1; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,FUNC %s
2; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,FUNC %s
3; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,FUNC %s
4; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck --check-prefixes=EG,FUNC %s
5
6
7; FUNC-LABEL: {{^}}constant_load_i64:
8; GCN: s_load_dwordx2 {{s\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0x0{{$}}
9; EG: VTX_READ_64
10define amdgpu_kernel void @constant_load_i64(i64 addrspace(1)* %out, i64 addrspace(4)* %in) #0 {
11  %ld = load i64, i64 addrspace(4)* %in
12  store i64 %ld, i64 addrspace(1)* %out
13  ret void
14}
15
16; FUNC-LABEL: {{^}}constant_load_v2i64:
17; GCN: s_load_dwordx4
18
19; EG: VTX_READ_128
20define amdgpu_kernel void @constant_load_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> addrspace(4)* %in) #0 {
21entry:
22  %ld = load <2 x i64>, <2 x i64> addrspace(4)* %in
23  store <2 x i64> %ld, <2 x i64> addrspace(1)* %out
24  ret void
25}
26
27; FUNC-LABEL: {{^}}constant_load_v3i64:
28; GCN-DAG: s_load_dwordx4 {{s\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0x0{{$}}
29; GCN-DAG: s_load_dwordx2 {{s\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0x{{[0-9]+}}{{$}}
30
31; EG-DAG: VTX_READ_128
32; EG-DAG: VTX_READ_128
33define amdgpu_kernel void @constant_load_v3i64(<3 x i64> addrspace(1)* %out, <3 x i64> addrspace(4)* %in) #0 {
34entry:
35  %ld = load <3 x i64>, <3 x i64> addrspace(4)* %in
36  store <3 x i64> %ld, <3 x i64> addrspace(1)* %out
37  ret void
38}
39
40; FUNC-LABEL: {{^}}constant_load_v4i64
41; GCN: s_load_dwordx8
42
43; EG: VTX_READ_128
44; EG: VTX_READ_128
45define amdgpu_kernel void @constant_load_v4i64(<4 x i64> addrspace(1)* %out, <4 x i64> addrspace(4)* %in) #0 {
46entry:
47  %ld = load <4 x i64>, <4 x i64> addrspace(4)* %in
48  store <4 x i64> %ld, <4 x i64> addrspace(1)* %out
49  ret void
50}
51
52; FUNC-LABEL: {{^}}constant_load_v8i64:
53; GCN: s_load_dwordx16
54
55; EG: VTX_READ_128
56; EG: VTX_READ_128
57; EG: VTX_READ_128
58; EG: VTX_READ_128
59define amdgpu_kernel void @constant_load_v8i64(<8 x i64> addrspace(1)* %out, <8 x i64> addrspace(4)* %in) #0 {
60entry:
61  %ld = load <8 x i64>, <8 x i64> addrspace(4)* %in
62  store <8 x i64> %ld, <8 x i64> addrspace(1)* %out
63  ret void
64}
65
66; FUNC-LABEL: {{^}}constant_load_v16i64:
67; GCN: s_load_dwordx16
68; GCN: s_load_dwordx16
69
70; EG: VTX_READ_128
71; EG: VTX_READ_128
72; EG: VTX_READ_128
73; EG: VTX_READ_128
74; EG: VTX_READ_128
75; EG: VTX_READ_128
76; EG: VTX_READ_128
77; EG: VTX_READ_128
78define amdgpu_kernel void @constant_load_v16i64(<16 x i64> addrspace(1)* %out, <16 x i64> addrspace(4)* %in) #0 {
79entry:
80  %ld = load <16 x i64>, <16 x i64> addrspace(4)* %in
81  store <16 x i64> %ld, <16 x i64> addrspace(1)* %out
82  ret void
83}
84
85attributes #0 = { nounwind }
86