label %for.body for.body: ; preds = %for.body, %entry @@ -34,7 +34,7 @@ for.body: ; preds = %for.body, %entry %a_p1 = add i32 %b, 2 store i32 %a_p1, i32* %Aidx_next, align 4 - %a = load i32, i32* %Aidx, align 4 + %a = load i32, i32* %Aidx, align 1 ; CHECK: %c = mul i32 %store_forwarded, 2 %c = mul i32 %a, 2 store i32 %c, i32* %Cidx, align 4 |
