1; RUN: llc -mtriple=amdgcn--amdpal -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
2; RUN: llc -mtriple=amdgcn--amdpal -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
3
4; TODO: Some of those tests fail with OS == amdhsa due to unreasonable register
5;       allocation differences.
6
7; SI-LABEL: {{^}}s_addk_i32_k0:
8; SI: s_load_dword [[VAL:s[0-9]+]]
9; SI: s_addk_i32 [[VAL]], 0x41
10; SI: v_mov_b32_e32 [[VRESULT:v[0-9]+]], [[VAL]]
11; SI: buffer_store_dword [[VRESULT]]
12; SI: s_endpgm
13define amdgpu_kernel void @s_addk_i32_k0(i32 addrspace(1)* %out, i32 %b) {
14  %add = add i32 %b, 65
15  store i32 %add, i32 addrspace(1)* %out
16  ret void
17}
18
19; FIXME: This should be folded with any number of uses.
20; SI-LABEL: {{^}}s_addk_i32_k0_x2:
21; SI: s_movk_i32 [[K:s[0-9]+]], 0x41
22; SI-DAG: s_add_i32 {{s[0-9]+}}, {{s[0-9]+}}, [[K]]
23; SI-DAG: s_add_i32 {{s[0-9]+}}, {{s[0-9]+}}, [[K]]
24; SI: s_endpgm
25define amdgpu_kernel void @s_addk_i32_k0_x2(i32 addrspace(1)* %out0, i32 addrspace(1)* %out1, i32 %a, i32 %b) {
26  %add0 = add i32 %a, 65
27  %add1 = add i32 %b, 65
28  store i32 %add0, i32 addrspace(1)* %out0
29  store i32 %add1, i32 addrspace(1)* %out1
30  ret void
31}
32
33; SI-LABEL: {{^}}s_addk_i32_k1:
34; SI: s_addk_i32 {{s[0-9]+}}, 0x7fff{{$}}
35; SI: s_endpgm
36define amdgpu_kernel void @s_addk_i32_k1(i32 addrspace(1)* %out, i32 %b) {
37  %add = add i32 %b, 32767 ; (1 << 15) - 1
38  store i32 %add, i32 addrspace(1)* %out
39  ret void
40}
41
42; SI-LABEL: {{^}}s_addk_i32_k2:
43; SI: s_sub_i32 s{{[0-9]+}}, s{{[0-9]+}}, 17
44; SI: s_endpgm
45define amdgpu_kernel void @s_addk_i32_k2(i32 addrspace(1)* %out, i32 %b) {
46  %add = add i32 %b, -17
47  store i32 %add, i32 addrspace(1)* %out
48  ret void
49}
50
51; SI-LABEL: {{^}}s_addk_i32_k3:
52; SI: s_addk_i32 {{s[0-9]+}}, 0xffbf{{$}}
53; SI: s_endpgm
54define amdgpu_kernel void @s_addk_i32_k3(i32 addrspace(1)* %out, i32 %b) {
55  %add = add i32 %b, -65
56  store i32 %add, i32 addrspace(1)* %out
57  ret void
58}
59
60; SI-LABEL: {{^}}s_addk_v2i32_k0:
61; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x41
62; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x42
63; SI: s_endpgm
64define amdgpu_kernel void @s_addk_v2i32_k0(<2 x i32> addrspace(1)* %out, <2 x i32> %b) {
65  %add = add <2 x i32> %b, <i32 65, i32 66>
66  store <2 x i32> %add, <2 x i32> addrspace(1)* %out
67  ret void
68}
69
70; SI-LABEL: {{^}}s_addk_v4i32_k0:
71; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x41
72; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x42
73; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x43
74; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x44
75; SI: s_endpgm
76define amdgpu_kernel void @s_addk_v4i32_k0(<4 x i32> addrspace(1)* %out, <4 x i32> %b) {
77  %add = add <4 x i32> %b, <i32 65, i32 66, i32 67, i32 68>
78  store <4 x i32> %add, <4 x i32> addrspace(1)* %out
79  ret void
80}
81
82; SI-LABEL: {{^}}s_addk_v8i32_k0:
83; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x41
84; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x42
85; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x43
86; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x44
87; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x45
88; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x46
89; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x47
90; SI-DAG: s_addk_i32 {{s[0-9]+}}, 0x48
91; SI: s_endpgm
92define amdgpu_kernel void @s_addk_v8i32_k0(<8 x i32> addrspace(1)* %out, <8 x i32> %b) {
93  %add = add <8 x i32> %b, <i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72>
94  store <8 x i32> %add, <8 x i32> addrspace(1)* %out
95  ret void
96}
97
98; SI-LABEL: {{^}}no_s_addk_i32_k0:
99; SI: s_add_i32 {{s[0-9]+}}, {{s[0-9]+}}, 0x8000{{$}}
100; SI: s_endpgm
101define amdgpu_kernel void @no_s_addk_i32_k0(i32 addrspace(1)* %out, i32 %b) {
102  %add = add i32 %b, 32768 ; 1 << 15
103  store i32 %add, i32 addrspace(1)* %out
104  ret void
105}
106
107@lds = addrspace(3) global [512 x i32] undef, align 4
108
109; SI-LABEL: {{^}}commute_s_addk_i32:
110; SI: s_addk_i32 s{{[0-9]+}}, 0x800{{$}}
111define amdgpu_kernel void @commute_s_addk_i32(i32 addrspace(1)* %out, i32 %b) #0 {
112  %size = call i32 @llvm.amdgcn.groupstaticsize()
113  %add = add i32 %size, %b
114  call void asm sideeffect "; foo $0, $1", "v,s"([512 x i32] addrspace(3)* @lds, i32 %add)
115  ret void
116}
117
118declare i32 @llvm.amdgcn.groupstaticsize() #1
119
120attributes #0 = { nounwind }
121attributes #1 = { nounwind readnone }
122