From 2aed6ca1d33468f7e6f8509213a1b9dcd8026ec1 Mon Sep 17 00:00:00 2001 From: Matt Arsenault Date: Sat, 19 Dec 2015 01:46:41 +0000 Subject: AMDGPU: Switch barrier intrinsics to using convergent noduplicate prevents unrolling of small loops that happen to have barriers in them. If a loop has a barrier in it, it is OK to duplicate it for the unroll. llvm-svn: 256075 --- llvm/test/CodeGen/AMDGPU/ds_write2st64.ll | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'llvm/test/CodeGen/AMDGPU/ds_write2st64.ll') diff --git a/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll b/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll index 358aa6a9e36..5a1024ccf6d 100644 --- a/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll +++ b/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll @@ -109,9 +109,9 @@ declare i32 @llvm.r600.read.tidig.x() #1 ; Function Attrs: nounwind readnone declare i32 @llvm.r600.read.tidig.y() #1 -; Function Attrs: noduplicate nounwind +; Function Attrs: convergent nounwind declare void @llvm.AMDGPU.barrier.local() #2 attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" } attributes #1 = { nounwind readnone } -attributes #2 = { noduplicate nounwind } +attributes #2 = { convergent nounwind } -- cgit v1.2.3