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
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx90a < %s | FileCheck -check-prefix=GFX90A %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx942 < %s | FileCheck -check-prefix=GFX942 %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefix=GFX1200 %s
define float @struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset(float %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
; GFX90A-NEXT: buffer_atomic_add_f32 v0, v[2:3], s[16:19], s20 idxen offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v3, v2
; GFX942-NEXT: v_mov_b32_e32 v2, v1
; GFX942-NEXT: buffer_atomic_add_f32 v0, v[2:3], s[0:3], s16 idxen offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: buffer_atomic_add_f32 v0, v[1:2], s[0:3], s16 idxen offen th:TH_ATOMIC_RETURN
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call float @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret float %ret
}
; Natural mapping, no voffset
define float @struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset(float %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 inreg %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: buffer_atomic_add_f32 v0, v1, s[16:19], s20 idxen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: buffer_atomic_add_f32 v0, v1, s[0:3], s16 idxen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: buffer_atomic_add_f32 v0, v1, s[0:3], s16 idxen th:TH_ATOMIC_RETURN
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call float @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 0, i32 %soffset, i32 0)
ret float %ret
}
define float @struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc(float %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
; GFX90A-NEXT: buffer_atomic_add_f32 v0, v[2:3], s[16:19], s20 idxen offen glc slc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v3, v2
; GFX942-NEXT: v_mov_b32_e32 v2, v1
; GFX942-NEXT: buffer_atomic_add_f32 v0, v[2:3], s[0:3], s16 idxen offen sc0 nt
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: buffer_atomic_add_f32 v0, v[1:2], s[0:3], s16 idxen offen th:TH_ATOMIC_NT_RETURN
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call float @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 2)
ret float %ret
}
define <2 x half> @struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<2 x half> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
; GFX90A-NEXT: buffer_atomic_pk_add_f16 v0, v[2:3], s[16:19], s20 idxen offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v3, v2
; GFX942-NEXT: v_mov_b32_e32 v2, v1
; GFX942-NEXT: buffer_atomic_pk_add_f16 v0, v[2:3], s[0:3], s16 idxen offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: buffer_atomic_pk_add_f16 v0, v[1:2], s[0:3], s16 idxen offen th:TH_ATOMIC_RETURN
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret <2 x half> %ret
}
; Test waterfall loop
define float @struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v9, v6
; GFX90A-NEXT: v_mov_b32_e32 v8, v5
; GFX90A-NEXT: v_mov_b32_e32 v5, v4
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
; GFX90A-NEXT: s_mov_b64 s[6:7], exec
; GFX90A-NEXT: .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: v_readfirstlane_b32 s8, v2
; GFX90A-NEXT: v_readfirstlane_b32 s9, v3
; GFX90A-NEXT: v_readfirstlane_b32 s10, v4
; GFX90A-NEXT: v_readfirstlane_b32 s11, v5
; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[2:3]
; GFX90A-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[4:5]
; GFX90A-NEXT: v_readfirstlane_b32 s12, v7
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, s12, v7
; GFX90A-NEXT: s_and_b64 s[4:5], s[4:5], vcc
