aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/llvm14/lib/Target/AArch64/AArch64SchedAmpere1.td
blob: 32f7299fbf87b2fddf31a75705b504b011bae877 (plain) (blame)
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
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
//=- AArch64SchedAmpere1.td - Ampere-1 scheduling def -----*- tablegen -*-=//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file defines the machine model for the Ampere Computing Ampere-1 to
// support instruction scheduling and other instruction cost heuristics.
//
//===----------------------------------------------------------------------===//

// The Ampere-1 core is an out-of-order micro-architecture.  The front
// end has branch prediction, with a 10-cycle recovery time from a
// mispredicted branch.  Instructions coming out of the front end are
// decoded into internal micro-ops (uops).

def Ampere1Model : SchedMachineModel {
  let IssueWidth            =   4;  // 4-way decode and dispatch
  let MicroOpBufferSize     = 174;  // micro-op re-order buffer size
  let LoadLatency           =   4;  // Optimistic load latency
  let MispredictPenalty     =  10;  // Branch mispredict penalty
  let LoopMicroOpBufferSize =  32;  // Instruction queue size
  let CompleteModel = 1;

  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
                                                    SMEUnsupported.F);
}

let SchedModel = Ampere1Model in {

//===----------------------------------------------------------------------===//
// Define each kind of processor resource and number available on Ampere-1.
// Ampere-1 has 12 pipelines that 8 independent scheduler (4 integer, 2 FP,
// and 2 memory) issue into.  The integer and FP schedulers can each issue
// one uop per cycle, while the memory schedulers can each issue one load
// and one store address calculation per cycle.

def Ampere1UnitA  : ProcResource<2>;  // integer single-cycle, branch, and flags r/w
def Ampere1UnitB  : ProcResource<2>;  // integer single-cycle, and complex shifts
def Ampere1UnitBS : ProcResource<1>;  // integer multi-cycle
def Ampere1UnitL  : ProcResource<2>;  // load
def Ampere1UnitS  : ProcResource<2>;  // store address calculation
def Ampere1UnitX  : ProcResource<1>;  // FP and vector operations, and flag write
def Ampere1UnitY  : ProcResource<1>;  // FP and vector operations, and crypto
def Ampere1UnitZ  : ProcResource<1>;  // FP store data and FP-to-integer moves

def Ampere1UnitAB : ProcResGroup<[Ampere1UnitA, Ampere1UnitB]>;
def Ampere1UnitXY : ProcResGroup<[Ampere1UnitX, Ampere1UnitY]>;

//===----------------------------------------------------------------------===//
// Define customized scheduler read/write types specific to the Ampere-1.

def Ampere1Write_1cyc_1A : SchedWriteRes<[Ampere1UnitA]> {
  let Latency = 1;
  let NumMicroOps = 1;
}

def Ampere1Write_1cyc_2A : SchedWriteRes<[Ampere1UnitA, Ampere1UnitA]> {
  let Latency = 1;
  let NumMicroOps = 2;
}

def Ampere1Write_1cyc_1B : SchedWriteRes<[Ampere1UnitB]> {
  let Latency = 1;
  let NumMicroOps = 1;
}

def Ampere1Write_1cyc_1AB : SchedWriteRes<[Ampere1UnitAB]> {
  let Latency = 1;
  let NumMicroOps = 1;
}

def Ampere1Write_1cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
  let Latency = 1;
  let NumMicroOps = 1;
}

def Ampere1Write_1cyc_1S : SchedWriteRes<[Ampere1UnitS]> {
  let Latency = 1;
  let NumMicroOps = 1;
}

def Ampere1Write_1cyc_2S : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS]> {
  let Latency = 1;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1Y : SchedWriteRes<[Ampere1UnitY]> {
  let Latency = 2;
  let NumMicroOps = 1;
}

def Ampere1Write_2cyc_2AB : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitAB]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1B_1AB : SchedWriteRes<[Ampere1UnitB, Ampere1UnitAB]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1B_1A : SchedWriteRes<[Ampere1UnitB, Ampere1UnitA]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1AB_1A : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitA]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1AB_2S : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitS,
                                                             Ampere1UnitS]> {
  let Latency = 2;
  let NumMicroOps = 3;
}

def Ampere1Write_2cyc_1AB_1S_1Z : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitS,
                                                                Ampere1UnitZ]> {
  let Latency = 2;
  let NumMicroOps = 3;
}

def Ampere1Write_2cyc_1B_1S : SchedWriteRes<[Ampere1UnitB, Ampere1UnitS]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_2cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 2;
  let NumMicroOps = 1;
}

def Ampere1Write_2cyc_1S_1Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitZ]> {
  let Latency = 2;
  let NumMicroOps = 2;
}

def Ampere1Write_3cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 3;
  let NumMicroOps = 1;
}

def Ampere1Write_3cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 3;
  let NumMicroOps = 1;
}

def Ampere1Write_3cyc_1B_1S_1AB : SchedWriteRes<[Ampere1UnitB, Ampere1UnitS,
                                                               Ampere1UnitAB]> {
  let Latency = 2;
  let NumMicroOps = 3;
}

def Ampere1Write_3cyc_1S_2Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 2;
  let NumMicroOps = 3;
}

def Ampere1Write_3cyc_2S_2Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS,
                                             Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 2;
  let NumMicroOps = 4;
}

def Ampere1Write_4cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_1X : SchedWriteRes<[Ampere1UnitX]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_1Y : SchedWriteRes<[Ampere1UnitY]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_1Z : SchedWriteRes<[Ampere1UnitZ]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_2L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL]> {
  let Latency = 4;
  let NumMicroOps = 2;
}

def Ampere1Write_4cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 4;
  let NumMicroOps = 1;
}

