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
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
|
//===-- VOP3PInstructions.td - Vector Instruction Definitions -------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// VOP3P Classes
//===----------------------------------------------------------------------===//
class VOP3P_Profile<VOPProfile P, VOP3Features Features = VOP3_REGULAR,
bit HasDPP = 0> : VOP3_Profile<P, Features> {
let IsVOP3P = 1;
let HasExtVOP3DPP = HasDPP;
// We do not want to print src modifiers for vop3p because the bits are
// overloaded in meaning and the logic in printOperandAndFPInputMods is
// wrong for vop3p
let AsmVOP3Base = AsmVOP3P;
}
// Used for FMA_MIX* and MAD_MIX* insts
// Their operands are only sort of f16 operands. Depending on
// op_sel_hi, these may be interpreted as f32. The inline immediate
// values are really f16 converted to f32, so we treat these as f16
// operands.
class VOP3P_Mix_Profile<VOPProfile P, VOP3Features Features = VOP3_REGULAR,
bit useTiedOutput = 0> : VOP3P_Profile<P, Features, 1> {
bit UseTiedOutput = useTiedOutput;
dag srcs =
(ins FP16InputMods:$src0_modifiers, VCSrc_f16:$src0,
FP16InputMods:$src1_modifiers, VCSrc_f16:$src1,
FP16InputMods:$src2_modifiers, VCSrc_f16:$src2);
dag dpp_srcs =
(ins FPVRegInputMods:$src0_modifiers, VGPRSrc_32:$src0,
FPVRegInputMods:$src1_modifiers, VRegSrc_32:$src1,
FP16InputMods:$src2_modifiers, VCSrc_f16:$src2);
// FIXME: clampmod0 misbehaves with the non-default vdst_in
// following it. For now workaround this by requiring clamp
// in tied patterns. This should use undef_tied_input, but it
// seems underdeveloped and doesn't apply the right register
// class constraints.
dag mods = !con(!if(UseTiedOutput, (ins clampmod:$clamp, VGPR_32:$vdst_in),
(ins clampmod0:$clamp)),
(ins op_sel0:$op_sel, op_sel_hi0:$op_sel_hi));
// We use Ins64 because that is the one which populates InOperandList
// due to the logic in class VOP3_Pseudo
let Ins64 = !con(srcs, mods);
let InsVOP3Base = !con(dpp_srcs, mods);
let AsmVOP3Base =
"$vdst, $src0_modifiers, $src1_modifiers, $src2_modifiers$op_sel$op_sel_hi$clamp";
}
multiclass VOP3PInst<string OpName, VOPProfile P,
SDPatternOperator node = null_frag, bit IsDOT = 0> {
def NAME : VOP3P_Pseudo<OpName, P,
!if (P.HasModifiers,
getVOP3PModPat<P, node, IsDOT, IsDOT>.ret,
getVOP3Pat<P, node>.ret)>;
let SubtargetPredicate = isGFX11Plus in {
if P.HasExtVOP3DPP then
def _dpp : VOP3_DPP_Pseudo<OpName, P> {
let VOP3P = 1;
let PseudoInstr = OpName #"_dpp";
}
} // end SubtargetPredicate = isGFX11Plus
}
// Non-packed instructions that use the VOP3P encoding.
// VOP3 neg/abs and VOP3P opsel/opsel_hi modifiers are allowed.
