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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
|
; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; FUNC-LABEL: @lds_atomic_xchg_ret_i32:
; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]],
; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4
; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]]
; SI: DS_WRXCHG_RTN_B32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]], 0x0, [M0]
; SI: BUFFER_STORE_DWORD [[RESULT]],
; SI: S_ENDPGM
define void @lds_atomic_xchg_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw xchg i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_xchg_ret_i32_offset:
; SI: DS_WRXCHG_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_xchg_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw xchg i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; XXX - Is it really necessary to load 4 into VGPR?
; FUNC-LABEL: @lds_atomic_add_ret_i32:
; SI: S_LOAD_DWORD [[SPTR:s[0-9]+]],
; SI: V_MOV_B32_e32 [[DATA:v[0-9]+]], 4
; SI: V_MOV_B32_e32 [[VPTR:v[0-9]+]], [[SPTR]]
; SI: DS_ADD_RTN_U32 [[RESULT:v[0-9]+]], [[VPTR]], [[DATA]], 0x0, [M0]
; SI: BUFFER_STORE_DWORD [[RESULT]],
; SI: S_ENDPGM
define void @lds_atomic_add_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw add i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_add_ret_i32_offset:
; SI: DS_ADD_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_add_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw add i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_inc_ret_i32:
; SI: DS_INC_RTN_U32
; SI: S_ENDPGM
define void @lds_atomic_inc_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw add i32 addrspace(3)* %ptr, i32 1 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_inc_ret_i32_offset:
; SI: DS_INC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_inc_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw add i32 addrspace(3)* %gep, i32 1 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_sub_ret_i32:
; SI: DS_SUB_RTN_U32
; SI: S_ENDPGM
define void @lds_atomic_sub_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw sub i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_sub_ret_i32_offset:
; SI: DS_SUB_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_sub_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw sub i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_dec_ret_i32:
; SI: DS_DEC_RTN_U32
; SI: S_ENDPGM
define void @lds_atomic_dec_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw sub i32 addrspace(3)* %ptr, i32 1 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_dec_ret_i32_offset:
; SI: DS_DEC_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_dec_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw sub i32 addrspace(3)* %gep, i32 1 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_and_ret_i32:
; SI: DS_AND_RTN_B32
; SI: S_ENDPGM
define void @lds_atomic_and_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw and i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_and_ret_i32_offset:
; SI: DS_AND_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_and_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw and i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_or_ret_i32:
; SI: DS_OR_RTN_B32
; SI: S_ENDPGM
define void @lds_atomic_or_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw or i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_or_ret_i32_offset:
; SI: DS_OR_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_or_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw or i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_xor_ret_i32:
; SI: DS_XOR_RTN_B32
; SI: S_ENDPGM
define void @lds_atomic_xor_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw xor i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_xor_ret_i32_offset:
; SI: DS_XOR_RTN_B32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_xor_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw xor i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FIXME: There is no atomic nand instr
; XFUNC-LABEL: @lds_atomic_nand_ret_i32:uction, so we somehow need to expand this.
; define void @lds_atomic_nand_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
; %result = atomicrmw nand i32 addrspace(3)* %ptr, i32 4 seq_cst
; store i32 %result, i32 addrspace(1)* %out, align 4
; ret void
; }
; FUNC-LABEL: @lds_atomic_min_ret_i32:
; SI: DS_MIN_RTN_I32
; SI: S_ENDPGM
define void @lds_atomic_min_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw min i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_min_ret_i32_offset:
; SI: DS_MIN_RTN_I32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_min_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw min i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_max_ret_i32:
; SI: DS_MAX_RTN_I32
; SI: S_ENDPGM
define void @lds_atomic_max_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw max i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_max_ret_i32_offset:
; SI: DS_MAX_RTN_I32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_max_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw max i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_umin_ret_i32:
; SI: DS_MIN_RTN_U32
; SI: S_ENDPGM
define void @lds_atomic_umin_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw umin i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_umin_ret_i32_offset:
; SI: DS_MIN_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_umin_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw umin i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_umax_ret_i32:
; SI: DS_MAX_RTN_U32
; SI: S_ENDPGM
define void @lds_atomic_umax_ret_i32(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%result = atomicrmw umax i32 addrspace(3)* %ptr, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: @lds_atomic_umax_ret_i32_offset:
; SI: DS_MAX_RTN_U32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 0x10
; SI: S_ENDPGM
define void @lds_atomic_umax_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr) nounwind {
%gep = getelementptr i32 addrspace(3)* %ptr, i32 4
%result = atomicrmw umax i32 addrspace(3)* %gep, i32 4 seq_cst
store i32 %result, i32 addrspace(1)* %out, align 4
ret void
}
|
