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
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
|
// RUN: mlir-opt %s -convert-linalg-to-loops | FileCheck %s
// RUN: mlir-opt %s -convert-linalg-to-parallel-loops | FileCheck --check-prefix=CHECKPARALLEL %s
// Test that we can lower all the way to LLVM without crashing, don't check results here.
// RUN: mlir-opt %s -convert-linalg-to-loops -convert-linalg-to-llvm='use-opaque-pointers=1' -o=/dev/null 2>&1
// CHECK: #[[$stride1Dilation1:.*]] = affine_map<(d0, d1) -> (d0 + d1)>
// CHECKPARALLEL: #[[$stride1Dilation1:.*]] = affine_map<(d0, d1) -> (d0 + d1)>
func.func @matmul(%arg0: memref<?xi8>, %M: index, %N: index, %K: index) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%A = memref.view %arg0[%c0][%M, %K] : memref<?xi8> to memref<?x?xf32>
%B = memref.view %arg0[%c0][%K, %N] : memref<?xi8> to memref<?x?xf32>
%C = memref.view %arg0[%c0][%M, %N] : memref<?xi8> to memref<?x?xf32>
linalg.matmul ins(%A, %B: memref<?x?xf32>, memref<?x?xf32>)
outs(%C: memref<?x?xf32>)
return
}
// CHECK-LABEL: func @matmul(%{{.*}}: memref<?xi8>,
// CHECK-SAME: [[M:arg[0-9]+]]: index
// CHECK-SAME: [[N:arg[0-9]+]]: index
// CHECK-SAME: [[K:arg[0-9]+]]: index
// CHECK: %[[A:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECK: %[[B:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECK: %[[C:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECK: scf.for {{.*}} to %[[M]]
// CHECK: scf.for {{.*}} to %[[N]]
// CHECK: scf.for {{.*}} to %[[K]]
// CHECK-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECK-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECK-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK-DAG: %[[c:.*]] = memref.load %[[C]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECK-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECK: store %[[res]], %[[C]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECKPARALLEL-LABEL: func @matmul(%{{.*}}: memref<?xi8>,
// CHECKPARALLEL-SAME: [[M:arg[0-9]+]]: index
// CHECKPARALLEL-SAME: [[N:arg[0-9]+]]: index
// CHECKPARALLEL-SAME: [[K:arg[0-9]+]]: index
// CHECKPARALLEL: %[[A:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECKPARALLEL: %[[B:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECKPARALLEL: %[[C:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECKPARALLEL: scf.parallel {{.*}} to (%[[M]], %[[N]]) step (%{{.*}}, %{{.*}} {
// CHECKPARALLEL: scf.for {{.*}} to %[[K]]
// CHECKPARALLEL-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECKPARALLEL-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECKPARALLEL-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL-DAG: %[[c:.*]] = memref.load %[[C]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECKPARALLEL-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[C]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
func.func @matvec(%arg0: memref<?xi8>, %M: index, %N: index) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%2 = memref.view %arg0[%c0][%M, %N] : memref<?xi8> to memref<?x?xf32>
%3 = memref.view %arg0[%c0][%M] : memref<?xi8> to memref<?xf32>
%4 = memref.view %arg0[%c0][%N] : memref<?xi8> to memref<?xf32>
linalg.matvec ins(%2, %3: memref<?x?xf32>, memref<?xf32>)
outs(%4 : memref<?xf32>)
return
}
// CHECK-LABEL: func @matvec(%{{.*}}: memref<?xi8>,
// CHECK-SAME: [[M:arg[0-9]+]]: index
// CHECK-SAME: [[K:arg[0-9]+]]: index
// CHECK: %[[A:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECK: %[[B:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECK: %[[C:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECK: scf.for {{.*}} to %[[M]]
// CHECK: scf.for {{.*}} to %[[K]]
// CHECK-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECK-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}] : memref<?xf32>
// CHECK-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK-DAG: %[[c:.*]] = memref.load %[[C]][%{{.*}}] : memref<?xf32>
// CHECK-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECK: store %[[res]], %[[C]][%{{.*}}] : memref<?xf32>
// CHECKPARALLEL-LABEL: func @matvec(%{{.*}}: memref<?xi8>,
// CHECKPARALLEL-SAME: [[M:arg[0-9]+]]: index
// CHECKPARALLEL-SAME: [[K:arg[0-9]+]]: index
// CHECKPARALLEL: %[[A:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?x?xf32>
// CHECKPARALLEL: %[[B:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECKPARALLEL: %[[C:.*]] = memref.view %{{.*}}[{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%[[M]]) step (%{{.*}}) {
// CHECKPARALLEL: scf.for {{.*}} to %[[K]]
// CHECKPARALLEL-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}, %{{.*}}] : memref<?x?xf32>
// CHECKPARALLEL-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}] : memref<?xf32>
// CHECKPARALLEL-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL-DAG: %[[c:.*]] = memref.load %[[C]][%{{.*}}] : memref<?xf32>
// CHECKPARALLEL-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[C]][%{{.*}}] : memref<?xf32>
