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
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
|
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX8 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX9 %s
--- |
define <2 x i16> @and_v2i16(<2 x i16> %a, <2 x i16> %b) #0 {
%and = and <2 x i16> %a, %b
ret <2 x i16> %and
}
define <3 x i16> @add_v3i16(<3 x i16> %a, <3 x i16> %b) #0 {
%add = add <3 x i16> %a, %b
ret <3 x i16> %add
}
define <3 x i16> @shl_v3i16(<3 x i16> %a, <3 x i16> %b) #0 {
%shl = shl <3 x i16> %a, %b
ret <3 x i16> %shl
}
define <4 x half> @fma_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
%fma = call <4 x half> @llvm.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c)
ret <4 x half> %fma
}
define amdgpu_ps <5 x half> @maxnum_v5i16(<5 x half> %a, <5 x half> %b) {
%fma = call <5 x half> @llvm.maxnum.v5f16(<5 x half> %a, <5 x half> %b)
ret <5 x half> %fma
}
declare <4 x half> @llvm.fma.v4f16(<4 x half>, <4 x half>, <4 x half>)
declare <5 x half> @llvm.maxnum.v5f16(<5 x half>, <5 x half>)
...
---
name: and_v2i16
body: |
bb.1:
liveins: $vgpr0, $vgpr1
; GFX8-LABEL: name: and_v2i16
; GFX8: liveins: $vgpr0, $vgpr1
; GFX8-NEXT: {{ $}}
; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX8-NEXT: [[AND:%[0-9]+]]:_(<2 x s16>) = G_AND [[COPY]], [[COPY1]]
; GFX8-NEXT: $vgpr0 = COPY [[AND]](<2 x s16>)
; GFX8-NEXT: SI_RETURN implicit $vgpr0
; GFX9-LABEL: name: and_v2i16
; GFX9: liveins: $vgpr0, $vgpr1
; GFX9-NEXT: {{ $}}
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9-NEXT: [[AND:%[0-9]+]]:_(<2 x s16>) = G_AND [[COPY]], [[COPY1]]
; GFX9-NEXT: $vgpr0 = COPY [[AND]](<2 x s16>)
; GFX9-NEXT: SI_RETURN implicit $vgpr0
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
%3:_(<2 x s16>) = G_AND %0, %1
$vgpr0 = COPY %3(<2 x s16>)
SI_RETURN implicit $vgpr0
...
---
name: add_v3i16
body: |
bb.1:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX8-LABEL: name: add_v3i16
; GFX8: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX8-NEXT: {{ $}}
; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
; GFX8-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC3]]
; GFX8-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[TRUNC4]]
; GFX8-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC5]]
; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ADD]](s16)
; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[ADD1]](s16)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[ADD2]](s16)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST4]](<2 x s16>)
; GFX8-NEXT: $vgpr1 = COPY [[BITCAST5]](<2 x s16>)
; GFX8-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
; GFX9-LABEL: name: add_v3i16
; GFX9: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX9-NEXT: {{ $}}
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX9-NEXT: [[ADD:%[0-9]+]]:_(<2 x s16>) = G_ADD [[COPY]], [[COPY2]]
; GFX9-NEXT: [[ADD1:%[0-9]+]]:_(<2 x s16>) = G_ADD [[BUILD_VECTOR]], [[BUILD_VECTOR1]]
; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[ADD1]](<2 x s16>)
; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[ADD]](<2 x s16>)
; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
%3:_(<2 x s16>) = COPY $vgpr0
%4:_(<2 x s16>) = COPY $vgpr1
%5:_(<4 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>)
%6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16) = G_UNMERGE_VALUES %5(<4 x s16>)
%0:_(<3 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16)
%10:_(<2 x s16>) = COPY $vgpr2
%11:_(<2 x s16>) = COPY $vgpr3
%12:_(<4 x s16>) = G_CONCAT_VECTORS %10(<2 x s16>), %11(<2 x s16>)
%13:_(s16), %14:_(s16), %15:_(s16), %16:_(s16) = G_UNMERGE_VALUES %12(<4 x s16>)
%1:_(<3 x s16>) = G_BUILD_VECTOR %13(s16), %14(s16), %15(s16)
%17:_(<3 x s16>) = G_ADD %0, %1
%21:_(s16), %22:_(s16), %23:_(s16) = G_UNMERGE_VALUES %17(<3 x s16>)
%24:_(s16) = G_IMPLICIT_DEF
%25:_(<4 x s16>) = G_BUILD_VECTOR %21(s16), %22(s16), %23(s16), %24(s16)
%19:_(<2 x s16>), %20:_(<2 x s16>) = G_UNMERGE_VALUES %25(<4 x s16>)
$vgpr0 = COPY %19(<2 x s16>)
$vgpr1 = COPY %20(<2 x s16>)
SI_RETURN implicit $vgpr0, implicit $vgpr1
...
