diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll | 10 | ||||
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll | 12 |
2 files changed, 21 insertions, 1 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll index fa12ee2fe92..e2c535baaa8 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll @@ -11,4 +11,14 @@ define void @ds_bpermute(i32 addrspace(1)* %out, i32 %index, i32 %src) nounwind ret void } +; CHECK-LABEL: {{^}}ds_bpermute_imm_offset: +; CHECK: ds_bpermute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:4 +; CHECK: s_waitcnt lgkmcnt +define void @ds_bpermute_imm_offset(i32 addrspace(1)* %out, i32 %base_index, i32 %src) nounwind { + %index = add i32 %base_index, 4 + %bpermute = call i32 @llvm.amdgcn.ds.bpermute(i32 %index, i32 %src) #0 + store i32 %bpermute, 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 f0ee48ff448..6d9c9419153 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll @@ -2,7 +2,7 @@ declare i32 @llvm.amdgcn.ds.permute(i32, i32) #0 -; FUNC-LABEL: {{^}}ds_permute: +; CHECK-LABEL: {{^}}ds_permute: ; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} ; CHECK: s_waitcnt lgkmcnt define void @ds_permute(i32 addrspace(1)* %out, i32 %index, i32 %src) nounwind { @@ -11,4 +11,14 @@ define void @ds_permute(i32 addrspace(1)* %out, i32 %index, i32 %src) nounwind { ret void } +; CHECK-LABEL: {{^}}ds_permute_imm_offset: +; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:4 +; CHECK: s_waitcnt lgkmcnt +define void @ds_permute_imm_offset(i32 addrspace(1)* %out, i32 %base_index, i32 %src) nounwind { + %index = add i32 %base_index, 4 + %bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #0 + store i32 %bpermute, i32 addrspace(1)* %out, align 4 + ret void +} + attributes #0 = { nounwind readnone convergent } |