def Ampere1Write_4cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 4;
  let NumMicroOps = 2;
}

def Ampere1Write_4cyc_1XY_1S_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitS, Ampere1UnitZ]> {
  let Latency = 4;
  let NumMicroOps = 3;
}

def Ampere1Write_4cyc_3S_3Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS, Ampere1UnitS,
                                             Ampere1UnitZ, Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 4;
  let NumMicroOps = 6;
}

def Ampere1Write_5cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
  let Latency = 5;
  let NumMicroOps = 2;
}

def Ampere1Write_5cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 5;
  let NumMicroOps = 1;
}

def Ampere1Write_5cyc_1X : SchedWriteRes<[Ampere1UnitX]> {
  let Latency = 5;
  let NumMicroOps = 1;
}

def Ampere1Write_5cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
  let Latency = 5;
  let NumMicroOps = 1;
}

def Ampere1Write_5cyc_2L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL]> {
  let Latency = 5;
  let NumMicroOps = 2;
}

def Ampere1Write_5cyc_1L_1BS : SchedWriteRes<[Ampere1UnitL, Ampere1UnitBS]> {
  let Latency = 5;
  let NumMicroOps = 2;
}

def Ampere1Write_5cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 5;
  let NumMicroOps = 1;
}

def Ampere1Write_5cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 5;
  let NumMicroOps = 2;
}

def Ampere1Write_5cyc_4S_4Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS,
                                             Ampere1UnitS, Ampere1UnitS,
                                             Ampere1UnitZ, Ampere1UnitZ,
                                             Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 5;
  let NumMicroOps = 8;
}

def Ampere1Write_5cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 5;
  let NumMicroOps = 6;
}

def Ampere1Write_6cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 6;
  let NumMicroOps = 6;
}

def Ampere1Write_6cyc_3XY_3S_3Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 6;
  let NumMicroOps = 9;
}

def Ampere1Write_6cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
  let Latency = 6;
  let NumMicroOps = 2;
}

def Ampere1Write_6cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 6;
  let NumMicroOps = 1;
}

def Ampere1Write_6cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 6;
  let NumMicroOps = 2;
}

def Ampere1Write_6cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 6;
  let NumMicroOps = 3;
}

def Ampere1Write_6cyc_3L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL]> {
  let Latency = 6;
  let NumMicroOps = 3;
}

def Ampere1Write_6cyc_4L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                          Ampere1UnitL, Ampere1UnitL]> {
  let Latency = 6;
  let NumMicroOps = 4;
}

def Ampere1Write_6cyc_1XY_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitZ]> {
  let Latency = 6;
  let NumMicroOps = 2;
}

def Ampere1Write_7cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 7;
  let NumMicroOps = 1;
}

def Ampere1Write_7cyc_1BS_1XY : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitXY]> {
  let Latency = 7;
  let NumMicroOps = 2;
}

def Ampere1Write_7cyc_1L_1XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitXY]> {
  let Latency = 7;
  let NumMicroOps = 2;
}

def Ampere1Write_7cyc_2L_2XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 7;
  let NumMicroOps = 4;
}

def Ampere1Write_7cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 7;
  let NumMicroOps = 2;
}

def Ampere1Write_7cyc_4XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ,
                                                 Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 7;
  let NumMicroOps = 12;
}

def Ampere1Write_8cyc_1BS_1A : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitA]> {
  let Latency = 8;
  let NumMicroOps = 2;
}

def Ampere1Write_8cyc_1BS_2A : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitA,
                                                             Ampere1UnitA]> {
  let Latency = 8;
  let NumMicroOps = 3;
}

def Ampere1Write_8cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 8;
  let NumMicroOps = 2;
}

def Ampere1Write_8cyc_4XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                           Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 8;
  let NumMicroOps = 4;
}

def Ampere1Write_8cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 8;
  let NumMicroOps = 6;
}

def Ampere1Write_8cyc_4L_4XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY,
                                              Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 8;
  let NumMicroOps = 8;
}

def Ampere1Write_9cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 9;
  let NumMicroOps = 6;
}

def Ampere1Write_9cyc_4L_4XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY,
                                              Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 9;
  let NumMicroOps = 8;
}

def Ampere1Write_9cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 9;
  let NumMicroOps = 3;
}

def Ampere1Write_9cyc_2L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                              Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 9;
  let NumMicroOps = 5;
}

def Ampere1Write_9cyc_6XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ,
                                                 Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 9;
  let NumMicroOps = 14;
}

def Ampere1Write_9cyc_8XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitXY, Ampere1UnitXY,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitS, Ampere1UnitS,
                                                 Ampere1UnitZ, Ampere1UnitZ,
                                                 Ampere1UnitZ, Ampere1UnitZ]> {
  let Latency = 9;
  let NumMicroOps = 16;
}

def Ampere1Write_10cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 10;
  let NumMicroOps = 2;
}

def Ampere1Write_10cyc_1XY_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitZ]> {
  let Latency = 10;
  let NumMicroOps = 2;
}

def Ampere1Write_10cyc_1X_1Z : SchedWriteRes<[Ampere1UnitX, Ampere1UnitZ]> {
  let Latency = 10;
  let NumMicroOps = 2;
}

def Ampere1Write_10cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
                                               Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 10;
  let NumMicroOps = 6;
}

def Ampere1Write_10cyc_1A_1BS_1X : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitX]> {
  let Latency = 10;
  let NumMicroOps = 3;
}

def Ampere1Write_10cyc_1A_1BS_1XY : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitXY]> {
  let Latency = 10;
  let NumMicroOps = 3;
}

def Ampere1Write_11cyc_1BS_1L : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitL]> {
  let Latency = 11;
  let NumMicroOps = 2;
}

