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
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
| # RUN: llc --mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs -run-pass si-fold-operands,si-shrink-instructions %s -o - | FileCheck %s
--- |
define amdgpu_kernel void @add_f32_1.0_one_f16_use() #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f32.add = fadd float %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile float %f32.add, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f32_1.0_multi_f16_use() #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f32.add = fadd float %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile float %f32.add, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f32_1.0_one_f32_use_one_f16_use () #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f32.add = fadd float %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile float %f32.add, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f32_1.0_one_f32_use_multi_f16_use () #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f16.add1 = fadd half %f16.val1, 0xH3C00
%f32.add = fadd float %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile half %f16.add1, half addrspace(1)* undef
store volatile float %f32.add, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_i32_1_multi_f16_use() #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH0001
%f16.add1 = fadd half %f16.val1, 0xH0001
store volatile half %f16.add0, half addrspace(1)* undef
store volatile half %f16.add1,half addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_i32_m2_one_f32_use_multi_f16_use () #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xHFFFE
%f16.add1 = fadd half %f16.val1, 0xHFFFE
%f32.add = fadd float %f32.val, 0xffffffffc0000000
store volatile half %f16.add0, half addrspace(1)* undef
store volatile half %f16.add1, half addrspace(1)* undef
store volatile float %f32.add, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f16_1.0_multi_f32_use() #0 {
%f32.val0 = load volatile float, float addrspace(1)* undef
%f32.val1 = load volatile float, float addrspace(1)* undef
%f32.val = load volatile float, float addrspace(1)* undef
%f32.add0 = fadd float %f32.val0, 1.0
%f32.add1 = fadd float %f32.val1, 1.0
store volatile float %f32.add0, float addrspace(1)* undef
store volatile float %f32.add1, float addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f16_1.0_other_high_bits_multi_f16_use() #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile half, half addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f32.add = fadd half %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile half %f32.add, half addrspace(1)* undef
ret void
}
define amdgpu_kernel void @add_f16_1.0_other_high_bits_use_f16_f32() #0 {
%f16.val0 = load volatile half, half addrspace(1)* undef
%f16.val1 = load volatile half, half addrspace(1)* undef
%f32.val = load volatile half, half addrspace(1)* undef
%f16.add0 = fadd half %f16.val0, 0xH3C00
%f32.add = fadd half %f32.val, 1.000000e+00
store volatile half %f16.add0, half addrspace(1)* undef
store volatile half %f32.add, half addrspace(1)* undef
ret void
}
attributes #0 = { nounwind }
...
---
# f32 1.0 with a single use should be folded as the low 32-bits of a
# literal constant.
# CHECK-LABEL: name: add_f32_1.0_one_f16_use
# CHECK: %13:vgpr_32 = V_ADD_F16_e32 1065353216, killed %11, implicit $exec
name: add_f32_1.0_one_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = V_MOV_B32_e32 1065353216, implicit $exec
%13 = V_ADD_F16_e64 0, killed %11, 0, %12, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %13, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
S_ENDPGM 0
...
---
# Materialized f32 inline immediate should not be folded into the f16
# operands
# CHECK-LABEL: name: add_f32_1.0_multi_f16_use
# CHECK: %13:vgpr_32 = V_MOV_B32_e32 1065353216, implicit $exec
# CHECK: %14:vgpr_32 = V_ADD_F16_e32 killed %11, %13, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 killed %12, killed %13, implicit $exec
name: add_f32_1.0_multi_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%13 = V_MOV_B32_e32 1065353216, implicit $exec
%14 = V_ADD_F16_e64 0, killed %11, 0, %13, 0, 0, implicit $exec
%15 = V_ADD_F16_e64 0, killed %12, 0, killed %13, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %14, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
S_ENDPGM 0
...
