121
|
1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
|
|
2
|
|
3 ; GCN-LABEL: {{^}}fold_mul_neg:
|
|
4 ; GCN: load_dword [[V:v[0-9]+]]
|
|
5 ; GCN: v_or_b32_e32 [[NEG:v[0-9]]], 0x80000000, [[V]]
|
|
6 ; GCN: store_dword [[NEG]]
|
|
7
|
|
8 define amdgpu_kernel void @fold_mul_neg(float addrspace(1)* %arg) {
|
|
9 %tid = tail call i32 @llvm.amdgcn.workitem.id.x()
|
|
10 %gep = getelementptr inbounds float, float addrspace(1)* %arg, i32 %tid
|
|
11 %v = load float, float addrspace(1)* %gep, align 4
|
|
12 %cmp = fcmp fast ogt float %v, 0.000000e+00
|
|
13 %sel = select i1 %cmp, float -1.000000e+00, float 1.000000e+00
|
|
14 %mul = fmul fast float %v, %sel
|
|
15 store float %mul, float addrspace(1)* %gep, align 4
|
|
16 ret void
|
|
17 }
|
|
18
|
|
19 ; GCN-LABEL: {{^}}fold_mul_abs:
|
|
20 ; GCN: load_dword [[V:v[0-9]+]]
|
|
21 ; GCN: v_and_b32_e32 [[ABS:v[0-9]]], 0x7fffffff, [[V]]
|
|
22 ; GCN: store_dword [[ABS]]
|
|
23
|
|
24 define amdgpu_kernel void @fold_mul_abs(float addrspace(1)* %arg) {
|
|
25 %tid = tail call i32 @llvm.amdgcn.workitem.id.x()
|
|
26 %gep = getelementptr inbounds float, float addrspace(1)* %arg, i32 %tid
|
|
27 %v = load float, float addrspace(1)* %gep, align 4
|
|
28 %cmp = fcmp fast olt float %v, 0.000000e+00
|
|
29 %sel = select i1 %cmp, float -1.000000e+00, float 1.000000e+00
|
|
30 %mul = fmul fast float %v, %sel
|
|
31 store float %mul, float addrspace(1)* %gep, align 4
|
|
32 ret void
|
|
33 }
|
|
34
|
|
35 declare i32 @llvm.amdgcn.workitem.id.x() #0
|
|
36
|
|
37 attributes #0 = { nounwind readnone speculatable }
|