multiclass VOP3_VOP3PInst<string OpName, VOP3P_Mix_Profile P> {
def NAME : VOP3P_Pseudo<OpName, P> {
let Constraints = !if(P.UseTiedOutput, "$vdst = $vdst_in", "");
let DisableEncoding = !if(P.UseTiedOutput, "$vdst_in", "");
}
let SubtargetPredicate = isGFX11Plus in {
if P.HasExtVOP3DPP then
def _dpp : VOP3_DPP_Pseudo<OpName, P> {
let VOP3P = 1;
let PseudoInstr = OpName#"_dpp";
let Constraints = !if(P.UseTiedOutput, "$vdst = $vdst_in", "");
let DisableEncoding = !if(P.UseTiedOutput, "$vdst_in", "");
}
} // end SubtargetPredicate = isGFX11Plus
}
let isReMaterializable = 1 in {
let isCommutable = 1 in {
defm V_PK_MAD_I16 : VOP3PInst<"v_pk_mad_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16_V2I16>>;
defm V_PK_MAD_U16 : VOP3PInst<"v_pk_mad_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16_V2I16>>;
let FPDPRounding = 1 in {
defm V_PK_FMA_F16 : VOP3PInst<"v_pk_fma_f16", VOP3P_Profile<VOP_V2F16_V2F16_V2F16_V2F16>, any_fma>;
defm V_PK_ADD_F16 : VOP3PInst<"v_pk_add_f16", VOP3P_Profile<VOP_V2F16_V2F16_V2F16>, any_fadd>;
defm V_PK_MUL_F16 : VOP3PInst<"v_pk_mul_f16", VOP3P_Profile<VOP_V2F16_V2F16_V2F16>, any_fmul>;
} // End FPDPRounding = 1
defm V_PK_MAX_F16 : VOP3PInst<"v_pk_max_f16", VOP3P_Profile<VOP_V2F16_V2F16_V2F16>, fmaxnum_like>;
defm V_PK_MIN_F16 : VOP3PInst<"v_pk_min_f16", VOP3P_Profile<VOP_V2F16_V2F16_V2F16>, fminnum_like>;
defm V_PK_ADD_U16 : VOP3PInst<"v_pk_add_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, add>;
defm V_PK_ADD_I16 : VOP3PInst<"v_pk_add_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>>;
defm V_PK_MUL_LO_U16 : VOP3PInst<"v_pk_mul_lo_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, mul>;
defm V_PK_MIN_I16 : VOP3PInst<"v_pk_min_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, smin>;
defm V_PK_MIN_U16 : VOP3PInst<"v_pk_min_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, umin>;
defm V_PK_MAX_I16 : VOP3PInst<"v_pk_max_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, smax>;
defm V_PK_MAX_U16 : VOP3PInst<"v_pk_max_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, umax>;
}
defm V_PK_SUB_U16 : VOP3PInst<"v_pk_sub_u16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>>;
defm V_PK_SUB_I16 : VOP3PInst<"v_pk_sub_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, sub>;
defm V_PK_LSHLREV_B16 : VOP3PInst<"v_pk_lshlrev_b16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, clshl_rev_16>;
defm V_PK_ASHRREV_I16 : VOP3PInst<"v_pk_ashrrev_i16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, cashr_rev_16>;
defm V_PK_LSHRREV_B16 : VOP3PInst<"v_pk_lshrrev_b16", VOP3P_Profile<VOP_V2I16_V2I16_V2I16>, clshr_rev_16>;
} // End isReMaterializable = 1
let SubtargetPredicate = HasVOP3PInsts in {
// Undo sub x, c -> add x, -c canonicalization since c is more likely
// an inline immediate than -c.
// The constant will be emitted as a mov, and folded later.
// TODO: We could directly encode the immediate now
def : GCNPat<
(add (v2i16 (VOP3PMods v2i16:$src0, i32:$src0_modifiers)), NegSubInlineConstV216:$src1),
(V_PK_SUB_U16 $src0_modifiers, $src0, SRCMODS.OP_SEL_1, NegSubInlineConstV216:$src1)
>;
// Integer operations with clamp bit set.
