1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
2 ; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
4 ; GCN-LABEL: {{^}}any_extend_vector_inreg_v16i8_to_v4i32:
6 ; GCN-DAG: s_load_dwordx4
7 ; GCN-DAG: s_load_dword
9 ; GCN: {{buffer|flat}}_store_byte
10 ; GCN: {{buffer|flat}}_store_byte
11 ; GCN: {{buffer|flat}}_store_byte
12 ; GCN: {{buffer|flat}}_store_byte
14 ; GCN: {{buffer|flat}}_store_byte
15 ; GCN: {{buffer|flat}}_store_byte
16 ; GCN: {{buffer|flat}}_store_byte
17 ; GCN: {{buffer|flat}}_store_byte
19 ; GCN: {{buffer|flat}}_store_byte
20 ; GCN: {{buffer|flat}}_store_byte
21 ; GCN: {{buffer|flat}}_store_byte
22 ; GCN: {{buffer|flat}}_store_byte
24 ; GCN: {{buffer|flat}}_store_byte
25 ; GCN: {{buffer|flat}}_store_byte
26 ; GCN: {{buffer|flat}}_store_byte
27 ; GCN: {{buffer|flat}}_store_byte
28 define amdgpu_kernel void @any_extend_vector_inreg_v16i8_to_v4i32(<8 x i8> addrspace(1)* nocapture readonly %arg, <16 x i8> addrspace(1)* %arg1) local_unnamed_addr #0 {
30 %tmp = bitcast <8 x i8> addrspace(1)* %arg to <16 x i8> addrspace(1)*
31 %tmp2 = load <16 x i8>, <16 x i8> addrspace(1)* %tmp, align 16
32 %tmp3 = extractelement <16 x i8> %tmp2, i64 4
33 %tmp6 = extractelement <16 x i8> %tmp2, i64 11
34 %tmp10 = getelementptr inbounds <8 x i8>, <8 x i8> addrspace(1)* %arg, i64 2
35 %tmp11 = bitcast <8 x i8> addrspace(1)* %tmp10 to <16 x i8> addrspace(1)*
36 %tmp12 = load <16 x i8>, <16 x i8> addrspace(1)* %tmp11, align 16
37 %tmp13 = extractelement <16 x i8> %tmp12, i64 7
38 %tmp17 = extractelement <16 x i8> %tmp12, i64 12
39 %tmp21 = getelementptr inbounds <8 x i8>, <8 x i8> addrspace(1)* %arg, i64 4
40 %tmp22 = bitcast <8 x i8> addrspace(1)* %tmp21 to <16 x i8> addrspace(1)*
41 %tmp23 = load <16 x i8>, <16 x i8> addrspace(1)* %tmp22, align 16
42 %tmp24 = extractelement <16 x i8> %tmp23, i64 3
43 %tmp1 = insertelement <16 x i8> undef, i8 %tmp3, i32 2
44 %tmp4 = insertelement <16 x i8> %tmp1, i8 0, i32 3
45 %tmp5 = insertelement <16 x i8> %tmp4, i8 0, i32 4
46 %tmp7 = insertelement <16 x i8> %tmp5, i8 %tmp6, i32 5
47 %tmp8 = insertelement <16 x i8> %tmp7, i8 0, i32 6
48 %tmp9 = insertelement <16 x i8> %tmp8, i8 %tmp13, i32 7
49 %tmp14 = insertelement <16 x i8> %tmp9, i8 0, i32 8
50 %tmp15 = insertelement <16 x i8> %tmp14, i8 %tmp17, i32 9
51 %tmp16 = insertelement <16 x i8> %tmp15, i8 0, i32 10
52 %tmp18 = insertelement <16 x i8> %tmp16, i8 0, i32 11
53 %tmp19 = insertelement <16 x i8> %tmp18, i8 %tmp24, i32 12
54 store <16 x i8> %tmp19, <16 x i8> addrspace(1)* %arg1, align 1
58 attributes #0 = { nounwind }