---
name: shl_v3i16
body: |
bb.1:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX8-LABEL: name: shl_v3i16
; GFX8: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX8-NEXT: {{ $}}
; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SHL]](s16)
; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SHL1]](s16)
; GFX8-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SHL2]](s16)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]]
; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST4]](<2 x s16>)
; GFX8-NEXT: $vgpr1 = COPY [[BITCAST5]](<2 x s16>)
; GFX8-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
; GFX9-LABEL: name: shl_v3i16
; GFX9: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; GFX9-NEXT: {{ $}}
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[COPY]], [[COPY2]](<2 x s16>)
; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC1]](s16)
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[SHL1]](s16), [[DEF]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[SHL]](<2 x s16>)
; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR]](<2 x s16>)
; GFX9-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
%3:_(<2 x s16>) = COPY $vgpr0
%4:_(<2 x s16>) = COPY $vgpr1
%5:_(<4 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>)
%6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16) = G_UNMERGE_VALUES %5(<4 x s16>)
%0:_(<3 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16)
%10:_(<2 x s16>) = COPY $vgpr2
%11:_(<2 x s16>) = COPY $vgpr3
%12:_(<4 x s16>) = G_CONCAT_VECTORS %10(<2 x s16>), %11(<2 x s16>)
%13:_(s16), %14:_(s16), %15:_(s16), %16:_(s16) = G_UNMERGE_VALUES %12(<4 x s16>)
%1:_(<3 x s16>) = G_BUILD_VECTOR %13(s16), %14(s16), %15(s16)
%17:_(<3 x s16>) = G_SHL %0, %1(<3 x s16>)
%21:_(s16), %22:_(s16), %23:_(s16) = G_UNMERGE_VALUES %17(<3 x s16>)
%24:_(s16) = G_IMPLICIT_DEF
%25:_(<4 x s16>) = G_BUILD_VECTOR %21(s16), %22(s16), %23(s16), %24(s16)
%19:_(<2 x s16>), %20:_(<2 x s16>) = G_UNMERGE_VALUES %25(<4 x s16>)
$vgpr0 = COPY %19(<2 x s16>)
$vgpr1 = COPY %20(<2 x s16>)
SI_RETURN implicit $vgpr0, implicit $vgpr1
...
---
name: fma_v4f16
body: |
bb.1:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX8-LABEL: name: fma_v4f16
; GFX8: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX8-NEXT: {{ $}}
; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX8-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX8-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX8-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; GFX8-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
; GFX8-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32)
; GFX8-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; GFX8-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
; GFX8-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32)
; GFX8-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
; GFX8-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
; GFX8-NEXT: [[FMA:%[0-9]+]]:_(s16) = G_FMA [[TRUNC]], [[TRUNC4]], [[TRUNC8]]
; GFX8-NEXT: [[FMA1:%[0-9]+]]:_(s16) = G_FMA [[TRUNC1]], [[TRUNC5]], [[TRUNC9]]
; GFX8-NEXT: [[FMA2:%[0-9]+]]:_(s16) = G_FMA [[TRUNC2]], [[TRUNC6]], [[TRUNC10]]
; GFX8-NEXT: [[FMA3:%[0-9]+]]:_(s16) = G_FMA [[TRUNC3]], [[TRUNC7]], [[TRUNC11]]
; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMA]](s16)
; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMA1]](s16)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMA2]](s16)
; GFX8-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FMA3]](s16)
; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32)
; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; GFX8-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; GFX8-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
; GFX9-LABEL: name: fma_v4f16
; GFX9: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX9-NEXT: {{ $}}
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
; GFX9-NEXT: [[FMA:%[0-9]+]]:_(<2 x s16>) = G_FMA [[COPY]], [[COPY2]], [[COPY4]]
; GFX9-NEXT: [[FMA1:%[0-9]+]]:_(<2 x s16>) = G_FMA [[COPY1]], [[COPY3]], [[COPY5]]
; GFX9-NEXT: $vgpr0 = COPY [[FMA]](<2 x s16>)
; GFX9-NEXT: $vgpr1 = COPY [[FMA1]](<2 x s16>)
; GFX9-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
%4:_(<2 x s16>) = COPY $vgpr0
%5:_(<2 x s16>) = COPY $vgpr1
%0:_(<4 x s16>) = G_CONCAT_VECTORS %4(<2 x s16>), %5(<2 x s16>)
%6:_(<2 x s16>) = COPY $vgpr2
%7:_(<2 x s16>) = COPY $vgpr3
%1:_(<4 x s16>) = G_CONCAT_VECTORS %6(<2 x s16>), %7(<2 x s16>)
%8:_(<2 x s16>) = COPY $vgpr4
%9:_(<2 x s16>) = COPY $vgpr5
%2:_(<4 x s16>) = G_CONCAT_VECTORS %8(<2 x s16>), %9(<2 x s16>)
%10:_(<4 x s16>) = G_FMA %0, %1, %2
%12:_(<2 x s16>), %13:_(<2 x s16>) = G_UNMERGE_VALUES %10(<4 x s16>)
$vgpr0 = COPY %12(<2 x s16>)
$vgpr1 = COPY %13(<2 x s16>)
SI_RETURN implicit $vgpr0, implicit $vgpr1
...