def Ampere1Write_11cyc_1A_1BS_1X : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitX]> {
  let Latency = 11;
  let NumMicroOps = 3;
}

def Ampere1Write_11cyc_1A_1BS_1XY : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitXY]> {
  let Latency = 11;
  let NumMicroOps = 3;
}

def Ampere1Write_11cyc_4L_8XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                               Ampere1UnitL, Ampere1UnitL,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 11;
  let NumMicroOps = 12;
}

def Ampere1Write_12cyc_4L_8XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
                                               Ampere1UnitL, Ampere1UnitL,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY,
                                               Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 12;
  let NumMicroOps = 12;
}

def Ampere1Write_12cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 12;
  let NumMicroOps = 3;
}

def Ampere1Write_12cyc_4XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
                                            Ampere1UnitXY, Ampere1UnitXY]> {
  let Latency = 12;
  let NumMicroOps = 4;
}

def Ampere1Write_18cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 18;
  let NumMicroOps = 1;
}

def Ampere1Write_19cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 19;
  let NumMicroOps = 1;
}

def Ampere1Write_25cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 25;
  let NumMicroOps = 1;
}

def Ampere1Write_32cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 32;
  let NumMicroOps = 1;
}

def Ampere1Write_34cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
  let Latency = 34;
  let NumMicroOps = 1;
}

def Ampere1Write_34cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 34;
  let NumMicroOps = 1;
}

def Ampere1Write_39cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 39;
  let NumMicroOps = 1;
}

def Ampere1Write_62cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
  let Latency = 62;
  let NumMicroOps = 1;
}

// For basic arithmetic, we have more flexibility for short shifts (LSL shift <= 4),
// which are a single uop, and for extended registers, which have full flexibility
// across Unit A or B for both uops.
def Ampere1Write_Arith : SchedWriteVariant<[
                                SchedVar<RegExtendedPred, [Ampere1Write_2cyc_2AB]>,
                                SchedVar<AmpereCheapLSL,  [Ampere1Write_1cyc_1AB]>,
                                SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1AB]>]>;

def Ampere1Write_ArithFlagsetting : SchedWriteVariant<[
                                SchedVar<RegExtendedPred, [Ampere1Write_2cyc_1AB_1A]>,
                                SchedVar<AmpereCheapLSL,  [Ampere1Write_1cyc_1A]>,
                                SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1A]>]>;

//===----------------------------------------------------------------------===//
// Map the target-defined scheduler read/write resources and latencies for Ampere-1.
// This provides a coarse model, which is then specialised below.

def : WriteRes<WriteImm,   [Ampere1UnitAB]>;  // MOVN, MOVZ
def : WriteRes<WriteI,     [Ampere1UnitAB]>;  // ALU
def : WriteRes<WriteISReg, [Ampere1UnitB, Ampere1UnitA]> {
  let Latency = 2;
  let NumMicroOps = 2;
}  // ALU of Shifted-Reg
def : WriteRes<WriteIEReg, [Ampere1UnitAB, Ampere1UnitA]> {
  let Latency = 2;
  let NumMicroOps = 2;
}  // ALU of Extended-Reg
def : WriteRes<WriteExtr,  [Ampere1UnitB]>;  // EXTR shifts a reg pair
def : WriteRes<WriteIS,    [Ampere1UnitB]>;  // Shift/Scale
def : WriteRes<WriteID32,  [Ampere1UnitBS]> {
  let Latency = 18;
}  // 32-bit Divide
def : WriteRes<WriteID64,  [Ampere1UnitBS]> {
  let Latency = 34;
}  // 64-bit Divide
def : WriteRes<WriteIM32,  [Ampere1UnitBS]> {
  let Latency = 3;
}  // 32-bit Multiply
def : WriteRes<WriteIM64,  [Ampere1UnitBS]> {
  let Latency = 3;
}  // 32-bit Multiply
def : WriteRes<WriteBr,    [Ampere1UnitA]>;
def : WriteRes<WriteBrReg, [Ampere1UnitA, Ampere1UnitA]>;
def : WriteRes<WriteLD,    [Ampere1UnitL]> {
  let Latency = 4;
}  // Load from base addr plus immediate offset
def : WriteRes<WriteST,    [Ampere1UnitS]> {
  let Latency = 1;
}  // Store to base addr plus immediate offset
def : WriteRes<WriteSTP,   [Ampere1UnitS, Ampere1UnitS]> {
  let Latency = 1;
  let NumMicroOps = 2;
}  // Store a register pair.
def : WriteRes<WriteAdr,   [Ampere1UnitAB]>;
def : WriteRes<WriteLDIdx, [Ampere1UnitAB, Ampere1UnitS]> {
  let Latency = 5;
  let NumMicroOps = 2;
}  // Load from a register index (maybe scaled).
def : WriteRes<WriteSTIdx, [Ampere1UnitS, Ampere1UnitS]> {
  let Latency = 1;
  let NumMicroOps = 2;
}  // Store to a register index (maybe scaled).
def : WriteRes<WriteF,  [Ampere1UnitXY]> {
  let Latency = 2;
}  // General floating-point ops.
def : WriteRes<WriteFCmp,  [Ampere1UnitX]> {
  let Latency = 5;
}  // Floating-point compare.
def : WriteRes<WriteFCvt,  [Ampere1UnitXY]> {
  let Latency = 6;
}  // Float conversion.
def : WriteRes<WriteFCopy, [Ampere1UnitXY]> {
}  // Float-int register copy.
def : WriteRes<WriteFImm,  [Ampere1UnitXY]> {
  let Latency = 2;
}  // Float-int register copy.
def : WriteRes<WriteFMul,  [Ampere1UnitXY]> {
  let Latency = 5;
}  // Floating-point multiply.
def : WriteRes<WriteFDiv,  [Ampere1UnitXY]> {
  let Latency = 34;
}  // Floating-point division.
def : WriteRes<WriteVd,    [Ampere1UnitXY]> {
  let Latency = 3;
}  // 64bit Vector D ops.
def : WriteRes<WriteVq,    [Ampere1UnitXY]> {
  let Latency = 3;
}  // 128bit Vector Q ops.
def : WriteRes<WriteVLD,   [Ampere1UnitL, Ampere1UnitL]> {
  let Latency = 5;
}  // Vector loads.
def : WriteRes<WriteVST,   [Ampere1UnitS, Ampere1UnitZ]> {
  let Latency = 2;
}  // Vector stores.