func.func @dot(%arg0: memref<?xi8>, %M: index) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%1 = memref.view %arg0[%c0][%M] : memref<?xi8> to memref<?xf32>
%2 = memref.view %arg0[%c0][%M] : memref<?xi8> to memref<?xf32>
%3 = memref.view %arg0[%c0][] : memref<?xi8> to memref<f32>
linalg.dot ins(%1, %2 : memref<?xf32>, memref<?xf32>)
outs(%3 : memref<f32>)
return
}
// CHECK-LABEL: func @dot(%{{.*}}: memref<?xi8>,
// CHECK-SAME: [[K:arg[0-9]+]]: index
// CHECK: %[[A:.*]] = memref.view %{{.*}}[{{.*}}][{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECK: %[[B:.*]] = memref.view %{{.*}}[{{.*}}][{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECK: %[[C:.*]] = memref.view %{{.*}}[{{.*}}][] : memref<?xi8> to memref<f32>
// CHECK: scf.for {{.*}} to %[[K]]
// CHECK-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}] : memref<?xf32>
// CHECK-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}] : memref<?xf32>
// CHECK-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK-DAG: %[[c:.*]] = memref.load %[[C]][] : memref<f32>
// CHECK-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECK: store %[[res]], %[[C]][] : memref<f32>
// CHECKPARALLEL-LABEL: func @dot(%{{.*}}: memref<?xi8>,
// CHECKPARALLEL-SAME: [[K:arg[0-9]+]]: index
// CHECKPARALLEL: %[[A:.*]] = memref.view %{{.*}}[{{.*}}][{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECKPARALLEL: %[[B:.*]] = memref.view %{{.*}}[{{.*}}][{{.*}}] : memref<?xi8> to memref<?xf32>
// CHECKPARALLEL: %[[C:.*]] = memref.view %{{.*}}[{{.*}}][] : memref<?xi8> to memref<f32>
// CHECKPARALLEL: scf.for {{.*}} to %[[K]]
// CHECKPARALLEL-DAG: %[[a:.*]] = memref.load %[[A]][%{{.*}}] : memref<?xf32>
// CHECKPARALLEL-DAG: %[[b:.*]] = memref.load %[[B]][%{{.*}}] : memref<?xf32>
// CHECKPARALLEL-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL-DAG: %[[c:.*]] = memref.load %[[C]][] : memref<f32>
// CHECKPARALLEL-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[C]][] : memref<f32>
func.func @dot_int(%arg0: memref<?xi32>, %arg1: memref<?xi32>,
%arg3: memref<i32>) {
// Verifies that we use the correct arith operations for integers.
linalg.dot ins(%arg0, %arg1 : memref<?xi32>, memref<?xi32>)
outs(%arg3 : memref<i32>)
return
}
// CHECK-LABEL: func @dot_int(
// CHECK: %[[inc:.*]] = arith.muli {{.*}} : i32
// CHECK-NEXT: %[[res:.*]] = arith.addi {{.*}}, %[[inc]] : i32
// CHECK-NEXT: store %[[res]], {{.*}} : memref<i32>
func.func @dot_bool(%arg0: memref<?xi1>, %arg1: memref<?xi1>,
%arg3: memref<i1>) {
// Verifies that we use the correct (saturating) arith operations for booleans.
linalg.dot ins(%arg0, %arg1 : memref<?xi1>, memref<?xi1>)
outs(%arg3 : memref<i1>)
return
}
// CHECK-LABEL: func @dot_bool(
// CHECK: %[[inc:.*]] = arith.andi {{.*}} : i1
// CHECK-NEXT: %[[res:.*]] = arith.ori {{.*}}, %[[inc]] : i1
// CHECK-NEXT: store %[[res]], {{.*}} : memref<i1>
func.func @dot_view(%arg0: memref<?xf32, strided<[1], offset: ?>>, %arg1: memref<?xf32, strided<[1], offset: ?>>, %arg2: memref<f32>) {
linalg.dot ins(%arg0, %arg1 : memref<?xf32, strided<[1], offset: ?>>,
memref<?xf32, strided<[1], offset: ?>>)
outs(%arg2: memref<f32>)
return
}
// CHECK-LABEL: func @dot_view(
// CHECK: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<f32>) {
// CHECK: %[[K:.*]] = memref.dim %arg0, %c0 : memref<?xf32, strided<[1], offset: ?>>
// CHECK: scf.for {{.*}} to %[[K]]
// CHECK-DAG: %[[a:.*]] = memref.load %arg0[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECK-DAG: %[[b:.*]] = memref.load %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECK-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK-DAG: %[[c:.*]] = memref.load %{{.*}}[] : memref<f32>
// CHECK-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECK: store %[[res]], %{{.*}}[] : memref<f32>
// CHECKPARALLEL-LABEL: func @dot_view(
// CHECKPARALLEL: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<f32>) {
// CHECKPARALLEL: %[[K:.*]] = memref.dim %arg0, %c0 : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL: scf.for {{.*}} to %[[K]]
// CHECKPARALLEL-DAG: %[[a:.*]] = memref.load %arg0[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL-DAG: %[[b:.*]] = memref.load %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL-DAG: %[[inc:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL-DAG: %[[c:.*]] = memref.load %{{.*}}[] : memref<f32>
// CHECKPARALLEL-DAG: %[[res:.*]] = arith.addf %[[c]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %{{.*}}[] : memref<f32>
func.func @fill_view(%arg0: memref<?xf32, strided<[1], offset: ?>>, %arg1: f32) {
linalg.fill ins(%arg1 : f32) outs(%arg0 : memref<?xf32, strided<[1], offset: ?>>)
return
}
// CHECK-LABEL: func @fill_view(
// CHECK: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: f32) {
// CHECK: scf.for {{.*}} to %{{.*}}
// CHECK: store %{{.*}}, %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL-LABEL: func @fill_view(
// CHECKPARALLEL: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: f32) {
// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