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_atomic_add_f32 v0, v[8:9], s[8:11], s12 idxen offen glc
; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3_vgpr4_vgpr5
; GFX90A-NEXT: ; implicit-def: $vgpr7
; GFX90A-NEXT: ; implicit-def: $vgpr8_vgpr9
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB4_1
; GFX90A-NEXT: ; %bb.2:
; GFX90A-NEXT: s_mov_b64 exec, s[6:7]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v9, v6
; GFX942-NEXT: v_mov_b32_e32 v8, v5
; GFX942-NEXT: v_mov_b32_e32 v5, v4
; GFX942-NEXT: v_mov_b32_e32 v4, v3
; GFX942-NEXT: v_mov_b32_e32 v3, v2
; GFX942-NEXT: v_mov_b32_e32 v2, v1
; GFX942-NEXT: s_mov_b64 s[2:3], exec
; GFX942-NEXT: .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: v_readfirstlane_b32 s4, v2
; GFX942-NEXT: v_readfirstlane_b32 s5, v3
; GFX942-NEXT: v_readfirstlane_b32 s6, v4
; GFX942-NEXT: v_readfirstlane_b32 s7, v5
; GFX942-NEXT: v_cmp_eq_u64_e32 vcc, s[4:5], v[2:3]
; GFX942-NEXT: v_readfirstlane_b32 s8, v7
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[4:5]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, s8, v7
; GFX942-NEXT: s_and_b64 s[0:1], s[0:1], vcc
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_atomic_add_f32 v0, v[8:9], s[4:7], s8 idxen offen sc0
; GFX942-NEXT: ; implicit-def: $vgpr2_vgpr3_vgpr4_vgpr5
; GFX942-NEXT: ; implicit-def: $vgpr7
; GFX942-NEXT: ; implicit-def: $vgpr8_vgpr9
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB4_1
; GFX942-NEXT: ; %bb.2:
; GFX942-NEXT: s_mov_b64 exec, s[2:3]
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_f32_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: s_mov_b32 s2, exec_lo
; GFX1200-NEXT: .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GFX1200-NEXT: v_readfirstlane_b32 s4, v1
; GFX1200-NEXT: v_readfirstlane_b32 s5, v2
; GFX1200-NEXT: v_readfirstlane_b32 s6, v3
; GFX1200-NEXT: v_readfirstlane_b32 s7, v4
; GFX1200-NEXT: v_readfirstlane_b32 s3, v7
; GFX1200-NEXT: s_wait_alu 0xf1ff
; GFX1200-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[1:2]
; GFX1200-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX1200-NEXT: v_cmp_eq_u64_e64 s0, s[6:7], v[3:4]
; GFX1200-NEXT: v_cmp_eq_u32_e64 s1, s3, v7
; GFX1200-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX1200-NEXT: s_wait_alu 0xfffe
; GFX1200-NEXT: s_and_b32 s0, s0, s1
; GFX1200-NEXT: s_wait_alu 0xfffe
; GFX1200-NEXT: s_and_saveexec_b32 s0, s0
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: buffer_atomic_add_f32 v0, v[5:6], s[4:7], s3 idxen offen th:TH_ATOMIC_RETURN
; GFX1200-NEXT: ; implicit-def: $vgpr1_vgpr2_vgpr3_vgpr4
; GFX1200-NEXT: ; implicit-def: $vgpr7
; GFX1200-NEXT: ; implicit-def: $vgpr5_vgpr6
; GFX1200-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX1200-NEXT: s_cbranch_execnz .LBB4_1
; GFX1200-NEXT: ; %bb.2:
; GFX1200-NEXT: s_mov_b32 exec_lo, s2
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call float @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret float %ret
}
; Test waterfall loop
define <2 x half> @struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset) #0 {
; GFX90A-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v9, v6
; GFX90A-NEXT: v_mov_b32_e32 v8, v5
; GFX90A-NEXT: v_mov_b32_e32 v5, v4
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
; GFX90A-NEXT: s_mov_b64 s[6:7], exec
; GFX90A-NEXT: .LBB5_1: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: v_readfirstlane_b32 s8, v2
; GFX90A-NEXT: v_readfirstlane_b32 s9, v3
; GFX90A-NEXT: v_readfirstlane_b32 s10, v4
; GFX90A-NEXT: v_readfirstlane_b32 s11, v5
; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[2:3]
; GFX90A-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[4:5]
; GFX90A-NEXT: v_readfirstlane_b32 s12, v7
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, s12, v7
