[PATCH] D29210: AMDGPU: Fix atomic_inc/atomic_dec not being divergent

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
Herald added a reviewer: tstellarAMD.
Herald added subscribers: tony-tye, yaxunl, nhaehnle, wdng, kzhuravl.

https://reviews.llvm.org/D29210

Files:
  lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
  test/Analysis/DivergenceAnalysis/AMDGPU/atomics.ll


Index: test/Analysis/DivergenceAnalysis/AMDGPU/atomics.ll
===================================================================
--- test/Analysis/DivergenceAnalysis/AMDGPU/atomics.ll
+++ test/Analysis/DivergenceAnalysis/AMDGPU/atomics.ll
@@ -12,4 +12,34 @@
   ret {i32, i1} %orig
 }
 
+; CHECK: DIVERGENT: %ret = call i32 @llvm.amdgcn.atomic.inc.i32.p1i32(i32 addrspace(1)* %ptr, i32 %val)
+define i32 @test_atomic_inc_i32(i32 addrspace(1)* %ptr, i32 %val) #0 {
+  %ret = call i32 @llvm.amdgcn.atomic.inc.i32.p1i32(i32 addrspace(1)* %ptr, i32 %val)
+  ret i32 %ret
+}
+
+; CHECK: DIVERGENT: %ret = call i64 @llvm.amdgcn.atomic.inc.i64.p1i64(i64 addrspace(1)* %ptr, i64 %val)
+define i64 @test_atomic_inc_i64(i64 addrspace(1)* %ptr, i64 %val) #0 {
+  %ret = call i64 @llvm.amdgcn.atomic.inc.i64.p1i64(i64 addrspace(1)* %ptr, i64 %val)
+  ret i64 %ret
+}
+
+; CHECK: DIVERGENT: %ret = call i32 @llvm.amdgcn.atomic.dec.i32.p1i32(i32 addrspace(1)* %ptr, i32 %val)
+define i32 @test_atomic_dec_i32(i32 addrspace(1)* %ptr, i32 %val) #0 {
+  %ret = call i32 @llvm.amdgcn.atomic.dec.i32.p1i32(i32 addrspace(1)* %ptr, i32 %val)
+  ret i32 %ret
+}
+
+; CHECK: DIVERGENT: %ret = call i64 @llvm.amdgcn.atomic.dec.i64.p1i64(i64 addrspace(1)* %ptr, i64 %val)
+define i64 @test_atomic_dec_i64(i64 addrspace(1)* %ptr, i64 %val) #0 {
+  %ret = call i64 @llvm.amdgcn.atomic.dec.i64.p1i64(i64 addrspace(1)* %ptr, i64 %val)
+  ret i64 %ret
+}
+
+declare i32 @llvm.amdgcn.atomic.inc.i32.p1i32(i32 addrspace(1)* nocapture, i32) #1
+declare i64 @llvm.amdgcn.atomic.inc.i64.p1i64(i64 addrspace(1)* nocapture, i64) #1
+declare i32 @llvm.amdgcn.atomic.dec.i32.p1i32(i32 addrspace(1)* nocapture, i32) #1
+declare i64 @llvm.amdgcn.atomic.dec.i64.p1i64(i64 addrspace(1)* nocapture, i64) #1
+
 attributes #0 = { nounwind }
+attributes #1 = { nounwind argmemonly }
Index: lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -249,6 +249,8 @@
   case Intrinsic::r600_read_tidig_x:
   case Intrinsic::r600_read_tidig_y:
   case Intrinsic::r600_read_tidig_z:
+  case Intrinsic::amdgcn_atomic_inc:
+  case Intrinsic::amdgcn_atomic_dec:
   case Intrinsic::amdgcn_image_atomic_swap:
   case Intrinsic::amdgcn_image_atomic_add:
   case Intrinsic::amdgcn_image_atomic_sub:


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D29210.86003.patch
Type: text/x-patch
Size: 2402 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170127/08c88a98/attachment.bin>