class VOP3PSatPat<SDPatternOperator pat, Instruction inst> : GCNPat<
(pat (v2i16 (VOP3PMods v2i16:$src0, i32:$src0_modifiers)),
(v2i16 (VOP3PMods v2i16:$src1, i32:$src1_modifiers))),
(inst $src0_modifiers, $src0, $src1_modifiers, $src1, DSTCLAMP.ENABLE)
>;
def : VOP3PSatPat<uaddsat, V_PK_ADD_U16>;
def : VOP3PSatPat<saddsat, V_PK_ADD_I16>;
def : VOP3PSatPat<usubsat, V_PK_SUB_U16>;
def : VOP3PSatPat<ssubsat, V_PK_SUB_I16>;
} // End SubtargetPredicate = HasVOP3PInsts
// TODO: Make sure we're doing the right thing with denormals. Note
// that FMA and MAD will differ.
multiclass MadFmaMixPats<SDPatternOperator fma_like,
Instruction mix_inst,
Instruction mixlo_inst,
Instruction mixhi_inst> {
// At least one of the operands needs to be an fpextend of an f16
// for this to be worthwhile, so we need three patterns here.
// TODO: Could we use a predicate to inspect src1/2/3 instead?
def : GCNPat <
(f32 (fma_like (f32 (VOP3PMadMixModsExt f16:$src0, i32:$src0_mods)),
(f32 (VOP3PMadMixMods f16:$src1, i32:$src1_mods)),
(f32 (VOP3PMadMixMods f16:$src2, i32:$src2_mods)))),
(mix_inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE)>;
def : GCNPat <
(f32 (fma_like (f32 (VOP3PMadMixMods f16:$src0, i32:$src0_mods)),
(f32 (VOP3PMadMixModsExt f16:$src1, i32:$src1_mods)),
(f32 (VOP3PMadMixMods f32:$src2, i32:$src2_mods)))),
(mix_inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE)>;
def : GCNPat <
(f32 (fma_like (f32 (VOP3PMadMixMods f16:$src0, i32:$src0_mods)),
(f32 (VOP3PMadMixMods f32:$src1, i32:$src1_mods)),
(f32 (VOP3PMadMixModsExt f16:$src2, i32:$src2_mods)))),
(mix_inst $src0_mods, $src0, $src1_mods, $src1, $src2_mods, $src2,
DSTCLAMP.NONE)>;
def : GCNPat <
(f16 (fpround (fma_like (f32 (VOP3PMadMixMods f16:$src0, i32:$src0_modifiers)),
(f32 (VOP3PMadMixMods f16:$src1, i32:$src1_modifiers)),
(f32 (VOP3PMadMixMods f16:$src2, i32:$src2_modifiers))))),
(mixlo_inst $src0_modifiers, $src0,
$src1_modifiers, $src1,
$src2_modifiers, $src2,
DSTCLAMP.NONE,
(i32 (IMPLICIT_DEF)))
>;
// FIXME: Special case handling for maxhi (especially for clamp)
// because dealing with the write to high half of the register is
// difficult.
def : GCNPat <
(build_vector f16:$elt0, (fpround (fma_like (f32 (VOP3PMadMixMods f16:$src0, i32:$src0_modifiers)),
(f32 (VOP3PMadMixMods f16:$src1, i32:$src1_modifiers)),
(f32 (VOP3PMadMixMods f16:$src2, i32:$src2_modifiers))))),
(v2f16 (mixhi_inst $src0_modifiers, $src0,
$src1_modifiers, $src1,
$src2_modifiers, $src2,
DSTCLAMP.NONE,
VGPR_32:$elt0))
>;
def : GCNPat <
(build_vector
f16:$elt0,
(AMDGPUclamp (fpround (fma_like (f32 (VOP3PMadMixMods f16:$src0, i32:$src0_modifiers)),
(f32 (VOP3PMadMixMods f16:$src1, i32:$src1_modifiers)),
(f32 (VOP3PMadMixMods f16:$src2, i32:$src2_modifiers)))))),