def : WriteRes<WriteAtomic,  []> { let Unsupported = 1; }

def : WriteRes<WriteSys,     []> { let Latency = 1; }
def : WriteRes<WriteBarrier, []> { let Latency = 1; }
def : WriteRes<WriteHint,    []> { let Latency = 1; }

def : WriteRes<WriteLDHi,    []> {
  let Latency = 4;
}  // The second register of a load-pair: LDP,LDPSW,LDNP,LDXP,LDAXP

// Forwarding logic.
def : ReadAdvance<ReadI,       0>;
def : ReadAdvance<ReadISReg,   0>;
def : ReadAdvance<ReadIEReg,   0>;
def : ReadAdvance<ReadIM,      0>;
def : ReadAdvance<ReadIMA,     1, [WriteIM32, WriteIM64]>;
def : ReadAdvance<ReadID,      0>;
def : ReadAdvance<ReadExtrHi,  0>;
def : ReadAdvance<ReadST,      0>;
def : ReadAdvance<ReadAdrBase, 0>;
def : ReadAdvance<ReadVLD,     0>;

//===----------------------------------------------------------------------===//
// Specialising the scheduling model further for Ampere-1.

def : InstRW<[Ampere1Write_1cyc_1AB], (instrs COPY)>;

// Branch instructions
def : InstRW<[Ampere1Write_1cyc_1A], (instrs Bcc, BL, RET)>;
def : InstRW<[Ampere1Write_1cyc_1A],
        (instrs CBZW, CBZX, CBNZW, CBNZX, TBZW, TBZX, TBNZW, TBNZX)>;
def : InstRW<[Ampere1Write_1cyc_2A], (instrs BLR)>;

// Cryptography instructions
// -- AES encryption/decryption
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^AES[DE]")>;
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^AESI?MC")>;
// -- Polynomial multiplication
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^PMUL", "^PMULL")>;
// -- SHA-256 hash
def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA256(H|H2)")>;
// -- SHA-256 schedule update
def : InstRW<[Ampere1Write_4cyc_1Y], (instregex "^SHA256SU[01]")>;
// -- SHA-3 instructions
def : InstRW<[Ampere1Write_2cyc_1XY],
        (instregex "^BCAX", "^EOR3", "^RAX1", "^XAR")>;
// -- SHA-512 hash
def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA512(H|H2)")>;
// -- SHA-512 schedule update
def : InstRW<[Ampere1Write_4cyc_1Y], (instregex "^SHA512SU[01]")>;
// -- SHA1 choose/majority/parity
def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA1[CMP]")>;
// -- SHA1 hash/schedule update
def : InstRW<[Ampere1Write_2cyc_1Y], (instregex "^SHA1SU[01]")>;
def : InstRW<[Ampere1Write_2cyc_1Y], (instregex "^SHA1H")>;

