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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=amdgcn-- -verify-machineinstrs | FileCheck %s -check-prefix=GCN
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
; Make sure the add and load are reduced to 32-bits even with the
; bitcast to vector.
define amdgpu_kernel void @bitcast_int_to_vector_extract_0(ptr addrspace(1) %out, ptr addrspace(1) %in, i64 %b) {
; GCN-LABEL: bitcast_int_to_vector_extract_0:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_load_dword s12, s[4:5], 0xd
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s10, 0
; GCN-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b64 s[8:9], s[2:3]
; GCN-NEXT: buffer_load_dword v0, v[0:1], s[8:11], 0 addr64
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: v_add_i32_e32 v0, vcc, s12, v0
; GCN-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GCN-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i64, ptr addrspace(1) %in, i32 %tid
%a = load i64, ptr addrspace(1) %gep
%add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0
store i32 %extract, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @bitcast_fp_to_vector_extract_0(ptr addrspace(1) %out, ptr addrspace(1) %in, double %b) {
; GCN-LABEL: bitcast_fp_to_vector_extract_0:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0xd
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s10, 0
; GCN-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b64 s[8:9], s[2:3]
; GCN-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[8:11], 0 addr64
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: v_add_f64 v[0:1], v[0:1], s[12:13]
; GCN-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GCN-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr double, ptr addrspace(1) %in, i32 %tid
%a = load double, ptr addrspace(1) %gep
%add = fadd double %a, %b
%val.bc = bitcast double %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0
store i32 %extract, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @bitcast_int_to_fpvector_extract_0(ptr addrspace(1) %out, ptr addrspace(1) %in, i64 %b) {
; GCN-LABEL: bitcast_int_to_fpvector_extract_0:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_load_dword s12, s[4:5], 0xd
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s10, 0
; GCN-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b64 s[8:9], s[2:3]
; GCN-NEXT: buffer_load_dword v0, v[0:1], s[8:11], 0 addr64
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: v_add_i32_e32 v0, vcc, s12, v0
; GCN-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GCN-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%gep = getelementptr i64, ptr addrspace(1) %in, i32 %tid
%a = load i64, ptr addrspace(1) %gep
%add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x float>
%extract = extractelement <2 x float> %val.bc, i32 0
store float %extract, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @no_extract_volatile_load_extract0(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; GCN-LABEL: no_extract_volatile_load_extract0:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s10, s6
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s8, s2
; GCN-NEXT: s_mov_b32 s9, s3
; GCN-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0 glc
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GCN-NEXT: s_endpgm
entry:
%vec = load volatile <4 x i32>, ptr addrspace(1) %in
%elt0 = extractelement <4 x i32> %vec, i32 0
store i32 %elt0, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @no_extract_volatile_load_extract2(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; GCN-LABEL: no_extract_volatile_load_extract2:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s10, s6
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s8, s2
; GCN-NEXT: s_mov_b32 s9, s3
; GCN-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0 glc
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: buffer_store_dword v2, off, s[4:7], 0
; GCN-NEXT: s_endpgm
entry:
%vec = load volatile <4 x i32>, ptr addrspace(1) %in
%elt2 = extractelement <4 x i32> %vec, i32 2
store i32 %elt2, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @no_extract_volatile_load_dynextract(ptr addrspace(1) %out, ptr addrspace(1) %in, i32 %idx) {
; GCN-LABEL: no_extract_volatile_load_dynextract:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GCN-NEXT: s_load_dword s12, s[4:5], 0xd
; GCN-NEXT: s_mov_b32 s7, 0xf000
; GCN-NEXT: s_mov_b32 s6, -1
; GCN-NEXT: s_mov_b32 s10, s6
; GCN-NEXT: s_mov_b32 s11, s7
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s4, s0
; GCN-NEXT: s_mov_b32 s5, s1
; GCN-NEXT: s_mov_b32 s8, s2
; GCN-NEXT: s_mov_b32 s9, s3
; GCN-NEXT: buffer_load_dwordx4 v[0:3], off, s[8:11], 0 glc
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_cmp_eq_u32 s12, 1
; GCN-NEXT: s_cselect_b64 vcc, -1, 0
; GCN-NEXT: s_cmp_eq_u32 s12, 2
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: s_cselect_b64 vcc, -1, 0
; GCN-NEXT: s_cmp_eq_u32 s12, 3
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: s_cselect_b64 vcc, -1, 0
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GCN-NEXT: s_endpgm
entry:
%vec = load volatile <4 x i32>, ptr addrspace(1) %in
%eltN = extractelement <4 x i32> %vec, i32 %idx
store i32 %eltN, ptr addrspace(1) %out
ret void
}
|