(v2f16 (mixhi_inst $src0_modifiers, $src0,
$src1_modifiers, $src1,
$src2_modifiers, $src2,
DSTCLAMP.ENABLE,
VGPR_32:$elt0))
>;
def : GCNPat <
(AMDGPUclamp (build_vector
(fpround (fma_like (f32 (VOP3PMadMixMods f16:$lo_src0, i32:$lo_src0_modifiers)),
(f32 (VOP3PMadMixMods f16:$lo_src1, i32:$lo_src1_modifiers)),
(f32 (VOP3PMadMixMods f16:$lo_src2, i32:$lo_src2_modifiers)))),
(fpround (fma_like (f32 (VOP3PMadMixMods f16:$hi_src0, i32:$hi_src0_modifiers)),
(f32 (VOP3PMadMixMods f16:$hi_src1, i32:$hi_src1_modifiers)),
(f32 (VOP3PMadMixMods f16:$hi_src2, i32:$hi_src2_modifiers)))))),
(v2f16 (mixhi_inst $hi_src0_modifiers, $hi_src0,
$hi_src1_modifiers, $hi_src1,
$hi_src2_modifiers, $hi_src2,
DSTCLAMP.ENABLE,
(mixlo_inst $lo_src0_modifiers, $lo_src0,
$lo_src1_modifiers, $lo_src1,
$lo_src2_modifiers, $lo_src2,
DSTCLAMP.ENABLE,
(i32 (IMPLICIT_DEF)))))
>;
def : GCNPat <
(f16 (fpround (fmul (f32 (VOP3PMadMixMods f32:$src0, i32:$src0_modifiers)),
(f32 (VOP3PMadMixMods f32:$src1, i32:$src1_modifiers))))),
(mixlo_inst $src0_modifiers, $src0,
$src1_modifiers, $src1,
(i32 0), (i32 0),
DSTCLAMP.NONE,
(i32 (IMPLICIT_DEF)))
>;
def : GCNPat <
(build_vector f16:$elt0, (fpround (fmul (f32 (VOP3PMadMixMods f32:$src0, i32:$src0_modifiers)),
(f32 (VOP3PMadMixMods f32:$src1, i32:$src1_modifiers))))),
(v2f16 (mixhi_inst $src0_modifiers, $src0,
$src1_modifiers, $src1,
(i32 0), (i32 0),
DSTCLAMP.NONE,
VGPR_32:$elt0))
>;
}
let SubtargetPredicate = HasMadMixInsts, OtherPredicates = [NoFP32Denormals] in {
// These are VOP3a-like opcodes which accept no omod.
// Size of src arguments (16/32) is controlled by op_sel.
// For 16-bit src arguments their location (hi/lo) are controlled by op_sel_hi.
let isCommutable = 1, mayRaiseFPException = 0 in {
let isReMaterializable = 1 in
defm V_MAD_MIX_F32 : VOP3_VOP3PInst<"v_mad_mix_f32", VOP3P_Mix_Profile<VOP_F32_F16_F16_F16, VOP3_OPSEL>>;
let FPDPRounding = 1 in {
// Clamp modifier is applied after conversion to f16.
defm V_MAD_MIXLO_F16 : VOP3_VOP3PInst<"v_mad_mixlo_f16", VOP3P_Mix_Profile<VOP_F16_F16_F16_F16, VOP3_OPSEL, 1>>;
let ClampLo = 0, ClampHi = 1 in {
defm V_MAD_MIXHI_F16 : VOP3_VOP3PInst<"v_mad_mixhi_f16", VOP3P_Mix_Profile<VOP_F16_F16_F16_F16, VOP3_OPSEL, 1>>;
}
} // End FPDPRounding = 1
}
defm : MadFmaMixPats<fmad, V_MAD_MIX_F32, V_MAD_MIXLO_F16, V_MAD_MIXHI_F16>;
} // End SubtargetPredicate = HasMadMixInsts, OtherPredicates = [NoFP32Denormals]
// Essentially the same as the mad_mix versions
let SubtargetPredicate = HasFmaMixInsts in {
let isCommutable = 1 in {
let isReMaterializable = 1 in
defm V_FMA_MIX_F32 : VOP3_VOP3PInst<"v_fma_mix_f32", VOP3P_Mix_Profile<VOP_F32_F16_F16_F16, VOP3_OPSEL>>;
let FPDPRounding = 1 in {
// Clamp modifier is applied after conversion to f16.
