diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.bf16.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.bf16.ll | 33 |
1 files changed, 33 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.bf16.ll new file mode 100644 index 000000000000..5287b5dba848 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sqrt.bf16.ll @@ -0,0 +1,33 @@ +; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN %s +; xUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefix=GCN %s + +; FIXME: GlobalISel does not work with bf16 + +declare bfloat @llvm.amdgcn.sqrt.bf16(bfloat) #0 + +; GCN-LABEL: {{^}}sqrt_bf16: +; GCN: v_sqrt_bf16_e32 {{v[0-9]+}}, {{s[0-9]+}} +define amdgpu_kernel void @sqrt_bf16(ptr addrspace(1) %out, bfloat %src) #1 { + %sqrt = call bfloat @llvm.amdgcn.sqrt.bf16(bfloat %src) #0 + store bfloat %sqrt, ptr addrspace(1) %out, align 2 + ret void +} + +; GCN-LABEL: {{^}}sqrt_bf16_constant_4 +; GCN: v_sqrt_bf16_e32 v0, 4.0 +define amdgpu_kernel void @sqrt_bf16_constant_4(ptr addrspace(1) %out) #1 { + %sqrt = call bfloat @llvm.amdgcn.sqrt.bf16(bfloat 4.0) #0 + store bfloat %sqrt, ptr addrspace(1) %out, align 2 + ret void +} + +; GCN-LABEL: {{^}}sqrt_bf16_constant_100 +; GCN: v_sqrt_bf16_e32 {{v[0-9]+}}, 0x42c8 +define amdgpu_kernel void @sqrt_bf16_constant_100(ptr addrspace(1) %out) #1 { + %sqrt = call bfloat @llvm.amdgcn.sqrt.bf16(bfloat 100.0) #0 + store bfloat %sqrt, ptr addrspace(1) %out, align 2 + ret void +} + +attributes #0 = { nounwind readnone } +attributes #1 = { nounwind } |