---
# f32 1.0 should be folded into the single f32 use as an inline
# immediate, and folded into the single f16 use as a literal constant
# CHECK-LABEL: name: add_f32_1.0_one_f32_use_one_f16_use
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 1065353216, %11, implicit $exec
# CHECK: %16:vgpr_32 = V_ADD_F32_e32 1065353216, killed %13, implicit $exec
name: add_f32_1.0_one_f32_use_one_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
- { id: 16, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%13 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%14 = V_MOV_B32_e32 1065353216, implicit $exec
%15 = V_ADD_F16_e64 0, %11, 0, %14, 0, 0, implicit $exec
%16 = V_ADD_F32_e64 0, killed %13, 0, killed %14, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_DWORD_OFFSET killed %16, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
S_ENDPGM 0
...
---
# f32 1.0 should be folded for the single f32 use as an inline
# constant, and not folded as a multi-use literal for the f16 cases
# CHECK-LABEL: name: add_f32_1.0_one_f32_use_multi_f16_use
# CHECK: %14:vgpr_32 = V_MOV_B32_e32 1065353216, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 %11, %14, implicit $exec
# CHECK: %16:vgpr_32 = V_ADD_F16_e32 %12, %14, implicit $exec
# CHECK: %17:vgpr_32 = V_ADD_F32_e32 1065353216, killed %13, implicit $exec
name: add_f32_1.0_one_f32_use_multi_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
- { id: 16, class: vgpr_32 }
- { id: 17, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%13 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%14 = V_MOV_B32_e32 1065353216, implicit $exec
%15 = V_ADD_F16_e64 0, %11, 0, %14, 0, 0, implicit $exec
%16 = V_ADD_F16_e64 0, %12, 0, %14, 0, 0, implicit $exec
%17 = V_ADD_F32_e64 0, killed %13, 0, killed %14, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %16, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_DWORD_OFFSET killed %17, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
S_ENDPGM 0
...
---
# CHECK-LABEL: name: add_i32_1_multi_f16_use
# CHECK: %13:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
# CHECK: %14:vgpr_32 = V_ADD_F16_e32 1, killed %11, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 1, killed %12, implicit $exec
name: add_i32_1_multi_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%13 = V_MOV_B32_e32 1, implicit $exec
%14 = V_ADD_F16_e64 0, killed %11, 0, %13, 0, 0, implicit $exec
%15 = V_ADD_F16_e64 0, killed %12, 0, killed %13, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %14, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
S_ENDPGM 0
...
---
# CHECK-LABEL: name: add_i32_m2_one_f32_use_multi_f16_use
# CHECK: %14:vgpr_32 = V_MOV_B32_e32 -2, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 -2, %11, implicit $exec
# CHECK: %16:vgpr_32 = V_ADD_F16_e32 -2, %12, implicit $exec
# CHECK: %17:vgpr_32 = V_ADD_F32_e32 -2, killed %13, implicit $exec
name: add_i32_m2_one_f32_use_multi_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
- { id: 16, class: vgpr_32 }
- { id: 17, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%13 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%14 = V_MOV_B32_e32 -2, implicit $exec
%15 = V_ADD_F16_e64 0, %11, 0, %14, 0, 0, implicit $exec
%16 = V_ADD_F16_e64 0, %12, 0, %14, 0, 0, implicit $exec
%17 = V_ADD_F32_e64 0, killed %13, 0, killed %14, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %16, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_DWORD_OFFSET killed %17, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
S_ENDPGM 0
...
---
# f32 1.0 should be folded for the single f32 use as an inline
# constant, and not folded as a multi-use literal for the f16 cases
# CHECK-LABEL: name: add_f16_1.0_multi_f32_use
# CHECK: %13:vgpr_32 = V_MOV_B32_e32 15360, implicit $exec
# CHECK: %14:vgpr_32 = V_ADD_F32_e32 %11, %13, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F32_e32 %12, %13, implicit $exec
name: add_f16_1.0_multi_f32_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%12 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%13 = V_MOV_B32_e32 15360, implicit $exec
%14 = V_ADD_F32_e64 0, %11, 0, %13, 0, 0, implicit $exec
%15 = V_ADD_F32_e64 0, %12, 0, %13, 0, 0, implicit $exec
BUFFER_STORE_DWORD_OFFSET killed %14, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
BUFFER_STORE_DWORD_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
S_ENDPGM 0
...