defm V_FMA_MIXLO_F16 : VOP3_VOP3PInst<"v_fma_mixlo_f16", VOP3P_Mix_Profile<VOP_F16_F16_F16_F16, VOP3_OPSEL, 1>>;
let ClampLo = 0, ClampHi = 1 in {
defm V_FMA_MIXHI_F16 : VOP3_VOP3PInst<"v_fma_mixhi_f16", VOP3P_Mix_Profile<VOP_F16_F16_F16_F16, VOP3_OPSEL, 1>>;
}
} // End FPDPRounding = 1
}
defm : MadFmaMixPats<fma, V_FMA_MIX_F32, V_FMA_MIXLO_F16, V_FMA_MIXHI_F16>;
}
// Defines patterns that extract signed 4bit from each Idx[0].
foreach Idx = [[0,28],[4,24],[8,20],[12,16],[16,12],[20,8],[24,4]] in
def ExtractSigned4bit_#Idx[0] : PatFrag<(ops node:$src),
(sra (shl node:$src, (i32 Idx[1])), (i32 28))>;
// Defines code pattern that extracts U(unsigned/signed) 4/8bit from FromBitIndex.
class Extract<int FromBitIndex, int BitMask, bit U>: PatFrag<
(ops node:$src),
!if (!or (!and (!eq (BitMask, 255), !eq (FromBitIndex, 24)), !eq (FromBitIndex, 28)), // last element
!if (U, (srl node:$src, (i32 FromBitIndex)), (sra node:$src, (i32 FromBitIndex))),
!if (!eq (FromBitIndex, 0), // first element
!if (U, (and node:$src, (i32 BitMask)),
!if (!eq (BitMask, 15), (!cast<PatFrag>("ExtractSigned4bit_"#FromBitIndex) node:$src),
(sext_inreg node:$src, i8))),
!if (U, (and (srl node:$src, (i32 FromBitIndex)), (i32 BitMask)),
!if (!eq (BitMask, 15), (!cast<PatFrag>("ExtractSigned4bit_"#FromBitIndex) node:$src),
(sext_inreg (srl node:$src, (i32 FromBitIndex)), i8)))))>;
foreach Type = ["I", "U"] in
foreach Index = 0-3 in {
// Defines patterns that extract each Index'ed 8bit from an unsigned
// 32bit scalar value;
def Type#Index#"_8bit" : Extract<!shl(Index, 3), 255, !eq (Type, "U")>;
// Defines multiplication patterns where the multiplication is happening on each
// Index'ed 8bit of a 32bit scalar value.
def Mul#Type#_Elt#Index : PatFrag<
(ops node:$src0, node:$src1),
(!cast<HasOneUseBinOp>(!if (!eq (Type, "I"), AMDGPUmul_i24_oneuse, AMDGPUmul_u24_oneuse))
(!cast<Extract>(Type#Index#"_8bit") node:$src0),
(!cast<Extract>(Type#Index#"_8bit") node:$src1))>;
}
// Different variants of dot8 patterns cause a huge increase in the compile time.
// Define non-associative/commutative add/mul to prevent permutation in the dot8
// pattern.
def NonACAdd : SDNode<"ISD::ADD" , SDTIntBinOp>;
def NonACAdd_oneuse : HasOneUseBinOp<NonACAdd>;
def NonACAMDGPUmul_u24 : SDNode<"AMDGPUISD::MUL_U24" , SDTIntBinOp>;
def NonACAMDGPUmul_u24_oneuse : HasOneUseBinOp<NonACAMDGPUmul_u24>;
def NonACAMDGPUmul_i24 : SDNode<"AMDGPUISD::MUL_I24" , SDTIntBinOp>;
def NonACAMDGPUmul_i24_oneuse : HasOneUseBinOp<NonACAMDGPUmul_i24>;
foreach Type = ["I", "U"] in
foreach Index = 0-7 in {
// Defines patterns that extract each Index'ed 4bit from an unsigned
// 32bit scalar value;
def Type#Index#"_4bit" : Extract<!shl(Index, 2), 15, !eq (Type, "U")>;
// Defines multiplication patterns where the multiplication is happening on each
// Index'ed 8bit of a 32bit scalar value.
