1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
|
; RUN: llc -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GCN-LABEL: {{^}}store_build_vector_multiple_uses_v4i32:
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_store_dwordx4
; GCN: buffer_store_dwordx4
; GCN: buffer_store_dword
; GCN: buffer_store_dword
; GCN: buffer_store_dword
; GCN: buffer_store_dword
define amdgpu_kernel void @store_build_vector_multiple_uses_v4i32(ptr addrspace(1) noalias %out0,
ptr addrspace(1) noalias %out1,
ptr addrspace(1) noalias %out2,
ptr addrspace(1) %in) {
%elt0 = load volatile i32, ptr addrspace(1) %in
%elt1 = load volatile i32, ptr addrspace(1) %in
%elt2 = load volatile i32, ptr addrspace(1) %in
%elt3 = load volatile i32, ptr addrspace(1) %in
%vec0 = insertelement <4 x i32> poison, i32 %elt0, i32 0
%vec1 = insertelement <4 x i32> %vec0, i32 %elt1, i32 1
%vec2 = insertelement <4 x i32> %vec1, i32 %elt2, i32 2
%vec3 = insertelement <4 x i32> %vec2, i32 %elt3, i32 3
store <4 x i32> %vec3, ptr addrspace(1) %out0
store <4 x i32> %vec3, ptr addrspace(1) %out1
%extract0 = extractelement <4 x i32> %vec3, i32 0
%extract1 = extractelement <4 x i32> %vec3, i32 1
%extract2 = extractelement <4 x i32> %vec3, i32 2
%extract3 = extractelement <4 x i32> %vec3, i32 3
store volatile i32 %extract0, ptr addrspace(1) %out2
store volatile i32 %extract1, ptr addrspace(1) %out2
store volatile i32 %extract2, ptr addrspace(1) %out2
store volatile i32 %extract3, ptr addrspace(1) %out2
ret void
}
; GCN-LABEL: {{^}}store_build_vector_multiple_extract_uses_v4i32:
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_store_dwordx4
; GCN: buffer_store_dword
; GCN: buffer_store_dword
; GCN: buffer_store_dword
; GCN: buffer_store_dword
define amdgpu_kernel void @store_build_vector_multiple_extract_uses_v4i32(ptr addrspace(1) noalias %out0,
ptr addrspace(1) noalias %out1,
ptr addrspace(1) noalias %out2,
ptr addrspace(1) %in) {
%elt0 = load volatile i32, ptr addrspace(1) %in
%elt1 = load volatile i32, ptr addrspace(1) %in
%elt2 = load volatile i32, ptr addrspace(1) %in
%elt3 = load volatile i32, ptr addrspace(1) %in
%vec0 = insertelement <4 x i32> poison, i32 %elt0, i32 0
%vec1 = insertelement <4 x i32> %vec0, i32 %elt1, i32 1
%vec2 = insertelement <4 x i32> %vec1, i32 %elt2, i32 2
%vec3 = insertelement <4 x i32> %vec2, i32 %elt3, i32 3
%extract0 = extractelement <4 x i32> %vec3, i32 0
%extract1 = extractelement <4 x i32> %vec3, i32 1
%extract2 = extractelement <4 x i32> %vec3, i32 2
%extract3 = extractelement <4 x i32> %vec3, i32 3
%op0 = add i32 %extract0, 3
%op1 = sub i32 %extract1, 9
%op2 = xor i32 %extract2, 1231412
%op3 = and i32 %extract3, 258233412312
store <4 x i32> %vec3, ptr addrspace(1) %out0
store volatile i32 %op0, ptr addrspace(1) %out2
store volatile i32 %op1, ptr addrspace(1) %out2
store volatile i32 %op2, ptr addrspace(1) %out2
store volatile i32 %op3, ptr addrspace(1) %out2
ret void
}
; GCN-LABEL: {{^}}store_build_vector_multiple_uses_v4i32_bitcast_to_v2i64:
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_load_dword
; GCN: buffer_store_dwordx4
; GCN: buffer_store_dwordx2
; GCN: buffer_store_dwordx2
define amdgpu_kernel void @store_build_vector_multiple_uses_v4i32_bitcast_to_v2i64(ptr addrspace(1) noalias %out0,
ptr addrspace(1) noalias %out1,
ptr addrspace(1) noalias %out2,
ptr addrspace(1) %in) {
%elt0 = load volatile i32, ptr addrspace(1) %in
%elt1 = load volatile i32, ptr addrspace(1) %in
%elt2 = load volatile i32, ptr addrspace(1) %in
%elt3 = load volatile i32, ptr addrspace(1) %in
%vec0 = insertelement <4 x i32> poison, i32 %elt0, i32 0
%vec1 = insertelement <4 x i32> %vec0, i32 %elt1, i32 1
%vec2 = insertelement <4 x i32> %vec1, i32 %elt2, i32 2
%vec3 = insertelement <4 x i32> %vec2, i32 %elt3, i32 3
%bc.vec3 = bitcast <4 x i32> %vec3 to <2 x i64>
store <2 x i64> %bc.vec3, ptr addrspace(1) %out0
%extract0 = extractelement <2 x i64> %bc.vec3, i32 0
%extract1 = extractelement <2 x i64> %bc.vec3, i32 1
store volatile i64 %extract0, ptr addrspace(1) %out2
store volatile i64 %extract1, ptr addrspace(1) %out2
ret void
}
|
