diff options
Diffstat (limited to 'clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl')
| -rw-r--r-- | clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl | 12 |
1 files changed, 12 insertions, 0 deletions
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl index e4ef3de..86c27d4 100644 --- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl +++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250-wmma-w32.cl @@ -157,6 +157,18 @@ void test_amdgcn_wmma_i32_16x16x64_iu8(global v8i* out, v8i a, v8i b, v8i c) *out = __builtin_amdgcn_wmma_i32_16x16x64_iu8(0, a, 0, b, c, false, true); } +// CHECK-GFX1250-LABEL: @test_amdgcn_wmma_f32_16x16x128_f8f6f4( +// CHECK-GFX1250-NEXT: entry: +// CHECK-GFX1250-NEXT: [[TMP0:%.*]] = shufflevector <16 x i32> [[B:%.*]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11> +// CHECK-GFX1250-NEXT: [[TMP1:%.*]] = tail call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A:%.*]], i32 2, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C:%.*]]) +// CHECK-GFX1250-NEXT: store <8 x float> [[TMP1]], ptr addrspace(1) [[OUT:%.*]], align 32, !tbaa [[TBAA4]] +// CHECK-GFX1250-NEXT: ret void +// +void test_amdgcn_wmma_f32_16x16x128_f8f6f4(global v8f* out, v16i a, v16i b, v8f c) +{ + *out = __builtin_amdgcn_wmma_f32_16x16x128_f8f6f4(1, a, 2, b, 0, c); +} + // CHECK-GFX1250-LABEL: @test_amdgcn_wmma_f32_16x16x32_f16( // CHECK-GFX1250-NEXT: entry: // CHECK-GFX1250-NEXT: [[TMP0:%.*]] = tail call <8 x float> @llvm.amdgcn.wmma.f32.16x16x32.f16.v8f32.v16f16(i1 false, <16 x half> [[A:%.*]], i1 false, <16 x half> [[B:%.*]], i16 0, <8 x float> [[C:%.*]], i1 false, i1 true) |