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
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -march=amdgcn -mcpu=hawaii -mattr=+flat-for-global -run-pass=amdgpu-regbankselect %s -verify-machineinstrs -o - | FileCheck %s
--- |
define amdgpu_kernel void @load_constant(ptr addrspace(4) %ptr0) {
ret void
}
define amdgpu_kernel void @load_constant_volatile(ptr addrspace(4) %ptr0) {
ret void
}
define amdgpu_kernel void @load_global_uniform_invariant(ptr addrspace(1) %ptr1) {
%tmp0 = load i32, ptr addrspace(1) %ptr1
ret void
}
define amdgpu_kernel void @load_global_uniform_noclobber(ptr addrspace(1) %ptr1) {
%tmp0 = load i32, ptr addrspace(1) %ptr1, !amdgpu.noclobber !0
ret void
}
define amdgpu_kernel void @load_global_uniform_variant(ptr addrspace(1) %ptr1) {
%tmp0 = load i32, ptr addrspace(1) %ptr1
ret void
}
define amdgpu_kernel void @load_global_uniform_volatile_invariant(ptr addrspace(1) %ptr1) {
%tmp0 = load i32, ptr addrspace(1) %ptr1
ret void
}
define amdgpu_kernel void @load_global_uniform_atomic_invariant(ptr addrspace(1) %ptr1) {
%tmp0 = load i32, ptr addrspace(1) %ptr1
ret void
}
define amdgpu_kernel void @load_global_non_uniform(ptr addrspace(1) %ptr2) {
%tmp0 = call i32 @llvm.amdgcn.workitem.id.x() #0
%tmp1 = getelementptr i32, ptr addrspace(1) %ptr2, i32 %tmp0
%tmp2 = load i32, ptr addrspace(1) %tmp1
ret void
}
define void @non_power_of_2() { ret void }
define amdgpu_kernel void @load_constant_v4i16_from_8_align8(ptr addrspace(4) %ptr0) {
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #0
attributes #0 = { nounwind readnone }
!0 = !{}
...
---
name: load_constant
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_constant
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p4) :: (load (s32) from %ir.ptr0, addrspace 4)
%0:_(p4) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32) from %ir.ptr0)
...
---
name: load_constant_volatile
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_constant_volatile
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p4) :: (volatile load (s32) from %ir.ptr0, addrspace 4)
%0:_(p4) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (volatile load (s32) from %ir.ptr0)
...
---
name: load_global_uniform_invariant
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_uniform_invariant
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p1) :: (invariant load (s32) from %ir.ptr1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (invariant load (s32) from %ir.ptr1)
...
---
name: load_global_uniform_noclobber
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_uniform_noclobber
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
; CHECK-NEXT: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load (s32) from %ir.ptr1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32) from %ir.ptr1)
...
---
name: load_global_uniform_variant
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_uniform_variant
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
; CHECK-NEXT: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load (s32) from %ir.ptr1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32) from %ir.ptr1)
...
---
name: load_global_uniform_volatile_invariant
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_uniform_volatile_invariant
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
; CHECK-NEXT: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (volatile invariant load (s32) from %ir.ptr1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (volatile invariant load (s32) from %ir.ptr1)
...
---
name: load_global_uniform_atomic_invariant
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_uniform_atomic_invariant
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
; CHECK-NEXT: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (invariant load acquire (s32) from %ir.ptr1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (invariant load acquire (s32) from %ir.ptr1)
...
---
name: load_global_non_uniform
legalized: true
body: |
bb.0:
liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: load_global_non_uniform
; CHECK: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
; CHECK-NEXT: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load (s32) from %ir.tmp1, addrspace 1)
%0:_(p1) = COPY $sgpr0_sgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32) from %ir.tmp1)
...
---
name: non_power_of_2
legalized: true
body: |
bb.0:
; CHECK-LABEL: name: non_power_of_2
; CHECK: [[DEF:%[0-9]+]]:sgpr(s448) = G_IMPLICIT_DEF
; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:sgpr(s32) = G_EXTRACT [[DEF]](s448), 0
; CHECK-NEXT: $sgpr0 = COPY [[EXTRACT]](s32)
; CHECK-NEXT: SI_RETURN_TO_EPILOG $sgpr0
%0:_(s448) = G_IMPLICIT_DEF
%1:_(s32) = G_EXTRACT %0:_(s448), 0
$sgpr0 = COPY %1:_(s32)
SI_RETURN_TO_EPILOG $sgpr0
...
---
name: load_constant_v4i16_from_8_align8
legalized: true
body: |
bb.0:
; CHECK-LABEL: name: load_constant_v4i16_from_8_align8
; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>) from %ir.ptr0, addrspace 4)
%0:_(p4) = COPY $sgpr0_sgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>) from %ir.ptr0, align 8, addrspace 4)
...
|
