1 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s 2 ; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s 3 4 declare float @llvm.amdgcn.cubeid(float, float, float) #0 5 declare float @llvm.amdgcn.cubesc(float, float, float) #0 6 declare float @llvm.amdgcn.cubetc(float, float, float) #0 7 declare float @llvm.amdgcn.cubema(float, float, float) #0 8 9 ; GCN-LABEL: {{^}}cube: 10 ; GCN-DAG: v_cubeid_f32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} 11 ; GCN-DAG: v_cubesc_f32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} 12 ; GCN-DAG: v_cubetc_f32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} 13 ; GCN-DAG: v_cubema_f32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} 14 ; GCN: _store_dwordx4 15 define amdgpu_kernel void @cube(<4 x float> addrspace(1)* %out, float %a, float %b, float %c) #1 { 16 %cubeid = call float @llvm.amdgcn.cubeid(float %a, float %b, float %c) 17 %cubesc = call float @llvm.amdgcn.cubesc(float %a, float %b, float %c) 18 %cubetc = call float @llvm.amdgcn.cubetc(float %a, float %b, float %c) 19 %cubema = call float @llvm.amdgcn.cubema(float %a, float %b, float %c) 20 21 %vec0 = insertelement <4 x float> undef, float %cubeid, i32 0 22 %vec1 = insertelement <4 x float> %vec0, float %cubesc, i32 1 23 %vec2 = insertelement <4 x float> %vec1, float %cubetc, i32 2 24 %vec3 = insertelement <4 x float> %vec2, float %cubema, i32 3 25 store <4 x float> %vec3, <4 x float> addrspace(1)* %out 26 ret void 27 } 28 29 attributes #0 = { nounwind readnone } 30 attributes #1 = { nounwind } 31
