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
diff --git a/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll b/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll
index 358aa6a..5a1024c 100644
--- a/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll
+++ b/llvm/test/CodeGen/AMDGPU/ds_write2st64.ll
@@ -109,9 +109,9 @@
; 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 }