// FP and vector load instructions
// -- Load 1-element structure to one/all lanes
// ---- all lanes
def : InstRW<[Ampere1Write_7cyc_1L_1XY],
        (instregex "^LD1Rv(8b|4h|2s|16b|8h|4s|2d)")>;
// ---- one lane
def : InstRW<[Ampere1Write_7cyc_1L_1XY],
        (instregex "^LD1i(8|16|32|64)")>;
// -- Load 1-element structure to one/all lanes, 1D size
def : InstRW<[Ampere1Write_5cyc_1L],
        (instregex "^LD1Rv1d")>;
// -- Load 1-element structures to 1 register
def : InstRW<[Ampere1Write_5cyc_1L],
        (instregex "^LD1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Load 1-element structures to 2 registers
def : InstRW<[Ampere1Write_5cyc_2L],
        (instregex "^LD1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Load 1-element structures to 3 registers
def : InstRW<[Ampere1Write_6cyc_3L],
        (instregex "^LD1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Load 1-element structures to 4 registers
def : InstRW<[Ampere1Write_6cyc_4L],
        (instregex "^LD1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Load 2-element structure to all lanes of 2 registers, 1D size
def : InstRW<[Ampere1Write_5cyc_2L],
        (instregex "^LD2Rv1d")>;
// -- Load 2-element structure to all lanes of 2 registers, other sizes
def : InstRW<[Ampere1Write_7cyc_2L_2XY],
        (instregex "^LD2Rv(8b|4h|2s|16b|8h|4s|2d)")>;
// -- Load 2-element structure to one lane of 2 registers
def : InstRW<[Ampere1Write_7cyc_2L_2XY],
        (instregex "^LD2i(8|16|32|64)")>;
// -- Load 2-element structures to 2 registers, 16B/8H/4S/2D size
def : InstRW<[Ampere1Write_7cyc_2L_2XY],
        (instregex "^LD2Twov(16b|8h|4s|2d)")>;
// -- Load 2-element structures to 2 registers, 8B/4H/2S size
def : InstRW<[Ampere1Write_9cyc_2L_3XY],
        (instregex "^LD2Twov(8b|4h|2s)")>;
// -- Load 3-element structure to all lanes of 3 registers, 1D size
def : InstRW<[Ampere1Write_6cyc_3L],
        (instregex "^LD3Rv1d")>;
// -- Load 3-element structure to all lanes of 3 registers, other sizes
def : InstRW<[Ampere1Write_8cyc_3L_3XY],
        (instregex "^LD3Rv(8b|4h|2s|16b|8h|4s|2d)")>;
// -- Load 3-element structure to one lane of 3 registers
def : InstRW<[Ampere1Write_8cyc_3L_3XY],
        (instregex "^LD3i(8|16|32|64)")>;
// -- Load 3-element structures to 3 registers, 16B/8H/4S sizes
def : InstRW<[Ampere1Write_9cyc_3L_3XY],
        (instregex "^LD3Threev(16b|8h|4s)")>;
// -- Load 3-element structures to 3 registers, 2D size
def : InstRW<[Ampere1Write_8cyc_3L_3XY],
        (instregex "^LD3Threev2d")>;
// -- Load 3-element structures to 3 registers, 8B/4H/2S sizes
def : InstRW<[Ampere1Write_10cyc_3L_3XY],
        (instregex "^LD3Threev(8b|4h|2s)")>;
// -- Load 4-element structure to all lanes of 4 registers, 1D size
def : InstRW<[Ampere1Write_6cyc_4L],
        (instregex "^LD4Rv1d")>;
// -- Load 4-element structure to all lanes of 4 registers, other sizes
def : InstRW<[Ampere1Write_8cyc_4L_4XY],
        (instregex "^LD4Rv(8b|4h|2s|16b|8h|4s|2d)")>;
// -- Load 4-element structure to one lane of 4 registers
def : InstRW<[Ampere1Write_6cyc_4L],
        (instregex "^LD4i(8|16|32|64)")>;
// -- Load 4-element structures to 4 registers, 2D size
def : InstRW<[Ampere1Write_9cyc_4L_4XY],
        (instregex "^LD4Fourv2d")>;
// -- Load 4-element structures to 4 registers, 2S size
def : InstRW<[Ampere1Write_12cyc_4L_8XY],
        (instregex "^LD4Fourv2s")>;
// -- Load 4-element structures to 4 registers, other sizes
def : InstRW<[Ampere1Write_11cyc_4L_8XY],
        (instregex "^LD4Fourv(8b|4h|16b|8h|4s)")>;
// -- Load pair, Q-form
def : InstRW<[Ampere1Write_5cyc_2L], (instregex "LDN?PQ")>;
// -- Load pair, S/D-form
def : InstRW<[Ampere1Write_5cyc_1L_1BS], (instregex "LDN?P(S|D)")>;
// -- Load register
def : InstRW<[Ampere1Write_5cyc_1L], (instregex "LDU?R[BHSDQ]i")>;
// -- Load register, sign-extended register
def : InstRW<[Ampere1Write_6cyc_1AB_1L], (instregex "LDR[BHSDQ]ro(W|X)")>;

// FP and vector store instructions
// -- Store 1-element structure from one lane of 1 register
def : InstRW<[Ampere1Write_4cyc_1XY_1S_1Z],
        (instregex "^ST1i(8|16|32|64)")>;
// -- Store 1-element structures from 1 register
def : InstRW<[Ampere1Write_2cyc_1S_1Z],
        (instregex "^ST1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Store 1-element structures from 2 registers
def : InstRW<[Ampere1Write_3cyc_2S_2Z],
        (instregex "^ST1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Store 1-element structures from 3 registers
def : InstRW<[Ampere1Write_4cyc_3S_3Z],
        (instregex "^ST1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Store 1-element structures from 4 registers
def : InstRW<[Ampere1Write_5cyc_4S_4Z],
        (instregex "^ST1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Store 2-element structure from one lane of 2 registers
def : InstRW<[Ampere1Write_5cyc_2XY_2S_2Z],
        (instregex "^ST2i(8|16|32|64)")>;
// -- Store 2-element structures from 2 registers, 16B/8H/4S/2D sizes
def : InstRW<[Ampere1Write_5cyc_2XY_2S_2Z],
        (instregex "^ST2Twov(16b|8h|4s|2d)")>;
// -- Store 2-element structures from 2 registers, 8B/4H/2S sizes
def : InstRW<[Ampere1Write_6cyc_2XY_2S_2Z],
        (instregex "^ST2Twov(8b|4h|2s)")>;
// -- Store 3-element structure from one lane of 3 registers
def : InstRW<[Ampere1Write_6cyc_3XY_3S_3Z],
        (instregex "^ST3i(8|16|32|64)")>;
// -- Store 3-element structures from 3 registers
def : InstRW<[Ampere1Write_6cyc_3XY_3S_3Z],
        (instregex "^ST3Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
// -- Store 4-element structure from one lane of 4 registers
def : InstRW<[Ampere1Write_7cyc_4XY_4S_4Z],
        (instregex "^ST4i(8|16|32|64)")>;
// -- Store 4-element structures from 4 registers, 16B/8H/4S sizes
def : InstRW<[Ampere1Write_9cyc_8XY_4S_4Z],
        (instregex "^ST4Fourv(16b|8h|4s)")>;
// -- Store 4-element structures from 4 registers, 2D sizes
def : InstRW<[Ampere1Write_7cyc_4XY_4S_4Z],
        (instregex "^ST4Fourv2d")>;
// -- Store 4-element structures from 4 registers, 8B/4H/2S sizes
def : InstRW<[Ampere1Write_9cyc_6XY_4S_4Z],
        (instregex "^ST4Fourv(8b|4h|2s)")>;
// -- Store pair, Q-form
def : InstRW<[Ampere1Write_3cyc_2S_2Z], (instregex "^STN?PQ")>;
// -- Store pair, S/D-form
def : InstRW<[Ampere1Write_3cyc_1S_2Z],	(instregex "^STN?P[SD]")>;
// -- Store register
def : InstRW<[Ampere1Write_2cyc_1S_1Z],	(instregex "^STU?R[BHSDQ](ui|i)")>;
// -- Store register, sign-extended register offset
def : InstRW<[Ampere1Write_2cyc_1AB_1S_1Z], (instregex "^STR[BHSDQ]ro[XW]")>;