---
name: maxnum_v5i16
body: |
bb.1:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX8-LABEL: name: maxnum_v5i16
; GFX8: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX8-NEXT: {{ $}}
; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX8-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX8-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; GFX8-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
; GFX8-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32)
; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; GFX8-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
; GFX8-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32)
; GFX8-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC]]
; GFX8-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC5]]
; GFX8-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
; GFX8-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC1]]
; GFX8-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC6]]
; GFX8-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
; GFX8-NEXT: [[FCANONICALIZE4:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC2]]
; GFX8-NEXT: [[FCANONICALIZE5:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC7]]
; GFX8-NEXT: [[FMAXNUM_IEEE2:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE4]], [[FCANONICALIZE5]]
; GFX8-NEXT: [[FCANONICALIZE6:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC3]]
; GFX8-NEXT: [[FCANONICALIZE7:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC8]]
; GFX8-NEXT: [[FMAXNUM_IEEE3:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE6]], [[FCANONICALIZE7]]
; GFX8-NEXT: [[FCANONICALIZE8:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC4]]
; GFX8-NEXT: [[FCANONICALIZE9:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC9]]
; GFX8-NEXT: [[FMAXNUM_IEEE4:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE8]], [[FCANONICALIZE9]]
; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE]](s16)
; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE1]](s16)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE2]](s16)
; GFX8-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE3]](s16)
; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32)
; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; GFX8-NEXT: [[ZEXT4:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE4]](s16)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT4]], [[SHL2]]
; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; GFX8-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; GFX8-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
; GFX8-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
; GFX9-LABEL: name: maxnum_v5i16
; GFX9: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX9-NEXT: {{ $}}
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[COPY]]
; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[COPY3]]
; GFX9-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[COPY1]]
; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[COPY4]]
; GFX9-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
; GFX9-NEXT: [[FCANONICALIZE4:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR]]
; GFX9-NEXT: [[FCANONICALIZE5:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR1]]
; GFX9-NEXT: [[FMAXNUM_IEEE2:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE4]], [[FCANONICALIZE5]]
; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE2]](<2 x s16>)
; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[FMAXNUM_IEEE]](<2 x s16>)
; GFX9-NEXT: $vgpr1 = COPY [[FMAXNUM_IEEE1]](<2 x s16>)
; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
%2:_(<2 x s16>) = COPY $vgpr0
%3:_(<2 x s16>) = COPY $vgpr1
%4:_(<2 x s16>) = COPY $vgpr2
%5:_(<6 x s16>) = G_CONCAT_VECTORS %2(<2 x s16>), %3(<2 x s16>), %4(<2 x s16>)
%6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16), %10:_(s16), %11:_(s16) = G_UNMERGE_VALUES %5(<6 x s16>)
%0:_(<5 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16), %9(s16), %10(s16)
%12:_(<2 x s16>) = COPY $vgpr3
%13:_(<2 x s16>) = COPY $vgpr4
%14:_(<2 x s16>) = COPY $vgpr5
%15:_(<6 x s16>) = G_CONCAT_VECTORS %12(<2 x s16>), %13(<2 x s16>), %14(<2 x s16>)
%16:_(s16), %17:_(s16), %18:_(s16), %19:_(s16), %20:_(s16), %21:_(s16) = G_UNMERGE_VALUES %15(<6 x s16>)
%1:_(<5 x s16>) = G_BUILD_VECTOR %16(s16), %17(s16), %18(s16), %19(s16), %20(s16)
%23:_(<5 x s16>) = G_FMAXNUM %0, %1
%27:_(s16), %28:_(s16), %29:_(s16), %30:_(s16), %31:_(s16) = G_UNMERGE_VALUES %23(<5 x s16>)
%32:_(s16) = G_IMPLICIT_DEF
%33:_(<6 x s16>) = G_BUILD_VECTOR %27(s16), %28(s16), %29(s16), %30(s16), %31(s16), %32(s16)
%24:_(<2 x s16>), %25:_(<2 x s16>), %26:_(<2 x s16>) = G_UNMERGE_VALUES %33(<6 x s16>)
$vgpr0 = COPY %24(<2 x s16>)
$vgpr1 = COPY %25(<2 x s16>)
$vgpr2 = COPY %26(<2 x s16>)
SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
...
|