// CHECKPARALLEL: store %{{.*}}, %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
func.func @fill_view0(%arg0: memref<f32>, %arg1: f32) {
linalg.fill ins(%arg1 : f32) outs(%arg0 : memref<f32>)
return
}
// CHECK-LABEL: func @fill_view0(%{{.*}}: memref<f32>, %{{.*}}: f32) {
// CHECK: store %{{.*}}, %{{.*}}[] : memref<f32>
// CHECKPARALLEL-LABEL: func @fill_view0(%{{.*}}: memref<f32>, %{{.*}}: f32) {
// CHECKPARALLEL: store %{{.*}}, %{{.*}}[] : memref<f32>
func.func @fill_view3(%arg0: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>, %arg1: f32) {
linalg.fill ins(%arg1 : f32) outs(%arg0 : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>)
return
}
// CHECK-LABEL: func @fill_view3(
// CHECK: %{{.*}}: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>, %{{.*}}: f32) {
// CHECK: scf.for {{.*}} to %{{.*}}
// CHECK: scf.for {{.*}} to %{{.*}}
// CHECK: scf.for {{.*}} to %{{.*}}
// CHECK: store %{{.*}}, {{.*}} : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECKPARALLEL-LABEL: func @fill_view3(
// CHECKPARALLEL: %{{.*}}: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>, %{{.*}}: f32) {
// CHECKPARALLEL: scf.parallel (%{{.*}}, %{{.*}}, %{{.*}}) = (%{{.*}}, %{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}}, %{{.*}}) {
// CHECKPARALLEL: store %{{.*}}, {{.*}} : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
func.func @copy_view(%arg0: memref<?xf32, strided<[1], offset: ?>>, %arg1: memref<?xf32, strided<[1], offset: ?>>) {
linalg.generic {
iterator_types = ["parallel"],
indexing_maps = [ affine_map<(i) -> (i)>, affine_map<(i) -> (i)>] }
ins(%arg0: memref<?xf32, strided<[1], offset: ?>>)
outs(%arg1: memref<?xf32, strided<[1], offset: ?>>) {
^bb0(%a: f32, %b: f32):
linalg.yield %a : f32
}
return
}
// CHECK-LABEL: func @copy_view(
// CHECK: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<?xf32, strided<[1], offset: ?>>) {
// CHECK: scf.for {{.*}} to %{{.*}}
// CHECK: %[[L:.*]] = memref.load %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECK: store %[[L]], %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL-LABEL: func @copy_view(
// CHECKPARALLEL: %{{.*}}: memref<?xf32, strided<[1], offset: ?>>, %{{.*}}: memref<?xf32, strided<[1], offset: ?>>) {
// CHECKPARALLEL: scf.parallel (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
// CHECKPARALLEL: %[[L:.*]] = memref.load %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
// CHECKPARALLEL: store %[[L]], %{{.*}}[%{{.*}}] : memref<?xf32, strided<[1], offset: ?>>
#accesses = [
affine_map<(i, j, k) -> (i, j)>,
affine_map<(i, j, k) -> (i, j, k)>,
affine_map<(i, j, k) -> (i, k, j)>
]
#trait2 = {
args_in = 1,
args_out = 2,
iterator_types = ["parallel", "parallel", "parallel"],
indexing_maps = #accesses,
library_call = "some_external_function_name_2",
doc = "B(i,j,k), C(i,k,j) = foo(A(i, j), B(i,j,k), C(i,k,j))"
}
func.func @generic_region(%arg0: memref<?x?xf32, strided<[?, 1], offset: ?>>, %arg1: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>, %arg2: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>) {
linalg.generic #trait2
ins(%arg0: memref<?x?xf32, strided<[?, 1], offset: ?>>)
outs(%arg1, %arg2 : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>,
memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>) {
^bb0(%a: f32, %b: f32, %c: f32):
%d = arith.mulf %a, %b : f32
%e = arith.addf %c, %d : f32
linalg.yield %d, %e : f32, f32
}
return
}
// CHECK-LABEL: @generic_region
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: scf.for %[[j:.*]] = {{.*}}
// CHECK: scf.for %[[k:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %{{.*}}[%[[i]], %[[j]]] : memref<?x?xf32, strided<[?, 1], offset: ?>>
// CHECK: %[[b:.*]] = memref.load %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECK: %[[c:.*]] = memref.load %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECK: %[[d:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK: %[[e:.*]] = arith.addf %[[c]], %[[d]] : f32
// CHECK: store %[[d]], %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECK: store %[[e]], %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECKPARALLEL-LABEL: @generic_region
// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]], %[[k:[a-zA-Z0-9_]*]])
// CHECKPARALLEL: %[[a:.*]] = memref.load %{{.*}}[%[[i]], %[[j]]] : memref<?x?xf32, strided<[?, 1], offset: ?>>
// CHECKPARALLEL: %[[b:.*]] = memref.load %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECKPARALLEL: %[[c:.*]] = memref.load %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECKPARALLEL: %[[d:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL: %[[e:.*]] = arith.addf %[[c]], %[[d]] : f32
// CHECKPARALLEL: store %[[d]], %{{.*}}[%[[i]], %[[j]], %[[k]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
// CHECKPARALLEL: store %[[e]], %{{.*}}[%[[i]], %[[k]], %[[j]]] : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>
#trait4 = {
args_in = 1,
args_out = 2,
iterator_types = ["parallel", "parallel", "parallel"],
