1 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
2 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9 %s
4 declare float @llvm.amdgcn.ds.fadd(float addrspace(3)* nocapture, float, i32, i32, i1)
5 declare float @llvm.amdgcn.ds.fmin(float addrspace(3)* nocapture, float, i32, i32, i1)
6 declare float @llvm.amdgcn.ds.fmax(float addrspace(3)* nocapture, float, i32, i32, i1)
8 ; GCN-LABEL: {{^}}lds_ds_fadd:
11 ; GCN-DAG: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
12 ; GCN: ds_add_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]] offset:32
13 ; GCN: ds_add_f32 [[V3:v[0-9]+]], [[V0]] offset:64
14 ; GCN: s_waitcnt lgkmcnt(1)
15 ; GCN: ds_add_rtn_f32 {{v[0-9]+}}, {{v[0-9]+}}, [[V2]]
16 define amdgpu_kernel void @lds_ds_fadd(float addrspace(1)* %out, float addrspace(3)* %ptrf, i32 %idx) {
17 %idx.add = add nuw i32 %idx, 4
18 %shl0 = shl i32 %idx.add, 3
19 %shl1 = shl i32 %idx.add, 4
20 %ptr0 = inttoptr i32 %shl0 to float addrspace(3)*
21 %ptr1 = inttoptr i32 %shl1 to float addrspace(3)*
22 %a1 = call float @llvm.amdgcn.ds.fadd(float addrspace(3)* %ptr0, float 4.2e+1, i32 0, i32 0, i1 false)
23 %a2 = call float @llvm.amdgcn.ds.fadd(float addrspace(3)* %ptr1, float 4.2e+1, i32 0, i32 0, i1 false)
24 %a3 = call float @llvm.amdgcn.ds.fadd(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
25 store float %a3, float addrspace(1)* %out
29 ; GCN-LABEL: {{^}}lds_ds_fmin:
30 ; VI-DAG: s_mov_b32 m0
32 ; GCN-DAG: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
33 ; GCN: ds_min_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]] offset:32
34 ; GCN: ds_min_f32 [[V3:v[0-9]+]], [[V0]] offset:64
35 ; GCN: s_waitcnt lgkmcnt(1)
36 ; GCN: ds_min_rtn_f32 {{v[0-9]+}}, {{v[0-9]+}}, [[V2]]
37 define amdgpu_kernel void @lds_ds_fmin(float addrspace(1)* %out, float addrspace(3)* %ptrf, i32 %idx) {
38 %idx.add = add nuw i32 %idx, 4
39 %shl0 = shl i32 %idx.add, 3
40 %shl1 = shl i32 %idx.add, 4
41 %ptr0 = inttoptr i32 %shl0 to float addrspace(3)*
42 %ptr1 = inttoptr i32 %shl1 to float addrspace(3)*
43 %a1 = call float @llvm.amdgcn.ds.fmin(float addrspace(3)* %ptr0, float 4.2e+1, i32 0, i32 0, i1 false)
44 %a2 = call float @llvm.amdgcn.ds.fmin(float addrspace(3)* %ptr1, float 4.2e+1, i32 0, i32 0, i1 false)
45 %a3 = call float @llvm.amdgcn.ds.fmin(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
46 store float %a3, float addrspace(1)* %out
50 ; GCN-LABEL: {{^}}lds_ds_fmax:
51 ; VI-DAG: s_mov_b32 m0
53 ; GCN-DAG: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
54 ; GCN: ds_max_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]] offset:32
55 ; GCN: ds_max_f32 [[V3:v[0-9]+]], [[V0]] offset:64
56 ; GCN: s_waitcnt lgkmcnt(1)
57 ; GCN: ds_max_rtn_f32 {{v[0-9]+}}, {{v[0-9]+}}, [[V2]]
58 define amdgpu_kernel void @lds_ds_fmax(float addrspace(1)* %out, float addrspace(3)* %ptrf, i32 %idx) {
59 %idx.add = add nuw i32 %idx, 4
60 %shl0 = shl i32 %idx.add, 3
61 %shl1 = shl i32 %idx.add, 4
62 %ptr0 = inttoptr i32 %shl0 to float addrspace(3)*
63 %ptr1 = inttoptr i32 %shl1 to float addrspace(3)*
64 %a1 = call float @llvm.amdgcn.ds.fmax(float addrspace(3)* %ptr0, float 4.2e+1, i32 0, i32 0, i1 false)
65 %a2 = call float @llvm.amdgcn.ds.fmax(float addrspace(3)* %ptr1, float 4.2e+1, i32 0, i32 0, i1 false)
66 %a3 = call float @llvm.amdgcn.ds.fmax(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
67 store float %a3, float addrspace(1)* %out