diff options
| author | Sam Kolton <Sam.Kolton@amd.com> | 2017-04-06 15:03:28 +0000 |
|---|---|---|
| committer | Sam Kolton <Sam.Kolton@amd.com> | 2017-04-06 15:03:28 +0000 |
| commit | 9fa169601f3cf42a32305ecf26024fc9c149c98c (patch) | |
| tree | 2301f35de20e5fd56182466cb403cd4e0f15cab6 /llvm/test/CodeGen/AMDGPU/max.i16.ll | |
| parent | da9e718fb4911cc4cf900b8428b4c7da5d19a47a (diff) | |
| download | bcm5719-llvm-9fa169601f3cf42a32305ecf26024fc9c149c98c.tar.gz bcm5719-llvm-9fa169601f3cf42a32305ecf26024fc9c149c98c.zip | |
[AMDGPU] Resubmit SDWA peephole: enable by default
Reviewers: vpykhtin, rampitec, arsenm
Subscribers: qcolombet, kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Differential Revision: https://reviews.llvm.org/D31671
llvm-svn: 299654
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/max.i16.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/max.i16.ll | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/max.i16.ll b/llvm/test/CodeGen/AMDGPU/max.i16.ll index 8fcf85b826f..abd75258c4d 100644 --- a/llvm/test/CodeGen/AMDGPU/max.i16.ll +++ b/llvm/test/CodeGen/AMDGPU/max.i16.ll @@ -20,7 +20,7 @@ define amdgpu_kernel void @v_test_imax_sge_i16(i16 addrspace(1)* %out, i16 addrs ; FIXME: Need to handle non-uniform case for function below (load without gep). ; GCN-LABEL: {{^}}v_test_imax_sge_v2i16: ; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} -; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; VI: v_max_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX9: v_pk_max_i16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} define amdgpu_kernel void @v_test_imax_sge_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %aptr, <2 x i16> addrspace(1)* %bptr) nounwind { @@ -39,7 +39,7 @@ define amdgpu_kernel void @v_test_imax_sge_v2i16(<2 x i16> addrspace(1)* %out, < ; FIXME: Need to handle non-uniform case for function below (load without gep). ; GCN-LABEL: {{^}}v_test_imax_sge_v3i16: ; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} -; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; VI: v_max_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} ; VI-NOT: v_max_i16 @@ -61,9 +61,9 @@ define amdgpu_kernel void @v_test_imax_sge_v3i16(<3 x i16> addrspace(1)* %out, < ; FIXME: Need to handle non-uniform case for function below (load without gep). ; GCN-LABEL: {{^}}v_test_imax_sge_v4i16: ; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; VI: v_max_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} -; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} -; VI: v_max_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} +; VI: v_max_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 ; GFX9: v_pk_max_i16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} ; GFX9: v_pk_max_i16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} @@ -130,7 +130,7 @@ define amdgpu_kernel void @v_test_umax_ugt_i16(i16 addrspace(1)* %out, i16 addrs ; GCN-LABEL: {{^}}v_test_umax_ugt_v2i16: ; VI: v_max_u16_e32 -; VI: v_max_u16_e32 +; VI: v_max_u16_sdwa ; GFX9: v_pk_max_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} define amdgpu_kernel void @v_test_umax_ugt_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %aptr, <2 x i16> addrspace(1)* %bptr) nounwind { |