indexing_maps = #accesses,
library_call = "some_external_function_name_2",
doc = "B(i,j,k), C(i,k,j) = foo(A(i, j) * B(i,j,k), i * j * k + C(i,k,j))"
}
func.func @generic_index_region(
%arg0: memref<?x?xf32, strided<[?, 1], offset: ?>>,
%arg1: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>,
%arg2: memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>) {
linalg.generic #trait4
ins(%arg0 : memref<?x?xf32, strided<[?, 1], offset: ?>>)
outs(%arg1, %arg2 : memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>,
memref<?x?x?xf32, strided<[?, ?, 1], offset: ?>>) {
^bb0(%a: f32, %b: f32, %c: f32):
%i = linalg.index 0 : index
%j = linalg.index 1 : index
%k = linalg.index 2 : index
%result_1 = arith.mulf %a, %b : f32
%ij = arith.addi %i, %j : index
%ijk = arith.addi %ij, %k : index
%ijk_int = arith.index_cast %ijk : index to i32
%ijk_float = arith.sitofp %ijk_int : i32 to f32
%result_2 = arith.addf %c, %ijk_float : f32
linalg.yield %result_1, %result_2 : f32, f32
}
return
}
// CHECK-LABEL: @generic_index_region
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: scf.for %[[j:.*]] = {{.*}}
// CHECK: scf.for %[[k:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %{{.*}}[%[[i]], %[[j]]]
// CHECK: %[[b:.*]] = memref.load %{{.*}}[%[[i]], %[[j]], %[[k]]]
// CHECK: %[[c:.*]] = memref.load %{{.*}}[%[[i]], %[[k]], %[[j]]]
// CHECK: %[[result_1:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECK: %[[ij:.*]] = arith.addi %[[i]], %[[j]] : index
// CHECK: %[[ijk:.*]] = arith.addi %[[ij]], %[[k]] : index
// CHECK: %[[ijk_int:.*]] = arith.index_cast %[[ijk]] : index to i32
// CHECK: %[[ijk_float:.*]] = arith.sitofp %[[ijk_int]] : i32 to f32
// CHECK: %[[result_2:.*]] = arith.addf %[[c]], %[[ijk_float]] : f32
// CHECK: store %[[result_1]], %{{.*}}[%[[i]], %[[j]], %[[k]]]
// CHECK: store %[[result_2]], %{{.*}}[%[[i]], %[[k]], %[[j]]]
// CHECKPARALLEL-LABEL: @generic_index_region
// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]], %[[k:[a-zA-Z0-9_]*]])
// CHECKPARALLEL: %[[a:.*]] = memref.load %{{.*}}[%[[i]], %[[j]]]
// CHECKPARALLEL: %[[b:.*]] = memref.load %{{.*}}[%[[i]], %[[j]], %[[k]]]
// CHECKPARALLEL: %[[c:.*]] = memref.load %{{.*}}[%[[i]], %[[k]], %[[j]]]
// CHECKPARALLEL: %[[result_1:.*]] = arith.mulf %[[a]], %[[b]] : f32
// CHECKPARALLEL: %[[ij:.*]] = arith.addi %[[i]], %[[j]] : index
// CHECKPARALLEL: %[[ijk:.*]] = arith.addi %[[ij]], %[[k]] : index
// CHECKPARALLEL: %[[ijk_int:.*]] = arith.index_cast %[[ijk]] : index to i32
// CHECKPARALLEL: %[[ijk_float:.*]] = arith.sitofp %[[ijk_int]] : i32 to f32
// CHECKPARALLEL: %[[result_2:.*]] = arith.addf %[[c]], %[[ijk_float]] : f32
// CHECKPARALLEL: store %[[result_1]], %{{.*}}[%[[i]], %[[j]], %[[k]]]
// CHECKPARALLEL: store %[[result_2]], %{{.*}}[%[[i]], %[[k]], %[[j]]]
// -----
#broadcast_access = [
affine_map<(i, j) -> ()>,
affine_map<(i, j) -> (i, j)>
]
#trait_broadcast = {
args_in = 1,
args_out = 1,
indexing_maps = #broadcast_access,
iterator_types = ["parallel", "parallel"],
library_call = "some_broadcast_external_fn"
}
func.func @generic_op_zero_rank(%arg0: memref<f32>, %arg1: memref<3x4xf32>)
{
linalg.generic #trait_broadcast
ins(%arg0 : memref<f32>)
outs(%arg1 : memref<3x4xf32>) {
^bb(%a: f32, %b: f32) :
linalg.yield %a : f32
}
return
}
// CHECK-LABEL: @generic_op_zero_rank
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32>
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: scf.for %[[j:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %[[ARG0]][]
// CHECK: store %[[a]], %[[ARG1]][%[[i]], %[[j]]]
// CHECKPARALLEL-LABEL: @generic_op_zero_rank
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32>
// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]])
// CHECKPARALLEL: %[[a:.*]] = memref.load %[[ARG0]][]
// CHECKPARALLEL: store %[[a]], %[[ARG1]][%[[i]], %[[j]]]
func.func @generic_op_scalar(%arg0: f32, %arg1: memref<3x4xf32>)
{
linalg.generic #trait_broadcast
ins(%arg0 : f32)
outs(%arg1 : memref<3x4xf32>) {
^bb(%a: f32, %b: f32) :
linalg.yield %a : f32
}
return
}
// CHECK-LABEL: @generic_op_scalar
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: f32
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32>
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: scf.for %[[j:.*]] = {{.*}}
// CHECK: store %[[ARG0]], %[[ARG1]][%[[i]], %[[j]]]
// CHECKPARALLEL-LABEL: @generic_op_scalar
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: f32
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xf32>
// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]])
// CHECKPARALLEL: store %[[ARG0]], %[[ARG1]][%[[i]], %[[j]]]
func.func @generic_index_op_zero_rank(%arg0: memref<i32>, %arg1: memref<3x4xi32>)
{
linalg.generic #trait_broadcast
ins(%arg0 : memref<i32>)
outs(%arg1 : memref<3x4xi32>) {
^bb(%a: i32, %b: i32) :
%i = linalg.index 0 : index
%j = linalg.index 1 : index
%ij = arith.addi %i, %j : index
%ij_int = arith.index_cast %ij : index to i32