// FP data processing, bfloat16 format
def : InstRW<[Ampere1Write_5cyc_1XY], (instrs BFCVT)>;
def : InstRW<[Ampere1Write_7cyc_2XY], (instrs BFCVTN, BFCVTN2)>;
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^BFDOTv", "^BF16DOT")>;
def : InstRW<[Ampere1Write_4cyc_2XY], (instrs BFMMLA)>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^BFMLAL")>;

// FP data processing, scalar/vector, half precision
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(ABD|ABS)v.[fi]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY],
        (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY],
        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY],
        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)16")>;
def : InstRW<[Ampere1Write_4cyc_1X],
        (instregex "^FCMPE?H")>;
def : InstRW<[Ampere1Write_10cyc_1A_1BS_1X],
        (instregex "^FCCMPE?H")>;
def : InstRW<[Ampere1Write_10cyc_1A_1BS_1XY],
        (instregex "^FCSELH")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^[SU]CVTFv.[fi]16")>;
def : InstRW<[Ampere1Write_25cyc_1XY], (instregex "^FDIVv.[if]16", "FDIVH")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if]16")>;
def : InstRW<[Ampere1Write_8cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv4[if]16")>;
def : InstRW<[Ampere1Write_12cyc_3XY], (instregex "^F(MAX|MIN)(NM)?Vv8[if]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FMULX?v.[if]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instrs FMULX16)>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FN?M(ADD|SUB)[H]rrr")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FML[AS]v.[if]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FRECPXv.[if]16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(RECP|RSQRT)S16")>;
def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if]16")>;
def : InstRW<[Ampere1Write_39cyc_1XY], (instregex "^FSQRTv.f16", "^FSQRTHr")>;

// FP data processing, scalar/vector, single/double precision
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^F(ABD|ABS)v.[fi](32|64)")>;
def : InstRW<[Ampere1Write_5cyc_1XY],
        (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi](32|64)")>;
def : InstRW<[Ampere1Write_5cyc_1XY],
        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi](32|64)")>;
def : InstRW<[Ampere1Write_5cyc_1XY],
        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)(32|64)")>;
def : InstRW<[Ampere1Write_5cyc_1X],
        (instregex "^FCMPE?(S|D)")>;
def : InstRW<[Ampere1Write_11cyc_1A_1BS_1X],
        (instregex "^FCCMPE?(S|D)")>;
def : InstRW<[Ampere1Write_11cyc_1A_1BS_1XY],
        (instregex "^FCSEL(S|D)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if](32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^[SU]CVTFv.[fi](32|64)")>;
def : InstRW<[Ampere1Write_34cyc_1XY], (instregex "^FDIVv.[if](64)", "FDIVD")>;
def : InstRW<[Ampere1Write_19cyc_1XY], (instregex "^FDIVv.[if](32)", "FDIVS")>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if](32|64)")>;
def : InstRW<[Ampere1Write_10cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv.[if](32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FMULX?v.[if](32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instrs FMULX32, FMULX64)>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FN?M(ADD|SUB)[SD]rrr")>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FML[AS]v.[if](32|64)")>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FRECPXv.[if](32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^F(RECP|RSQRT)S(32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if](32|64)")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FRINT(32|64)")>;
def : InstRW<[Ampere1Write_62cyc_1XY], (instregex "^FSQRTv.f64", "^FSQRTDr")>;
def : InstRW<[Ampere1Write_32cyc_1XY], (instregex "^FSQRTv.f32", "^FSQRTSr")>;

// FP miscellaneous instructions
def : InstRW<[Ampere1Write_10cyc_1XY_1Z], (instregex "^FCVT[AMNPZ][SU][SU][XW][HSD]r")>;
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FCVT[HSD]Hr")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVT[HSD][SD]r")>;
def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVTLv")>;
def : InstRW<[Ampere1Write_8cyc_2XY], (instregex "^FCVT(N|XN)v")>;
def : InstRW<[Ampere1Write_10cyc_1X_1Z], (instrs FJCVTZS)>;
def : InstRW<[Ampere1Write_5cyc_1BS], (instregex "^FMOV[HSD][WX]r")>;
def : InstRW<[Ampere1Write_7cyc_1BS_1XY], (instregex "^FMOVDXHighr")>;
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^FMOV[HSD][ri]")>;
def : InstRW<[Ampere1Write_6cyc_1XY_1Z], (instregex "^FMOVXDHighr")>;
def : InstRW<[Ampere1Write_4cyc_1Z], (instregex "^FMOV[WX][HSD]r")>;

