diff options
| author | Tom Stellard <thomas.stellard@amd.com> | 2015-06-13 03:28:10 +0000 |
|---|---|---|
| committer | Tom Stellard <thomas.stellard@amd.com> | 2015-06-13 03:28:10 +0000 |
| commit | 45bb48ea197fe496865387120c7c55b56f0717d6 (patch) | |
| tree | 3914fb8c4ace9ea4dee024fc944d51dc45adf401 /llvm/test/CodeGen/AMDGPU/mad_int24.ll | |
| parent | 8fa9677d4eb1cd43973eb59acc12b0534691d604 (diff) | |
| download | bcm5719-llvm-45bb48ea197fe496865387120c7c55b56f0717d6.tar.gz bcm5719-llvm-45bb48ea197fe496865387120c7c55b56f0717d6.zip | |
R600 -> AMDGPU rename
llvm-svn: 239657
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/mad_int24.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/mad_int24.ll | 35 |
1 files changed, 35 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/mad_int24.ll b/llvm/test/CodeGen/AMDGPU/mad_int24.ll new file mode 100644 index 00000000000..86d75a63ca4 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/mad_int24.ll @@ -0,0 +1,35 @@ +; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s --check-prefix=EG --check-prefix=FUNC +; RUN: llc < %s -march=r600 -mcpu=cayman | FileCheck %s --check-prefix=CM --check-prefix=FUNC +; RUN: llc < %s -march=amdgcn -mcpu=SI -verify-machineinstrs | FileCheck %s --check-prefix=SI --check-prefix=FUNC +; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s --check-prefix=SI --check-prefix=FUNC + +declare i32 @llvm.AMDGPU.imul24(i32, i32) nounwind readnone + +; FUNC-LABEL: {{^}}i32_mad24: +; Signed 24-bit multiply is not supported on pre-Cayman GPUs. +; EG: MULLO_INT +; Make sure we aren't masking the inputs. +; CM-NOT: AND +; CM: MULADD_INT24 +; SI-NOT: and +; SI: v_mad_i32_i24 +define void @i32_mad24(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c) { +entry: + %0 = shl i32 %a, 8 + %a_24 = ashr i32 %0, 8 + %1 = shl i32 %b, 8 + %b_24 = ashr i32 %1, 8 + %2 = mul i32 %a_24, %b_24 + %3 = add i32 %2, %c + store i32 %3, i32 addrspace(1)* %out + ret void +} + +; FUNC-LABEL: @test_imul24 +; SI: v_mad_i32_i24 +define void @test_imul24(i32 addrspace(1)* %out, i32 %src0, i32 %src1, i32 %src2) nounwind { + %mul = call i32 @llvm.AMDGPU.imul24(i32 %src0, i32 %src1) nounwind readnone + %add = add i32 %mul, %src2 + store i32 %add, i32 addrspace(1)* %out, align 4 + ret void +} |