; GFX90A-NEXT: s_and_b64 s[4:5], s[4:5], vcc
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_atomic_pk_add_f16 v0, v[8:9], s[8:11], s12 idxen offen glc
; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3_vgpr4_vgpr5
; GFX90A-NEXT: ; implicit-def: $vgpr7
; GFX90A-NEXT: ; implicit-def: $vgpr8_vgpr9
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB5_1
; GFX90A-NEXT: ; %bb.2:
; GFX90A-NEXT: s_mov_b64 exec, s[6:7]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX942-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v9, v6
; GFX942-NEXT: v_mov_b32_e32 v8, v5
; GFX942-NEXT: v_mov_b32_e32 v5, v4
; GFX942-NEXT: v_mov_b32_e32 v4, v3
; GFX942-NEXT: v_mov_b32_e32 v3, v2
; GFX942-NEXT: v_mov_b32_e32 v2, v1
; GFX942-NEXT: s_mov_b64 s[2:3], exec
; GFX942-NEXT: .LBB5_1: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: v_readfirstlane_b32 s4, v2
; GFX942-NEXT: v_readfirstlane_b32 s5, v3
; GFX942-NEXT: v_readfirstlane_b32 s6, v4
; GFX942-NEXT: v_readfirstlane_b32 s7, v5
; GFX942-NEXT: v_cmp_eq_u64_e32 vcc, s[4:5], v[2:3]
; GFX942-NEXT: v_readfirstlane_b32 s8, v7
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[4:5]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, s8, v7
; GFX942-NEXT: s_and_b64 s[0:1], s[0:1], vcc
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_atomic_pk_add_f16 v0, v[8:9], s[4:7], s8 idxen offen sc0
; GFX942-NEXT: ; implicit-def: $vgpr2_vgpr3_vgpr4_vgpr5
; GFX942-NEXT: ; implicit-def: $vgpr7
; GFX942-NEXT: ; implicit-def: $vgpr8_vgpr9
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB5_1
; GFX942-NEXT: ; %bb.2:
; GFX942-NEXT: s_mov_b64 exec, s[2:3]
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX1200-LABEL: struct_ptr_buffer_atomic_add_v2f16_rtn__vgpr_val__vgpr_rsrc__vgpr_voffset__vgpr_soffset:
; GFX1200: ; %bb.0:
; GFX1200-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1200-NEXT: s_wait_expcnt 0x0
; GFX1200-NEXT: s_wait_samplecnt 0x0
; GFX1200-NEXT: s_wait_bvhcnt 0x0
; GFX1200-NEXT: s_wait_kmcnt 0x0
; GFX1200-NEXT: s_mov_b32 s2, exec_lo
; GFX1200-NEXT: .LBB5_1: ; =>This Inner Loop Header: Depth=1
; GFX1200-NEXT: v_readfirstlane_b32 s4, v1
; GFX1200-NEXT: v_readfirstlane_b32 s5, v2
; GFX1200-NEXT: v_readfirstlane_b32 s6, v3
; GFX1200-NEXT: v_readfirstlane_b32 s7, v4
; GFX1200-NEXT: v_readfirstlane_b32 s3, v7
; GFX1200-NEXT: s_wait_alu 0xf1ff
; GFX1200-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[1:2]
; GFX1200-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX1200-NEXT: v_cmp_eq_u64_e64 s0, s[6:7], v[3:4]
; GFX1200-NEXT: v_cmp_eq_u32_e64 s1, s3, v7
; GFX1200-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX1200-NEXT: s_wait_alu 0xfffe
; GFX1200-NEXT: s_and_b32 s0, s0, s1
; GFX1200-NEXT: s_wait_alu 0xfffe
; GFX1200-NEXT: s_and_saveexec_b32 s0, s0
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: buffer_atomic_pk_add_f16 v0, v[5:6], s[4:7], s3 idxen offen th:TH_ATOMIC_RETURN
; GFX1200-NEXT: ; implicit-def: $vgpr1_vgpr2_vgpr3_vgpr4
; GFX1200-NEXT: ; implicit-def: $vgpr7
; GFX1200-NEXT: ; implicit-def: $vgpr5_vgpr6
; GFX1200-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX1200-NEXT: s_cbranch_execnz .LBB5_1
; GFX1200-NEXT: ; %bb.2:
; GFX1200-NEXT: s_mov_b32 exec_lo, s2
; GFX1200-NEXT: s_wait_loadcnt 0x0
; GFX1200-NEXT: s_setpc_b64 s[30:31]
%ret = call <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
ret <2 x half> %ret
}
declare float @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f32(float, ptr addrspace(8), i32, i32, i32, i32 immarg)
declare <2 x half> @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.v2f16(<2 x half>, ptr addrspace(8), i32, i32, i32, i32 immarg)
attributes #0 = { nounwind }
|
