diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-07-19 23:16:53 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2016-07-19 23:16:53 +0000 |
| commit | a1fe17c9adb2b6093f1ce848a48fb8954c27c595 (patch) | |
| tree | 2fcb8b6fcd3f50a1c72634d2808ad3fdc7206d90 /llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll | |
| parent | 1986030b62601d8cd6d74cfc083e4638be3d8b46 (diff) | |
| download | bcm5719-llvm-a1fe17c9adb2b6093f1ce848a48fb8954c27c595.tar.gz bcm5719-llvm-a1fe17c9adb2b6093f1ce848a48fb8954c27c595.zip | |
AMDGPU: Change fdiv lowering based on !fpmath metadata
If 2.5 ulp is acceptable, denormals are not required, and
isn't a reciprocal which will already be handled, replace
with a faster fdiv.
Simplify the lowering tests by using per function
subtarget features.
llvm-svn: 276051
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll | 18 |
1 files changed, 18 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll new file mode 100644 index 00000000000..54d7848da3b --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdiv.fast.ll @@ -0,0 +1,18 @@ +; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck %s + +declare float @llvm.amdgcn.fdiv.fast(float, float) #0 + +; CHECK-LABEL: {{^}}test_fdiv_fast: +; CHECK: v_cndmask_b32_e32 v{{[0-9]+}}, 1.0, v{{[0-9]+}}, vcc +; CHECK: v_mul_f32_e32 +; CHECK: v_rcp_f32_e32 +; CHECK: v_mul_f32_e32 +; CHECK: v_mul_f32_e32 +define void @test_fdiv_fast(float addrspace(1)* %out, float %a, float %b) #1 { + %fdiv = call float @llvm.amdgcn.fdiv.fast(float %a, float %b) + store float %fdiv, float addrspace(1)* %out + ret void +} + +attributes #0 = { nounwind readnone } +attributes #1 = { nounwind } |
