[PATCH] D33576: [AMDGPU] Require waitcnt before barrier for all targets; adjust tests.
Mark Searles via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu May 25 18:10:09 PDT 2017
msearles updated this revision to Diff 100350.
msearles added a comment.
Doh! Add lib/Target/AMDGPU/AMDGPUSubtarget.h back into the diff.
https://reviews.llvm.org/D33576
Files:
lib/Target/AMDGPU/AMDGPUSubtarget.h
test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.ll
Index: test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.ll
===================================================================
--- test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.ll
+++ test/CodeGen/AMDGPU/llvm.amdgcn.s.barrier.ll
@@ -3,9 +3,8 @@
; GCN-LABEL: {{^}}test_barrier:
; GFX8: buffer_store_dword
-; GFX8: s_waitcnt
; GFX9: flat_store_dword
-; GFX9-NOT: s_waitcnt
+; GCN: s_waitcnt
; GCN: s_barrier
define amdgpu_kernel void @test_barrier(i32 addrspace(1)* %out, i32 %size) #0 {
entry:
Index: lib/Target/AMDGPU/AMDGPUSubtarget.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -730,7 +730,7 @@
/// \returns True if waitcnt instruction is needed before barrier instruction,
/// false otherwise.
bool needWaitcntBeforeBarrier() const {
- return getGeneration() < GFX9;
+ return true;
}
/// \returns true if the flat_scratch register should be initialized with the
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D33576.100350.patch
Type: text/x-patch
Size: 990 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170526/3d913be6/attachment-0001.bin>
More information about the llvm-commits
mailing list