1; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
2; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s
3; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
4; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
5
6
7@b = internal addrspace(4) constant [1 x i16] [ i16 7 ], align 2
8
9@float_gv = internal unnamed_addr addrspace(4) constant [5 x float] [float 0.0, float 1.0, float 2.0, float 3.0, float 4.0], align 4
10
11; FUNC-LABEL: {{^}}float:
12; GCN: s_load_dword
13
14; EG: VTX_READ_32
15; EG: @float_gv
16; EG-NOT: MOVA_INT
17; EG-NOT: MOV
18define amdgpu_kernel void @float(float addrspace(1)* %out, i32 %index) {
19entry:
20  %0 = getelementptr inbounds [5 x float], [5 x float] addrspace(4)* @float_gv, i32 0, i32 %index
21  %1 = load float, float addrspace(4)* %0
22  store float %1, float addrspace(1)* %out
23  ret void
24}
25
26@i32_gv = internal unnamed_addr addrspace(4) constant [5 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4], align 4
27
28; FUNC-LABEL: {{^}}i32:
29
30; GCN: s_load_dword
31
32; EG: VTX_READ_32
33; EG: @i32_gv
34; EG-NOT: MOVA_INT
35; EG-NOT: MOV
36define amdgpu_kernel void @i32(i32 addrspace(1)* %out, i32 %index) {
37entry:
38  %0 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(4)* @i32_gv, i32 0, i32 %index
39  %1 = load i32, i32 addrspace(4)* %0
40  store i32 %1, i32 addrspace(1)* %out
41  ret void
42}
43
44
45%struct.foo = type { float, [5 x i32] }
46
47@struct_foo_gv = internal unnamed_addr addrspace(4) constant [1 x %struct.foo] [ %struct.foo { float 16.0, [5 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4] } ]
48
49; FUNC-LABEL: {{^}}struct_foo_gv_load:
50; GCN: s_load_dword
51
52; EG: VTX_READ_32
53; EG: @struct_foo_gv
54; EG-NOT: MOVA_INT
55; EG-NOT: MOV
56define amdgpu_kernel void @struct_foo_gv_load(i32 addrspace(1)* %out, i32 %index) {
57  %gep = getelementptr inbounds [1 x %struct.foo], [1 x %struct.foo] addrspace(4)* @struct_foo_gv, i32 0, i32 0, i32 1, i32 %index
58  %load = load i32, i32 addrspace(4)* %gep, align 4
59  store i32 %load, i32 addrspace(1)* %out, align 4
60  ret void
61}
62
63@array_v1_gv = internal addrspace(4) constant [4 x <1 x i32>] [ <1 x i32> <i32 1>,
64                                                                <1 x i32> <i32 2>,
65                                                                <1 x i32> <i32 3>,
66                                                                <1 x i32> <i32 4> ]
67
68; FUNC-LABEL: {{^}}array_v1_gv_load:
69; GCN: s_load_dword
70
71; EG: VTX_READ_32
72; EG: @array_v1_gv
73; EG-NOT: MOVA_INT
74; EG-NOT: MOV
75define amdgpu_kernel void @array_v1_gv_load(<1 x i32> addrspace(1)* %out, i32 %index) {
76  %gep = getelementptr inbounds [4 x <1 x i32>], [4 x <1 x i32>] addrspace(4)* @array_v1_gv, i32 0, i32 %index
77  %load = load <1 x i32>, <1 x i32> addrspace(4)* %gep, align 4
78  store <1 x i32> %load, <1 x i32> addrspace(1)* %out, align 4
79  ret void
80}
81
82; FUNC-LABEL: {{^}}gv_addressing_in_branch:
83
84; EG: VTX_READ_32
85; EG: @float_gv
86; EG-NOT: MOVA_INT
87define amdgpu_kernel void @gv_addressing_in_branch(float addrspace(1)* %out, i32 %index, i32 %a) {
88entry:
89  %0 = icmp eq i32 0, %a
90  br i1 %0, label %if, label %else
91
92if:
93  %1 = getelementptr inbounds [5 x float], [5 x float] addrspace(4)* @float_gv, i32 0, i32 %index
94  %2 = load float, float addrspace(4)* %1
95  store float %2, float addrspace(1)* %out
96  br label %endif
97
98else:
99  store float 1.0, float addrspace(1)* %out
100  br label %endif
101
102endif:
103  ret void
104}
105