%result = arith.addi %a, %ij_int : i32
linalg.yield %result : i32
}
return
}
// CHECK-LABEL: @generic_index_op_zero_rank
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<i32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xi32>
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: scf.for %[[j:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %[[ARG0]][
// CHECK: %[[ij:.*]] = arith.addi %[[i]], %[[j]] : index
// CHECK: %[[ij_int:.*]] = arith.index_cast %[[ij]] : index to i32
// CHECK: %[[result:.*]] = arith.addi %[[a]], %[[ij_int]] : i32
// CHECK: store %[[result]], %[[ARG1]][%[[i]], %[[j]]]
// CHECKPARALLEL-LABEL: @generic_index_op_zero_rank
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<i32>
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<3x4xi32>
// CHECKPARALLEL: scf.parallel (%[[i:[a-zA-Z0-9_]*]], %[[j:[a-zA-Z0-9_]*]])
// CHECKPARALLEL: %[[a:.*]] = memref.load %[[ARG0]][
// CHECKPARALLEL: %[[ij:.*]] = arith.addi %[[i]], %[[j]] : index
// CHECKPARALLEL: %[[ij_int:.*]] = arith.index_cast %[[ij]] : index to i32
// CHECKPARALLEL: %[[result:.*]] = arith.addi %[[a]], %[[ij_int]] : i32
// CHECKPARALLEL: store %[[result]], %[[ARG1]][%[[i]], %[[j]]]
#reduce_1D_access = [
affine_map<(i) -> (i)>,
affine_map<(i) -> ()>
]
#trait_reduce_1D = {
args_in = 1,
args_out = 1,
indexing_maps = #reduce_1D_access,
iterator_types = ["reduction"],
library_call = "some_reduce_external_fn"
}
func.func @generic_op_1D_reduce(%arg0: memref<?xf32>, %arg1: memref<f32>)
{
linalg.generic #trait_reduce_1D
ins(%arg0 : memref<?xf32>)
outs(%arg1 : memref<f32>) {
^bb(%a: f32, %b: f32) :
%0 = arith.addf %a, %b : f32
linalg.yield %0 : f32
}
return
}
// CHECK-LABEL: @generic_op_1D_reduce
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<?xf32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %[[ARG0]][%[[i]]]
// CHECK: %[[b:.*]] = memref.load %[[ARG1]][]
// CHECK: %[[c:.*]] = arith.addf %[[a]], %[[b]] : f32
// CHECK: store %[[c]], %[[ARG1]][]
// CHECKPARALLEL-LABEL: @generic_op_1D_reduce
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<?xf32>
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL: scf.for %[[i:.*]] = {{.*}}
// CHECKPARALLEL: %[[a:.*]] = memref.load %[[ARG0]][%[[i]]]
// CHECKPARALLEL: %[[b:.*]] = memref.load %[[ARG1]][]
// CHECKPARALLEL: %[[c:.*]] = arith.addf %[[a]], %[[b]] : f32
// CHECKPARALLEL: store %[[c]], %[[ARG1]][]
#reduce_init_1D_access = [
affine_map<(i) -> (i)>,
affine_map<(i) -> ()>,
affine_map<(i) -> ()>
]
#trait_reduce_init_1D = {
args_in = 2,
args_out = 1,
indexing_maps = #reduce_init_1D_access,
iterator_types = ["reduction"],
library_call = "some_reduce_external_fn"
}
func.func @generic_index_op_1D_reduce(%arg0: memref<?xf32>,
%arg1: memref<f32>,
%arg2: memref<f32>)
{
linalg.generic #trait_reduce_init_1D
ins(%arg0, %arg1 : memref<?xf32>, memref<f32>)
outs(%arg2 : memref<f32>) {
^bb(%a: f32, %b: f32, %c: f32) :
%i = linalg.index 0 : index
%0 = arith.constant 0 : index
%1 = arith.cmpi eq, %0, %i : index
%2 = arith.select %1, %b, %c : f32
%3 = arith.addf %a, %2 : f32
linalg.yield %3 : f32
}
return
}
// CHECK-LABEL: @generic_index_op_1D_reduce
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<?xf32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: %[[a:.*]] = memref.load %[[ARG0]][%[[i]]]
// CHECK: %[[b:.*]] = memref.load %[[ARG1]][]
// CHECK: %[[c:.*]] = memref.load %[[ARG2]][]
// CHECK: %[[d:.*]] = arith.select %{{.*}}, %[[b]], %[[c]]
// CHECK: %[[e:.*]] = arith.addf %[[a]], %[[d]]
// CHECK: store %[[e]], %[[ARG2]][]
// CHECKPARALLEL-LABEL: @generic_index_op_1D_reduce
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<?xf32>
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL: scf.for %[[i:.*]] = {{.*}}
// CHECKPARALLEL: %[[a:.*]] = memref.load %[[ARG0]][%[[i]]]
// CHECKPARALLEL: %[[b:.*]] = memref.load %[[ARG1]][]
// CHECKPARALLEL: %[[c:.*]] = memref.load %[[ARG2]][]
// CHECKPARALLEL: %[[d:.*]] = arith.select %{{.*}}, %[[b]], %[[c]]
// CHECKPARALLEL: %[[e:.*]] = arith.addf %[[a]], %[[d]]
// CHECKPARALLEL: store %[[e]], %[[ARG2]][]
#trait_const_fill = {
args_in = 0,
args_out = 1,
indexing_maps = [affine_map<(i) -> (i)>],
iterator_types = ["parallel"],
library_call = "some_external_fn"
}
func.func @generic_const_init(%arg0: memref<?xf32>) {
%cst = arith.constant 1.0 : f32
linalg.generic #trait_const_fill outs(%arg0 : memref<?xf32>) {
^bb0(%arg1: f32):
linalg.yield %cst : f32
}
return
}
// CHECK-LABEL: @generic_const_init
// CHECK-SAME: %[[ARG0:.*]]: memref<?xf32>
// CHECK: %[[CONST:.*]] = arith.constant 1.000000e+00 : f32
// CHECK: scf.for %[[i:.*]] = {{.*}}
// CHECK: store %[[CONST]], %[[ARG0]]
// CHECKPARALLEL-LABEL: @generic_const_init
// CHECKPARALLEL-SAME: %[[ARG0:.*]]: memref<?xf32>
// CHECKPARALLEL: %[[CONST:.*]] = arith.constant 1.000000e+00 : f32
// CHECKPARALLEL: scf.parallel (%[[i:.*]])
// CHECKPARALLEL: store %[[CONST]], %[[ARG0]]
#scalar_access = [
affine_map<() -> ()>,
affine_map<() -> ()>,
affine_map<() -> ()>
]
