1; RUN: llc -march=amdgcn -mcpu=gfx902 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
2
3; GCN-LABEL: {{^}}vector_clause:
4; GCN:      global_load_dwordx4
5; GCN-NEXT: global_load_dwordx4
6; GCN-NEXT: global_load_dwordx4
7; GCN-NEXT: global_load_dwordx4
8; GCN-NEXT: s_nop
9define amdgpu_kernel void @vector_clause(<4 x i32> addrspace(1)* noalias nocapture readonly %arg, <4 x i32> addrspace(1)* noalias nocapture %arg1) {
10bb:
11  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
12  %tmp2 = zext i32 %tmp to i64
13  %tmp3 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp2
14  %tmp4 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp3, align 16
15  %tmp5 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp2
16  %tmp6 = add nuw nsw i64 %tmp2, 1
17  %tmp7 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp6
18  %tmp8 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp7, align 16
19  %tmp9 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp6
20  %tmp10 = add nuw nsw i64 %tmp2, 2
21  %tmp11 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp10
22  %tmp12 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp11, align 16
23  %tmp13 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp10
24  %tmp14 = add nuw nsw i64 %tmp2, 3
25  %tmp15 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp14
26  %tmp16 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp15, align 16
27  %tmp17 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp14
28  store <4 x i32> %tmp4, <4 x i32> addrspace(1)* %tmp5, align 16
29  store <4 x i32> %tmp8, <4 x i32> addrspace(1)* %tmp9, align 16
30  store <4 x i32> %tmp12, <4 x i32> addrspace(1)* %tmp13, align 16
31  store <4 x i32> %tmp16, <4 x i32> addrspace(1)* %tmp17, align 16
32  ret void
33}
34
35; GCN-LABEL: {{^}}scalar_clause:
36; GCN:      s_load_dwordx2
37; GCN-NEXT: s_load_dwordx2
38; GCN-NEXT: s_nop
39; GCN-NEXT: s_waitcnt lgkmcnt(0)
40; GCN-NEXT: s_load_dwordx4
41; GCN-NEXT: s_load_dwordx4
42; GCN-NEXT: s_load_dwordx4
43; GCN-NEXT: s_load_dwordx4
44define amdgpu_kernel void @scalar_clause(<4 x i32> addrspace(1)* noalias nocapture readonly %arg, <4 x i32> addrspace(1)* noalias nocapture %arg1) {
45bb:
46  %tmp = load <4 x i32>, <4 x i32> addrspace(1)* %arg, align 16
47  %tmp2 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 1
48  %tmp3 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp2, align 16
49  %tmp4 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 1
50  %tmp5 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 2
51  %tmp6 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp5, align 16
52  %tmp7 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 2
53  %tmp8 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 3
54  %tmp9 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp8, align 16
55  %tmp10 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 3
56  store <4 x i32> %tmp, <4 x i32> addrspace(1)* %arg1, align 16
57  store <4 x i32> %tmp3, <4 x i32> addrspace(1)* %tmp4, align 16
58  store <4 x i32> %tmp6, <4 x i32> addrspace(1)* %tmp7, align 16
59  store <4 x i32> %tmp9, <4 x i32> addrspace(1)* %tmp10, align 16
60  ret void
61}
62
63; GCN-LABEL: {{^}}mubuf_clause:
64; GCN:      buffer_load_dword
65; GCN-NEXT: buffer_load_dword
66; GCN-NEXT: buffer_load_dword
67; GCN-NEXT: buffer_load_dword
68; GCN-NEXT: buffer_load_dword
69; GCN-NEXT: buffer_load_dword
70; GCN-NEXT: buffer_load_dword
71; GCN-NEXT: buffer_load_dword
72; GCN-NEXT: buffer_load_dword
73; GCN-NEXT: buffer_load_dword
74; GCN-NEXT: buffer_load_dword
75; GCN-NEXT: buffer_load_dword
76; GCN-NEXT: buffer_load_dword
77; GCN-NEXT: buffer_load_dword
78; GCN-NEXT: buffer_load_dword
79; GCN-NEXT: s_nop
80; GCN-NEXT: buffer_load_dword
81define void @mubuf_clause(<4 x i32> addrspace(5)* noalias nocapture readonly %arg, <4 x i32> addrspace(5)* noalias nocapture %arg1) {
82bb:
83  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
84  %tmp2 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg, i32 %tmp
85  %tmp3 = load <4 x i32>, <4 x i32> addrspace(5)* %tmp2, align 16
86  %tmp4 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg1, i32 %tmp
87  %tmp5 = add nuw nsw i32 %tmp, 1
