1 ; RUN: llc -march=amdgcn -mcpu=gfx902 -verify-machineinstrs -amdgpu-enable-global-sgpr-addr -stop-after=si-form-memory-clauses < %s | FileCheck -check-prefix=GCN %s
3 ; GCN-LABEL: {{^}}name:{{[ ]*}}vector_clause
5 ; GCN-NEXT: LOAD_DWORDX2
6 ; GCN-NEXT: LOAD_DWORDX2
8 define amdgpu_kernel void @vector_clause(<4 x i32> addrspace(1)* noalias nocapture readonly %arg, <4 x i32> addrspace(1)* noalias nocapture %arg1) {
10 %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
11 %tmp2 = zext i32 %tmp to i64
12 %tmp3 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp2
13 %tmp4 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp3, align 16
14 %tmp5 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp2
15 %tmp6 = add nuw nsw i64 %tmp2, 1
16 %tmp7 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp6
17 %tmp8 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp7, align 16
18 %tmp9 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp6
19 %tmp10 = add nuw nsw i64 %tmp2, 2
20 %tmp11 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp10
21 %tmp12 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp11, align 16
22 %tmp13 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp10
23 %tmp14 = add nuw nsw i64 %tmp2, 3
24 %tmp15 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp14
25 %tmp16 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp15, align 16
26 %tmp17 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp14
27 store <4 x i32> %tmp4, <4 x i32> addrspace(1)* %tmp5, align 16
28 store <4 x i32> %tmp8, <4 x i32> addrspace(1)* %tmp9, align 16
29 store <4 x i32> %tmp12, <4 x i32> addrspace(1)* %tmp13, align 16
30 store <4 x i32> %tmp16, <4 x i32> addrspace(1)* %tmp17, align 16
34 ; GCN-LABEL: {{^}}name:{{[ ]*}}no_vector_clause
36 define amdgpu_kernel void @no_vector_clause(<4 x i32> addrspace(1)* noalias nocapture readonly %arg, <4 x i32> addrspace(1)* noalias nocapture %arg1) #0 {
38 %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
39 %tmp2 = zext i32 %tmp to i64
40 %tmp3 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp2
41 %tmp4 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp3, align 16
42 %tmp5 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp2
43 %tmp6 = add nuw nsw i64 %tmp2, 1
44 %tmp7 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp6
45 %tmp8 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp7, align 16
46 %tmp9 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp6
47 %tmp10 = add nuw nsw i64 %tmp2, 2
48 %tmp11 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp10
49 %tmp12 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp11, align 16
50 %tmp13 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp10
51 %tmp14 = add nuw nsw i64 %tmp2, 3
52 %tmp15 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg, i64 %tmp14
53 %tmp16 = load <4 x i32>, <4 x i32> addrspace(1)* %tmp15, align 16
54 %tmp17 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %arg1, i64 %tmp14
55 store <4 x i32> %tmp4, <4 x i32> addrspace(1)* %tmp5, align 16
56 store <4 x i32> %tmp8, <4 x i32> addrspace(1)* %tmp9, align 16
57 store <4 x i32> %tmp12, <4 x i32> addrspace(1)* %tmp13, align 16
58 store <4 x i32> %tmp16, <4 x i32> addrspace(1)* %tmp17, align 16
62 declare i32 @llvm.amdgcn.workitem.id.x()
64 attributes #0 = { "amdgpu-max-memory-clause"="1" }