def Mul#Type#Index#"_4bit" : PatFrag<
(ops node:$src0, node:$src1),
(!cast<HasOneUseBinOp>(!if (!eq (Type, "I"), NonACAMDGPUmul_i24_oneuse, NonACAMDGPUmul_u24_oneuse))
(!cast<Extract>(Type#Index#"_4bit") node:$src0),
(!cast<Extract>(Type#Index#"_4bit") node:$src1))>;
}
class UDot2Pat<Instruction Inst> : GCNPat <
(add (add_oneuse (AMDGPUmul_u24_oneuse (srl i32:$src0, (i32 16)),
(srl i32:$src1, (i32 16))), i32:$src2),
(AMDGPUmul_u24_oneuse (and i32:$src0, (i32 65535)),
(and i32:$src1, (i32 65535)))
),
(Inst (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))> {
let SubtargetPredicate = !cast<VOP_Pseudo>(Inst).SubtargetPredicate;
}
class SDot2Pat<Instruction Inst> : GCNPat <
(add (add_oneuse (AMDGPUmul_i24_oneuse (sra i32:$src0, (i32 16)),
(sra i32:$src1, (i32 16))), i32:$src2),
(AMDGPUmul_i24_oneuse (sext_inreg i32:$src0, i16),
(sext_inreg i32:$src1, i16))),
(Inst (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))> {
let SubtargetPredicate = !cast<VOP_Pseudo>(Inst).SubtargetPredicate;
}
let IsDOT = 1 in {
let SubtargetPredicate = HasDot2Insts in {
defm V_DOT2_I32_I16 : VOP3PInst<"v_dot2_i32_i16",
VOP3P_Profile<VOP_I32_V2I16_V2I16_I32>, int_amdgcn_sdot2, 1>;
defm V_DOT2_U32_U16 : VOP3PInst<"v_dot2_u32_u16",
VOP3P_Profile<VOP_I32_V2I16_V2I16_I32>, int_amdgcn_udot2, 1>;
} // End SubtargetPredicate = HasDot2Insts
let SubtargetPredicate = HasDot10Insts in
defm V_DOT2_F32_F16 : VOP3PInst<"v_dot2_f32_f16",
VOP3P_Profile<VOP_F32_V2F16_V2F16_F32, VOP3_REGULAR, /*HasDPP*/ 1>,
AMDGPUfdot2, 1/*ExplicitClamp*/>;
let SubtargetPredicate = HasDot7Insts in {
defm V_DOT4_U32_U8 : VOP3PInst<"v_dot4_u32_u8",
VOP3P_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>, int_amdgcn_udot4, 1>;
defm V_DOT8_U32_U4 : VOP3PInst<"v_dot8_u32_u4",
VOP3P_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>, int_amdgcn_udot8, 1>;
} // End SubtargetPredicate = HasDot7Insts
let SubtargetPredicate = HasDot1Insts in {
defm V_DOT4_I32_I8 : VOP3PInst<"v_dot4_i32_i8",
VOP3P_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>, int_amdgcn_sdot4, 1>;
defm V_DOT8_I32_I4 : VOP3PInst<"v_dot8_i32_i4",
VOP3P_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>, int_amdgcn_sdot8, 1>;
} // End SubtargetPredicate = HasDot1Insts
def DOT2_BF16_Profile
: VOP3P_Profile<VOP_F32_V2I16_V2I16_F32, VOP3_REGULAR, /*HasDPP*/ 1> {
let HasSrc1Mods = 1;
}
let SubtargetPredicate = HasDot9Insts in {
defm V_DOT2_F32_BF16 : VOP3PInst<"v_dot2_f32_bf16", DOT2_BF16_Profile,
int_amdgcn_fdot2_f32_bf16, 1>;
} // End SubtargetPredicate = HasDot9Insts
} // End let IsDOT = 1
multiclass VOP3PDOTIUInst <string OpName, SDPatternOperator intrinsic_node> {
let IsDOT = 1 in
defm NAME : VOP3PInst<OpName, VOP3P_Profile<VOP_I32_I32_I32_I32, VOP3_PACKED>,
null_frag, 1>;
// Dot-iu instructions consider input as signed if imod neg bits are set. Thus
// Dot-iu Intrinsics have extra operands and require separate codegen pattern.