// Integer arithmetic and logical instructions
def : InstRW<[Ampere1Write_1cyc_1A],
        (instregex "ADC(W|X)r", "SBC(W|X)r")>;
def : InstRW<[Ampere1Write_Arith],
        (instregex "(ADD|AND|BIC|EON|EOR|ORN|ORR|SUB)(W|X)r")>;
def : InstRW<[Ampere1Write_ArithFlagsetting],
        (instregex "(ADD|AND|BIC|SUB)S(W|X)r")>;
def : InstRW<[Ampere1Write_1cyc_1A],
        (instregex "(ADC|SBC)S(W|X)r")>;
def : InstRW<[Ampere1Write_1cyc_1A], (instrs RMIF)>;
def : InstRW<[Ampere1Write_1cyc_1A],
        (instregex "(CCMN|CCMP)(X|W)")>;
def : InstRW<[Ampere1Write_1cyc_1A],
        (instregex "(CSEL|CSINC|CSINV|CSNEG)(X|W)")>;
def : InstRW<[Ampere1Write_18cyc_1BS], (instrs SDIVWr, UDIVWr)>;
def : InstRW<[Ampere1Write_34cyc_1BS], (instrs SDIVXr, UDIVXr)>;
def : InstRW<[Ampere1Write_3cyc_1BS],
        (instregex "(S|U)MULHr")>;
def : InstRW<[Ampere1Write_4cyc_1BS],
        (instregex "(S|U)?M(ADD|SUB)L?r")>;

// Integer load instructions
def : InstRW<[Ampere1Write_4cyc_2L],
        (instregex "(LDNP|LDP|LDPSW)(X|W)")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDR(B|D|H|Q|S)ui")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDR(D|Q|W|X)l")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDTR(B|H|W|X)i")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDTRS(BW|BX|HW|HX|W)i")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDUR(BB|HH|X|W)i")>;
def : InstRW<[Ampere1Write_4cyc_1L],
        (instregex "LDURS(BW|BX|HW|HX|W)i")>;
def : InstRW<[Ampere1Write_5cyc_1AB_1L],
        (instregex "LDR(HH|SHW|SHX|W|X)ro(W|X)")>;
def : InstRW<[Ampere1Write_1cyc_1L],
        (instrs PRFMl, PRFUMi, PRFUMi)>;
def : InstRW<[Ampere1Write_2cyc_1AB_1L],
        (instrs PRFMroW, PRFMroX)>;

// Integer miscellaneous instructions
def : InstRW<[Ampere1Write_1cyc_1A],  (instrs ADR, ADRP)>;
def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "EXTR(W|X)")>;
def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "(S|U)?BFM(W|X)")>;
def : InstRW<[Ampere1Write_3cyc_1BS], (instregex "^CRC32C?[BHWX]")>;
def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "CLS(W|X)")>;
def : InstRW<[Ampere1Write_1cyc_1A],  (instrs SETF8, SETF16)>;
def : InstRW<[Ampere1Write_1cyc_1AB],
        (instrs MOVKWi, MOVKXi, MOVNWi, MOVNXi, MOVZWi, MOVZXi)>;
def : InstRW<[Ampere1Write_1cyc_1B],
        (instregex "(RBIT|REV|REV16)(W|X)r", "REV32Xr")>;
def : InstRW<[Ampere1Write_1cyc_1B],
        (instregex "(ASR|LSL|LSR|ROR)V(W|X)r")>;

// Integer store instructions
def : InstRW<[Ampere1Write_1cyc_2S],  (instregex "STNP(X|W)i")>;
def : InstRW<[Ampere1Write_2cyc_1B_1S],
        (instrs STPWi, STPXi)>;
def : InstRW<[Ampere1Write_3cyc_1B_1S_1AB],
        (instregex "STP(W|X)(pre|post)")>;
def : InstRW<[Ampere1Write_1cyc_1S],
        (instrs STTRBi, STTRHi, STTRWi, STTRXi)>;
def : InstRW<[Ampere1Write_1cyc_1S],
        (instregex "STUR(BB|HH|X|W)i",
                   "STR(X|W)ui",
                   "STUR(BB|HH|X|W)i")>;
def : InstRW<[Ampere1Write_1cyc_2S], (instrs STRWroX, STRXroX)>;
def : InstRW<[Ampere1Write_2cyc_1AB_2S], (instrs STRWroW, STRXroW)>;

// Pointer authentication
//def : InstRW<[Ampere1Write_7cyc_1BS],
//	(instrs AUTIAZ, AUTIBZ, AUTIASP, AUTIBSP, AUTIA1716, AUTIB1716)>;
def : InstRW<[Ampere1Write_8cyc_1BS_1A],
        (instregex "BRA(A|AZ|B|BZ)", "RETA(A|B)", "ERETA(A|B)")>;
def : InstRW<[Ampere1Write_8cyc_1BS_2A],
        (instrs BLRAA, BLRAAZ, BLRAB, BLRABZ)>;
//def : InstRW<[Ampere1Write_7cyc_1BS],
//	(instrs PACIAZ, PACIBZ, PACIASP, PACIBSP, PACIA1716, PACIB1716)>;
def : InstRW<[Ampere1Write_11cyc_1BS_1L], (instregex "^LDRA(A|B)")>;
def : InstRW<[Ampere1Write_7cyc_1BS], (instrs XPACD, XPACI)>;