#scalar_trait = {
args_in = 2,
args_out = 1,
iterator_types = [],
indexing_maps = #scalar_access,
library_call = "some_external_fn"
}
func.func @scalar_code(%arg0: memref<f32>, %arg1 : memref<f32>, %arg2 : memref<f32>, %arg3 : i1)
{
linalg.generic #scalar_trait
ins(%arg0, %arg1 : memref<f32>, memref<f32>)
outs(%arg2 : memref<f32>) {
^bb(%a : f32, %b : f32, %c : f32) :
%result = scf.if %arg3 -> (f32) {
scf.yield %a : f32
} else {
scf.yield %b : f32
}
linalg.yield %result : f32
}
return
}
// CHECK-LABEL: @scalar_code
// CHECK-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref<f32>
// CHECK-NOT: scf.for
// CHECK: memref.load %[[ARG0]][]
// CHECK: memref.load %[[ARG1]][]
// CHECK: scf.if
// CHECK: scf.yield
// CHECK: else
// CHECK: scf.yield
// CHECK: store %{{.*}}, %[[ARG2]][]
// CHECKPARALLEL-LABEL: @scalar_code
// CHECKPARALLEL-SAME: %[[ARG0:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL-SAME: %[[ARG1:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL-SAME: %[[ARG2:[a-zA-Z0-9_]*]]: memref<f32>
// CHECKPARALLEL-NOT: scf.for
// CHECKPARALLEL: memref.load %[[ARG0]][]
// CHECKPARALLEL: memref.load %[[ARG1]][]
// CHECKPARALLEL: scf.if
// CHECKPARALLEL: scf.yield
// CHECKPARALLEL: else
// CHECKPARALLEL: scf.yield
// CHECKPARALLEL: store %{{.*}}, %[[ARG2]][]
//----------------------------------------------------------------------------//
// Named ops to loops.
//----------------------------------------------------------------------------//
func.func @named_batch_matmul(%A: memref<?x?x?xf32>, %B: memref<?x?x?xf32>, %C: memref<?x?x?xf32>) {
linalg.batch_matmul ins(%A, %B : memref<?x?x?xf32>, memref<?x?x?xf32>)
outs(%C : memref<?x?x?xf32>)
return
}
// CHECK-LABEL: @named_batch_matmul
// CHECK-SAME: %[[mA:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK-SAME: %[[mB:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK-SAME: %[[mC:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK: %[[B:.*]] = memref.dim %[[mA]], %c0 : memref<?x?x?xf32>
// CHECK: %[[M:.*]] = memref.dim %[[mA]], %c1 : memref<?x?x?xf32>
// CHECK: %[[K:.*]] = memref.dim %[[mA]], %c2 : memref<?x?x?xf32>
// CHECK: %[[N:.*]] = memref.dim %[[mB]], %c2 : memref<?x?x?xf32>
// CHECK: scf.for %[[b:.*]] = %{{.*}} to %[[B]]
// CHECK: scf.for %[[m:.*]] = %{{.*}} to %[[M]]
// CHECK: scf.for %[[n:.*]] = %{{.*}} to %[[N]]
// CHECK: scf.for %[[k:.*]] = %{{.*}} to %[[K]]
// CHECK: %[[va:.*]] = memref.load %[[mA]][%[[b]], %[[m]], %[[k]]] : memref<?x?x?xf32>
// CHECK: %[[vb:.*]] = memref.load %[[mB]][%[[b]], %[[k]], %[[n]]] : memref<?x?x?xf32>
// CHECK: %[[vc:.*]] = memref.load %[[mC]][%[[b]], %[[m]], %[[n]]] : memref<?x?x?xf32>
// CHECK: %[[inc:.*]] = arith.mulf %[[va]], %[[vb]] : f32
// CHECK: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECK: store %[[res]], %[[mC]][%[[b]], %[[m]], %[[n]]] : memref<?x?x?xf32>
// CHECKPARALLEL-LABEL: @named_batch_matmul
// CHECKPARALLEL-SAME: %[[mA:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL-SAME: %[[mB:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL-SAME: %[[mC:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL: %[[B:.*]] = memref.dim %[[mA]], %c0 : memref<?x?x?xf32>
// CHECKPARALLEL: %[[M:.*]] = memref.dim %[[mA]], %c1 : memref<?x?x?xf32>
// CHECKPARALLEL: %[[K:.*]] = memref.dim %[[mA]], %c2 : memref<?x?x?xf32>
// CHECKPARALLEL: %[[N:.*]] = memref.dim %[[mB]], %c2 : memref<?x?x?xf32>
// CHECKPARALLEL: scf.parallel (%[[b:.*]], %[[m:.*]], %[[n:.*]]) = ({{.*}}) to (%[[B]], %[[M]], %[[N]]) step ({{.*}}) {
// CHECKPARALLEL: scf.for %[[k:.*]] = %{{.*}} to %[[K]]
// CHECKPARALLEL: %[[va:.*]] = memref.load %[[mA]][%[[b]], %[[m]], %[[k]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[vb:.*]] = memref.load %[[mB]][%[[b]], %[[k]], %[[n]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[vc:.*]] = memref.load %[[mC]][%[[b]], %[[m]], %[[n]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[inc:.*]] = arith.mulf %[[va]], %[[vb]] : f32
// CHECKPARALLEL: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[mC]][%[[b]], %[[m]], %[[n]]] : memref<?x?x?xf32>
func.func @conv1d_no_symbols(%in : memref<?xf32>, %filter : memref<?xf32>, %out : memref<?xf32>) -> () {
linalg.conv_1d ins(%in, %filter : memref<?xf32>, memref<?xf32>)
outs(%out : memref<?xf32>)
return
}
// CHECK-LABEL: @conv1d_no_symbols
// CHECK-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECK-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECK-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECK: %[[c0:.*]] = arith.constant 0 : index
// CHECK: %[[c1:.*]] = arith.constant 1 : index
// CHECK: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?xf32>
// CHECK: %[[dim1:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?xf32>
// CHECK: scf.for %[[b:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] {
// CHECK: scf.for %[[m:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECK: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[b]], %[[m]])
