1 ; RUN: llc -march=r600 -mcpu=SI -mattr=-fp32-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI-UNSAFE -check-prefix=SI %s
2 ; RUN: llc -march=r600 -mcpu=SI -mattr=-fp32-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI %s
4 declare float @llvm.sqrt.f32(float) nounwind readnone
5 declare double @llvm.sqrt.f64(double) nounwind readnone
7 ; SI-LABEL: {{^}}rsq_f32:
10 define void @rsq_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
11 %val = load float addrspace(1)* %in, align 4
12 %sqrt = call float @llvm.sqrt.f32(float %val) nounwind readnone
13 %div = fdiv float 1.0, %sqrt
14 store float %div, float addrspace(1)* %out, align 4
18 ; SI-LABEL: {{^}}rsq_f64:
19 ; SI-UNSAFE: V_RSQ_F64_e32
20 ; SI-SAFE: V_SQRT_F64_e32
22 define void @rsq_f64(double addrspace(1)* noalias %out, double addrspace(1)* noalias %in) nounwind {
23 %val = load double addrspace(1)* %in, align 4
24 %sqrt = call double @llvm.sqrt.f64(double %val) nounwind readnone
25 %div = fdiv double 1.0, %sqrt
26 store double %div, double addrspace(1)* %out, align 4
30 ; SI-LABEL: {{^}}rsq_f32_sgpr:
31 ; SI: V_RSQ_F32_e32 {{v[0-9]+}}, {{s[0-9]+}}
33 define void @rsq_f32_sgpr(float addrspace(1)* noalias %out, float %val) nounwind {
34 %sqrt = call float @llvm.sqrt.f32(float %val) nounwind readnone
35 %div = fdiv float 1.0, %sqrt
36 store float %div, float addrspace(1)* %out, align 4