diff options
| author | Changpeng Fang <changpeng.fang@gmail.com> | 2016-03-17 16:43:50 +0000 |
|---|---|---|
| committer | Changpeng Fang <changpeng.fang@gmail.com> | 2016-03-17 16:43:50 +0000 |
| commit | 234fcb81d3e48dd46fc6f610df77511883b69fd8 (patch) | |
| tree | 0dee865863b8c485e3542afc6918e51902649634 /llvm/test | |
| parent | 79cad857a0e8fddccad46fcbc81c0ea66da7cfb7 (diff) | |
| download | bcm5719-llvm-234fcb81d3e48dd46fc6f610df77511883b69fd8.tar.gz bcm5719-llvm-234fcb81d3e48dd46fc6f610df77511883b69fd8.zip | |
AMDGPU/SI: Do not generate s_waitcnt after ds_permute/ds_bpermute
Symmary:
ds_permute/ds_bpermute do not read memory so s_waitcnt is not needed.
Reviewers
arsenm, tstellarAMD
Subscribers
llvm-commits, arsenm
Differential Revision:
http://reviews.llvm.org/D18197
llvm-svn: 263720
Diffstat (limited to 'llvm/test')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll | 24 | ||||
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll | 28 |
2 files changed, 50 insertions, 2 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll index 45cd45d830e..d9c87b13604 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll @@ -10,4 +10,28 @@ define void @ds_bpermute(i32 addrspace(1)* %out, i32 %index, i32 %src) nounwind ret void } +; FUNC-LABEL: {{^}}bpermute_no_waitcnt_test: +; CHECK: s_cbranch_scc1 +; CHECK: ds_bpermute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; CHECK-NOT: s_waitcnt +define void @bpermute_no_waitcnt_test(i32 addrspace(1)* %out, i32 %cond) { +entry: + + %tmp = icmp eq i32 %cond, 0 + br i1 %tmp, label %if, label %else + +if: ; preds = %entry + + %bpermute = call i32 @llvm.amdgcn.ds.bpermute(i32 0, i32 0) #0 + br label %endif + +else: ; preds = %entry + br label %endif + +endif: + %val = phi i32 [ %bpermute, %if ], [0, %else] ; preds = %else, %if + store i32 %val, i32 addrspace(1)* %out, align 4 + ret void +} + attributes #0 = { nounwind readnone convergent } diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll index e2172868ed9..2f774abd1c6 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll @@ -5,9 +5,33 @@ declare i32 @llvm.amdgcn.ds.permute(i32, i32) #0 ; FUNC-LABEL: {{^}}ds_permute: ; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} define void @ds_permute(i32 addrspace(1)* %out, i32 %index, i32 %src) nounwind { - %bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #0 - store i32 %bpermute, i32 addrspace(1)* %out, align 4 + %permute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #0 + store i32 %permute, i32 addrspace(1)* %out, align 4 ret void } +; FUNC-LABEL: {{^}}permute_no_waitcnt_test: +; CHECK: s_cbranch_scc1 +; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; CHECK-NOT: s_waitcnt +define void @permute_no_waitcnt_test(i32 addrspace(1)* %out, i32 %cond) { +entry: + + %tmp = icmp eq i32 %cond, 0 + br i1 %tmp, label %if, label %else + +if: ; preds = %entry + %permute = call i32 @llvm.amdgcn.ds.permute(i32 0, i32 0) #0 + br label %endif + +else: ; preds = %entry + br label %endif + +endif: + %val = phi i32 [ %permute, %if ], [0, %else] ; preds = %else, %if + store i32 %val, i32 addrspace(1)* %out, align 4 + ret void +} + + attributes #0 = { nounwind readnone convergent } |