88  %tmp6 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg, i32 %tmp5
89  %tmp7 = load <4 x i32>, <4 x i32> addrspace(5)* %tmp6, align 16
90  %tmp8 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg1, i32 %tmp5
91  %tmp9 = add nuw nsw i32 %tmp, 2
92  %tmp10 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg, i32 %tmp9
93  %tmp11 = load <4 x i32>, <4 x i32> addrspace(5)* %tmp10, align 16
94  %tmp12 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg1, i32 %tmp9
95  %tmp13 = add nuw nsw i32 %tmp, 3
96  %tmp14 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg, i32 %tmp13
97  %tmp15 = load <4 x i32>, <4 x i32> addrspace(5)* %tmp14, align 16
98  %tmp16 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(5)* %arg1, i32 %tmp13
99  store <4 x i32> %tmp3, <4 x i32> addrspace(5)* %tmp4, align 16
100  store <4 x i32> %tmp7, <4 x i32> addrspace(5)* %tmp8, align 16
101  store <4 x i32> %tmp11, <4 x i32> addrspace(5)* %tmp12, align 16
102  store <4 x i32> %tmp15, <4 x i32> addrspace(5)* %tmp16, align 16
103  ret void
104}
105
106; GCN-LABEL: {{^}}vector_clause_indirect:
107; GCN: global_load_dwordx2 [[ADDR:v\[[0-9:]+\]]], v[{{[0-9:]+}}], off
108; GCN-NEXT: s_nop
109; GCN-NEXT: s_waitcnt vmcnt(0)
110; GCN-NEXT: global_load_dwordx4 v[{{[0-9:]+}}], [[ADDR]], off
111; GCN-NEXT: global_load_dwordx4 v[{{[0-9:]+}}], [[ADDR]], off offset:16
112define amdgpu_kernel void @vector_clause_indirect(i64 addrspace(1)* noalias nocapture readonly %arg, <4 x i32> addrspace(1)* noalias nocapture readnone %arg1, <4 x i32> addrspace(1)* noalias nocapture %arg2) {
113bb:
114  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
115  %tmp3 = zext i32 %tmp to i64
116  %tmp4 = getelementptr inbounds i64, i64 addrspace(1)* %arg, i64 %tmp3
117  %tmp5 = bitcast i64 addrspace(1)* %tmp4 to <4 x i32> addrspace(1)* addrspace(1)*
118  %tmp6 = load <4 x i32> addrspace(1)*, <4 x i32> addrspace(1)* addrspace(1)* %tmp5, align 8
119  %tmp7 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp6, align 16
120  %tmp8 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %tmp6, i64 1
121  %tmp9 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp8, align 16
122  store <4 x i32> %tmp7, <4 x i32> addrspace(1)* %arg2, align 16
123  %tmp10 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg2, i64 1
124  store <4 x i32> %tmp9, <4 x i32> addrspace(1)* %tmp10, align 16
125  ret void
126}
127
128; GCN-LABEL: {{^}}load_global_d16_hi:
129; GCN:      global_load_short_d16_hi v
130; GCN-NEXT: s_nop
131; GCN-NEXT: global_load_short_d16_hi v
132define void @load_global_d16_hi(i16 addrspace(1)* %in, i16 %reg, <2 x i16> addrspace(1)* %out) {
133entry:
134  %gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32
135  %load1 = load i16, i16 addrspace(1)* %in
136  %load2 = load i16, i16 addrspace(1)* %gep
137  %build0 = insertelement <2 x i16> undef, i16 %reg, i32 0
138  %build1 = insertelement <2 x i16> %build0, i16 %load1, i32 1
139  store <2 x i16> %build1, <2 x i16> addrspace(1)* %out
140  %build2 = insertelement <2 x i16> undef, i16 %reg, i32 0
141  %build3 = insertelement <2 x i16> %build2, i16 %load2, i32 1
142  %gep2 = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %out, i64 32
143  store <2 x i16> %build3, <2 x i16> addrspace(1)* %gep2
144  ret void
145}
146
147; GCN-LABEL: {{^}}load_global_d16_lo:
148; GCN:      global_load_short_d16 v
149; GCN-NEXT: s_nop
150; GCN-NEXT: global_load_short_d16 v
151define void @load_global_d16_lo(i16 addrspace(1)* %in, i32 %reg, <2 x i16> addrspace(1)* %out) {
152entry:
153  %gep = getelementptr inbounds i16, i16 addrspace(1)* %in, i64 32
154  %reg.bc1 = bitcast i32 %reg to <2 x i16>
155  %reg.bc2 = bitcast i32 %reg to <2 x i16>
156  %load1 = load i16, i16 addrspace(1)* %in
157  %load2 = load i16, i16 addrspace(1)* %gep
158  %build1 = insertelement <2 x i16> %reg.bc1, i16 %load1, i32 0
159  %build2 = insertelement <2 x i16> %reg.bc2, i16 %load2, i32 0
160  %gep2 = getelementptr inbounds <2 x i16>, <2 x i16> addrspace(1)* %out, i64 32
161  store <2 x i16> %build1, <2 x i16> addrspace(1)* %out
162  store <2 x i16> %build2, <2 x i16> addrspace(1)* %gep2
163  ret void
164}
165
166declare i32 @llvm.amdgcn.workitem.id.x()
167