// CHECK: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]]] : memref<?xf32>
// CHECK: %[[va:.*]] = memref.load %[[arg1]][%[[m]]] : memref<?xf32>
// CHECK: %[[vc:.*]] = memref.load %[[arg2]][%[[b]]] : memref<?xf32>
// CHECK: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECK: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECK: store %[[res]], %[[arg2]][%[[b]]] : memref<?xf32>
// CHECKPARALLEL-LABEL: @conv1d_no_symbols
// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?xf32>
// CHECKPARALLEL: %[[c0:.*]] = arith.constant 0 : index
// CHECKPARALLEL: %[[c1:.*]] = arith.constant 1 : index
// CHECKPARALLEL: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?xf32>
// CHECKPARALLEL: %[[dim1:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?xf32>
// CHECKPARALLEL: scf.parallel (%[[b:.*]]) = (%[[c0]]) to (%[[dim1]]) step (%[[c1]]) {
// CHECKPARALLEL: scf.for %[[m:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[b]], %[[m]])
// CHECKPARALLEL: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]]] : memref<?xf32>
// CHECKPARALLEL: %[[va:.*]] = memref.load %[[arg1]][%[[m]]] : memref<?xf32>
// CHECKPARALLEL: %[[vc:.*]] = memref.load %[[arg2]][%[[b]]] : memref<?xf32>
// CHECKPARALLEL: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECKPARALLEL: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[b]]] : memref<?xf32>
func.func @conv2d_no_symbols(%in : memref<?x?xf32>, %filter : memref<?x?xf32>, %out : memref<?x?xf32>) -> () {
linalg.conv_2d ins(%in, %filter : memref<?x?xf32>, memref<?x?xf32>)
outs(%out: memref<?x?xf32>)
return
}
// CHECK-LABEL: @conv2d_no_symbols
// CHECK-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECK-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECK-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECK: %[[c0:.*]] = arith.constant 0 : index
// CHECK: %[[c1:.*]] = arith.constant 1 : index
// CHECK: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?x?xf32>
// CHECK: %[[dim1:.*]] = memref.dim %[[arg1]], %[[c1]] : memref<?x?xf32>
// CHECK: %[[dim2:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?x?xf32>
// CHECK: %[[dim3:.*]] = memref.dim %[[arg2]], %[[c1]] : memref<?x?xf32>
// CHECK: scf.for %[[arg3:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] {
// CHECK: scf.for %[[arg4:.*]] = %[[c0]] to %[[dim3]] step %[[c1]] {
// CHECK: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECK: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] {
// CHECK: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg5]])
// CHECK: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg6]])
// CHECK: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]], %[[aff2]]] : memref<?x?xf32>
// CHECK: %[[va:.*]] = memref.load %[[arg1]][%[[arg5]], %[[arg6]]] : memref<?x?xf32>
// CHECK: %[[vc:.*]] = memref.load %[[arg2]][%[[arg3]], %[[arg4]]] : memref<?x?xf32>
// CHECK: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECK: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECK: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]]] : memref<?x?xf32>
// CHECKPARALLEL-LABEL: @conv2d_no_symbols
// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?x?xf32>
// CHECKPARALLEL: %[[c0:.*]] = arith.constant 0 : index
// CHECKPARALLEL: %[[c1:.*]] = arith.constant 1 : index
// CHECKPARALLEL: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?x?xf32>
// CHECKPARALLEL: %[[dim1:.*]] = memref.dim %[[arg1]], %[[c1]] : memref<?x?xf32>
// CHECKPARALLEL: %[[dim2:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?x?xf32>
// CHECKPARALLEL: %[[dim3:.*]] = memref.dim %[[arg2]], %[[c1]] : memref<?x?xf32>
// CHECKPARALLEL: scf.parallel (%[[arg3:.*]], %[[arg4:.*]]) = (%[[c0]], %[[c0]]) to (%[[dim2]], %[[dim3]]) step (%[[c1]], %[[c1]]) {
// CHECKPARALLEL: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECKPARALLEL: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] {
// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg5]])
// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg6]])
// CHECKPARALLEL: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]], %[[aff2]]] : memref<?x?xf32>
// CHECKPARALLEL: %[[va:.*]] = memref.load %[[arg1]][%[[arg5]], %[[arg6]]] : memref<?x?xf32>
// CHECKPARALLEL: %[[vc:.*]] = memref.load %[[arg2]][%[[arg3]], %[[arg4]]] : memref<?x?xf32>
// CHECKPARALLEL: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECKPARALLEL: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]]] : memref<?x?xf32>
func.func @conv3d_no_symbols(%in : memref<?x?x?xf32>, %filter : memref<?x?x?xf32>, %out : memref<?x?x?xf32>) -> () {
linalg.conv_3d ins(%in, %filter : memref<?x?x?xf32>, memref<?x?x?xf32>)
outs(%out : memref<?x?x?xf32>)
return
}
// CHECK-LABEL: @conv3d_no_symbols
// CHECK-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECK-DAG: %[[c2:.*]] = arith.constant 2 : index
// CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[c1:.*]] = arith.constant 1 : index
// CHECK: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?x?x?xf32>