// Vector integer instructions
// -- absolute difference
def : InstRW<[Ampere1Write_3cyc_1XY],
             (instregex "^SABAv", "^SABALv", "^SABDv", "^SABDLv",
                        "^UABAv", "^UABALv", "^UABDv", "^UABDLv")>;
// -- arithmetic
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^ABSv", "^(ADD|SUB)v", "^SADDLv", "^SADDW", "SHADD",
                   "SHSUB", "^SRHADD", "^URHADD", "SSUBL", "SSUBW",
                   "^UADDLv", "^UADDW", "UHADD", "UHSUB", "USUBL", "USUBW")>;
// -- arithmetic, horizontal, 16B
def : InstRW<[Ampere1Write_12cyc_4XY],
            (instregex "^ADDVv16i8v", "^SADDLVv16i8v", "^UADDLVv16i8v")>;
def : InstRW<[Ampere1Write_12cyc_4XY],
            (instregex "^[SU](MIN|MAX)Vv16i8v")>;
// -- arithmetic, horizontal, 4H/4S
def : InstRW<[Ampere1Write_6cyc_2XY],
            (instregex "^[SU]?ADDL?V(v8i8|v4i16|v2i32)v")>;
def : InstRW<[Ampere1Write_6cyc_2XY],
            (instregex "^[SU](MIN|MAX)V(v4i16|v4i32)v")>;
// -- arithmetic, horizontal, 8B/8H
def : InstRW<[Ampere1Write_9cyc_3XY],
            (instregex "^[SU]?ADDL?V(v8i16|v4i32)v")>;
def : InstRW<[Ampere1Write_9cyc_3XY],
            (instregex "^[SU](MIN|MAX)V(v8i8|v8i16)v")>;
// -- arithmetic, narrowing
def : InstRW<[Ampere1Write_5cyc_2XY], (instregex "(ADD|SUB)HNv.*")>;
def : InstRW<[Ampere1Write_5cyc_2XY], (instregex "(RADD|RSUB)HNv.*")>;
// -- arithmetic, pairwise
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^ADDPv", "^SADALP", "^UADALP", "^SADDLPv", "^UADDLPv")>;
// -- arithmetic, saturating
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^SQADD", "^SQSUB", "^SUQADD", "^UQADD", "^UQSUB", "^USQADD")>;
// -- bit count
def : InstRW<[Ampere1Write_2cyc_1XY],
        (instregex "^(CLS|CLZ|CNT)v")>;
// -- compare
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^CMEQv", "^CMGEv", "^CMGTv", "^CMLEv", "^CMLTv",
                   "^CMHIv", "^CMHSv")>;
// -- compare non-zero
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^CMTSTv")>;
// -- dot product
def : InstRW<[Ampere1Write_3cyc_1XY], (instregex "^(S|SU|U|US)DOTv")>;
// -- fp reciprocal estimate
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FRECPEv", "^FRSQRTEv")>;
// -- integer reciprocal estimate
def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^URECPEv", "^URSQRTEv")>;
// -- logical
def : InstRW<[Ampere1Write_2cyc_1XY],
        (instregex "^ANDv", "^BICv", "^EORv", "^ORRv", "^ORNv", "^NOTv")>;
// -- logical, narrowing
def : InstRW<[Ampere1Write_5cyc_2XY],
        (instregex "RSHRNv",
                   "SHRNv", "SQSHRNv", "SQSHRUNv",
                   "UQXTNv")>;
// -- matrix multiply
def : InstRW<[Ampere1Write_6cyc_2XY],
        (instrs SMMLA, UMMLA, USMMLA)>;
// -- max/min
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^SMAXv", "^SMINv", "^UMAXv", "^UMINv")>;
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^SMAXPv", "^SMINPv", "^UMAXPv", "^UMINPv")>;
// -- move immediate
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^MOVIv", "^MVNIv")>;
// -- multiply
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "MULv", "SMULLv", "UMULLv", "SQDMUL(H|L)v", "SQRDMULHv")>;
// -- multiply accumulate
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "MLAv", "MLSv", "(S|U|SQD)(MLAL|MLSL)v", "SQRDML(A|S)Hv")>;
// -- negation, saturating
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^SQABS", "^SQNEG")>;
// -- reverse bits/bytes
def : InstRW<[Ampere1Write_2cyc_1XY],
        (instregex "^RBITv", "^REV16v", "^REV32v", "^REV64v")>;
// -- shift
def : InstRW<[Ampere1Write_3cyc_1XY], (instregex "^[SU]SHL(v16i8|v8i16|v4i32|v2i64)")>;
// -- shift and accumulate
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "SRSRAv", "SSRAv", "URSRAv", "USRAv")>;
// -- shift, saturating
def : InstRW<[Ampere1Write_3cyc_1XY],
        (instregex "^SQRSHLv", "^SQRSHRNv", "^SQRSHRUNv", "^SQSHL", "^SQSHLU",
                   "^SQXTNv", "^SQXTUNv", "^UQSHRNv", "UQRSHRNv", "^UQRSHL",
                   "^UQSHL")>;

// Vector miscellaneous instructions
// -- duplicate element
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^DUPv.+lane")>;
// -- duplicate from GPR
def : InstRW<[Ampere1Write_5cyc_1BS], (instregex "^DUPv.+gpr")>;
// -- extract narrow
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^XTNv")>;
// -- insert/extract element
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^EXTv", "^INSv.+lane")>;
// -- move FP immediate
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^FMOVv")>;
// -- move element to GPR
def : InstRW<[Ampere1Write_6cyc_1XY_1Z], (instregex "(S|U)MOVv")>;
// -- move from GPR to any element
def : InstRW<[Ampere1Write_7cyc_1BS_1XY], (instregex "^INSv.+gpr")>;
// -- table lookup
def : InstRW<[Ampere1Write_2cyc_1XY],
            (instrs TBLv8i8One, TBLv16i8One, TBXv8i8One, TBXv16i8One)>;
def : InstRW<[Ampere1Write_4cyc_2XY],
            (instrs TBLv8i8Two, TBLv16i8Two, TBXv8i8Two, TBXv16i8Two)>;
def : InstRW<[Ampere1Write_6cyc_3XY],
            (instrs TBLv8i8Three, TBLv16i8Three, TBXv8i8Three, TBXv16i8Three)>;
def : InstRW<[Ampere1Write_8cyc_4XY],
            (instrs TBLv8i8Four, TBLv16i8Four, TBXv8i8Four, TBXv16i8Four)>;
// -- transpose
def : InstRW<[Ampere1Write_2cyc_1XY],
              (instregex "^TRN1v", "^TRN2v", "^UZP1v", "^UZP2v")>;
// -- zip/unzip
def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^ZIP1v", "^ZIP2v")>;

} // SchedModel = Ampere1Model