---
# The low 16-bits are an inline immediate, but the high bits are junk
# FIXME: Should be able to fold this
# CHECK-LABEL: name: add_f16_1.0_other_high_bits_multi_f16_use
# CHECK: %13:vgpr_32 = V_MOV_B32_e32 80886784, implicit $exec
# CHECK: %14:vgpr_32 = V_ADD_F16_e32 %11, %13, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 %12, %13, implicit $exec
name: add_f16_1.0_other_high_bits_multi_f16_use
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%12 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%13 = V_MOV_B32_e32 80886784, implicit $exec
%14 = V_ADD_F16_e64 0, %11, 0, %13, 0, 0, implicit $exec
%15 = V_ADD_F16_e64 0, %12, 0, %13, 0, 0, implicit $exec
BUFFER_STORE_SHORT_OFFSET killed %14, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
S_ENDPGM 0
...
---
# FIXME: Should fold inline immediate into f16 and literal use into
# f32 instruction.
# CHECK-LABEL: name: add_f16_1.0_other_high_bits_use_f16_f32
# CHECK: %13:vgpr_32 = V_MOV_B32_e32 305413120, implicit $exec
# CHECK: %14:vgpr_32 = V_ADD_F32_e32 %11, %13, implicit $exec
# CHECK: %15:vgpr_32 = V_ADD_F16_e32 %12, %13, implicit $exec
name: add_f16_1.0_other_high_bits_use_f16_f32
alignment: 1
exposesReturnsTwice: false
legalized: false
regBankSelected: false
selected: false
tracksRegLiveness: true
registers:
- { id: 0, class: sreg_64 }
- { id: 1, class: sreg_32 }
- { id: 2, class: sgpr_32 }
- { id: 3, class: vgpr_32 }
- { id: 4, class: sreg_64 }
- { id: 5, class: sreg_32 }
- { id: 6, class: sreg_64 }
- { id: 7, class: sreg_32 }
- { id: 8, class: sreg_32 }
- { id: 9, class: sreg_32 }
- { id: 10, class: sgpr_128 }
- { id: 11, class: vgpr_32 }
- { id: 12, class: vgpr_32 }
- { id: 13, class: vgpr_32 }
- { id: 14, class: vgpr_32 }
- { id: 15, class: vgpr_32 }
frameInfo:
isFrameAddressTaken: false
isReturnAddressTaken: false
hasStackMap: false
hasPatchPoint: false
stackSize: 0
offsetAdjustment: 0
maxAlignment: 0
adjustsStack: false
hasCalls: false
maxCallFrameSize: 0
hasOpaqueSPAdjustment: false
hasVAStart: false
hasMustTailInVarArgFunc: false
body: |
bb.0 (%ir-block.0):
%4 = IMPLICIT_DEF
%5 = COPY %4.sub1
%6 = IMPLICIT_DEF
%7 = COPY %6.sub0
%8 = S_MOV_B32 61440
%9 = S_MOV_B32 -1
%10 = REG_SEQUENCE killed %7, 1, killed %5, 2, killed %9, 3, killed %8, 4
%11 = BUFFER_LOAD_DWORD_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 4 from `float addrspace(1)* undef`)
%12 = BUFFER_LOAD_USHORT_OFFSET %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile load 2 from `half addrspace(1)* undef`)
%13 = V_MOV_B32_e32 305413120, implicit $exec
%14 = V_ADD_F32_e64 0, %11, 0, %13, 0, 0, implicit $exec
%15 = V_ADD_F16_e64 0, %12, 0, %13, 0, 0, implicit $exec
BUFFER_STORE_DWORD_OFFSET killed %14, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 4 into `float addrspace(1)* undef`)
BUFFER_STORE_SHORT_OFFSET killed %15, %10, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (volatile store 2 into `half addrspace(1)* undef`)
S_ENDPGM 0
...
|