// CHECK: %[[dim1:.*]] = memref.dim %[[arg1]], %[[c1]] : memref<?x?x?xf32>
// CHECK: %[[dim2:.*]] = memref.dim %[[arg1]], %[[c2]] : memref<?x?x?xf32>
// CHECK: %[[dim3:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?x?x?xf32>
// CHECK: %[[dim4:.*]] = memref.dim %[[arg2]], %[[c1]] : memref<?x?x?xf32>
// CHECK: %[[dim5:.*]] = memref.dim %[[arg2]], %[[c2]] : memref<?x?x?xf32>
// CHECK: scf.for %[[arg3:.*]] = %[[c0]] to %[[dim3]] step %[[c1]] {
// CHECK: scf.for %[[arg4:.*]] = %[[c0]] to %[[dim4]] step %[[c1]] {
// CHECK: scf.for %[[arg5:.*]] = %[[c0]] to %[[dim5]] step %[[c1]] {
// CHECK: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECK: scf.for %[[arg7:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] {
// CHECK: scf.for %[[arg8:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] {
// CHECK: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg6]])
// CHECK: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg7]])
// CHECK: %[[aff3:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg5]], %[[arg8]])
// CHECK: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]], %[[aff2]], %[[aff3]]] : memref<?x?x?xf32>
// CHECK: %[[va:.*]] = memref.load %[[arg1]][%[[arg6]], %[[arg7]], %[[arg8]]] : memref<?x?x?xf32>
// CHECK: %[[vc:.*]] = memref.load %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref<?x?x?xf32>
// CHECK: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECK: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECK: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref<?x?x?xf32>
// CHECKPARALLEL-LABEL: @conv3d_no_symbols
// CHECKPARALLEL-SAME: %[[arg0:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL-SAME: %[[arg1:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL-SAME: %[[arg2:[a-zA-Z0-9]+]]: memref<?x?x?xf32>
// CHECKPARALLEL-DAG: %[[c2:.*]] = arith.constant 2 : index
// CHECKPARALLEL-DAG: %[[c0:.*]] = arith.constant 0 : index
// CHECKPARALLEL-DAG: %[[c1:.*]] = arith.constant 1 : index
// CHECKPARALLEL: %[[dim0:.*]] = memref.dim %[[arg1]], %[[c0]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[dim1:.*]] = memref.dim %[[arg1]], %[[c1]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[dim2:.*]] = memref.dim %[[arg1]], %[[c2]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[dim3:.*]] = memref.dim %[[arg2]], %[[c0]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[dim4:.*]] = memref.dim %[[arg2]], %[[c1]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[dim5:.*]] = memref.dim %[[arg2]], %[[c2]] : memref<?x?x?xf32>
// CHECKPARALLEL: scf.parallel (%[[arg3:.*]], %[[arg4:.*]], %[[arg5:.*]]) = (%[[c0]], %[[c0]], %[[c0]]) to (%[[dim3]], %[[dim4]], %[[dim5]]) step (%[[c1]], %[[c1]], %[[c1]]) {
// CHECKPARALLEL: scf.for %[[arg6:.*]] = %[[c0]] to %[[dim0]] step %[[c1]] {
// CHECKPARALLEL: scf.for %[[arg7:.*]] = %[[c0]] to %[[dim1]] step %[[c1]] {
// CHECKPARALLEL: scf.for %[[arg8:.*]] = %[[c0]] to %[[dim2]] step %[[c1]] {
// CHECKPARALLEL: %[[aff:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg3]], %[[arg6]])
// CHECKPARALLEL: %[[aff2:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg4]], %[[arg7]])
// CHECKPARALLEL: %[[aff3:.*]] = affine.apply #[[$stride1Dilation1]](%[[arg5]], %[[arg8]])
// CHECKPARALLEL: %[[vb:.*]] = memref.load %[[arg0]][%[[aff]], %[[aff2]], %[[aff3]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[va:.*]] = memref.load %[[arg1]][%[[arg6]], %[[arg7]], %[[arg8]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[vc:.*]] = memref.load %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref<?x?x?xf32>
// CHECKPARALLEL: %[[inc:.*]] = arith.mulf %[[vb]], %[[va]] : f32
// CHECKPARALLEL: %[[res:.*]] = arith.addf %[[vc]], %[[inc]] : f32
// CHECKPARALLEL: store %[[res]], %[[arg2]][%[[arg3]], %[[arg4]], %[[arg5]]] : memref<?x?x?xf32>
// -----
func.func @lower_to_loops_with_rank_reducing_subviews(
%arg0 : memref<?xi32>, %arg1 : memref<?x?xi32>, %arg2 : index,
%arg3 : index, %arg4 : index) {
%0 = memref.subview %arg0[%arg2] [%arg3] [1]
: memref<?xi32> to memref<?xi32, strided<[1], offset: ?>>
%1 = memref.subview %arg1[0, %arg4] [1, %arg3] [1, 1]
: memref<?x?xi32> to memref<?xi32, strided<[1], offset: ?>>
linalg.generic {
iterator_types = ["parallel"],
indexing_maps = [affine_map<(i) -> (i)>, affine_map<(i) -> (i)>]}
ins(%0: memref<?xi32, strided<[1], offset: ?>>)
outs(%1: memref<?xi32, strided<[1], offset: ?>>) {
^bb0(%a: i32, %b: i32):
linalg.yield %a : i32
}
return
}
// CHECK-LABEL: func @lower_to_loops_with_rank_reducing_subviews
// CHECK: scf.for %[[IV:.+]] = %{{.+}} to %{{.+}} step %{{.+}} {
// CHECK: %[[VAL:.+]] = memref.load %{{.+}}[%[[IV]]]
// CHECK: memref.store %[[VAL]], %{{.+}}[%[[IV]]]
// CHECK: }
// CHECKPARALLEL-LABEL: func @lower_to_loops_with_rank_reducing_subviews
// CHECKPARALLEL: scf.parallel (%[[IV:.+]]) = (%{{.+}}) to (%{{.+}}) step (%{{.+}}) {
// CHECKPARALLEL: %[[VAL:.+]] = memref.load %{{.+}}[%[[IV]]]
// CHECKPARALLEL: memref.store %[[VAL]], %{{.+}}[%[[IV]]]
// CHECKPARALLEL: }
|