def : GCNPat < (intrinsic_node (DotIUVOP3PMods i32:$src0_mods), i32:$src0,
(DotIUVOP3PMods i32:$src1_mods), i32:$src1,
i32:$src2, (i1 timm:$clamp)),
(!cast<Instruction>(NAME) $src0_mods, i32:$src0,
$src1_mods, i32:$src1,
(i32 8), i32:$src2, i1:$clamp)
>;
}
let SubtargetPredicate = HasDot8Insts in {
defm V_DOT4_I32_IU8 : VOP3PDOTIUInst<"v_dot4_i32_iu8", int_amdgcn_sudot4>;
defm V_DOT8_I32_IU4 : VOP3PDOTIUInst<"v_dot8_i32_iu4", int_amdgcn_sudot8>;
} // End SubtargetPredicate = HasDot8Insts
def : UDot2Pat<V_DOT2_U32_U16>;
def : SDot2Pat<V_DOT2_I32_I16>;
foreach Type = ["U", "I"] in
let SubtargetPredicate = !cast<VOP_Pseudo>("V_DOT4_"#Type#"32_"#Type#8).SubtargetPredicate in
def : GCNPat <
!cast<dag>(!foldl((i32 i32:$src2), [0, 1, 2, 3], lhs, y,
(add_oneuse lhs, (!cast<PatFrag>("Mul"#Type#"_Elt"#y) i32:$src0, i32:$src1)))),
(!cast<VOP3P_Pseudo>("V_DOT4_"#Type#"32_"#Type#8) (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))>;
foreach Type = ["U", "I"] in
let SubtargetPredicate = !cast<VOP_Pseudo>("V_DOT8_"#Type#"32_"#Type#4).SubtargetPredicate in
def : GCNPat <
!cast<dag>(!foldl((add_oneuse i32:$src2, (!cast<PatFrag>("Mul"#Type#"0_4bit") i32:$src0, i32:$src1)),
[1, 2, 3, 4, 5, 6, 7], lhs, y,
(NonACAdd_oneuse lhs, (!cast<PatFrag>("Mul"#Type#y#"_4bit") i32:$src0, i32:$src1)))),
(!cast<VOP3P_Pseudo>("V_DOT8_"#Type#"32_"#Type#4) (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))>;
// Different variants of dot8 code-gen dag patterns are not generated through table-gen due to a huge increase
// in the compile time. Directly handle the pattern generated by the FE here.
foreach Type = ["U", "I"] in
let SubtargetPredicate = !cast<VOP_Pseudo>("V_DOT8_"#Type#"32_"#Type#4).SubtargetPredicate in
def : GCNPat <
!cast<dag>(!foldl((add_oneuse i32:$src2, (!cast<PatFrag>("Mul"#Type#"0_4bit") i32:$src0, i32:$src1)),
[7, 1, 2, 3, 4, 5, 6], lhs, y,
(NonACAdd_oneuse lhs, (!cast<PatFrag>("Mul"#Type#y#"_4bit") i32:$src0, i32:$src1)))),
(!cast<VOP3P_Pseudo>("V_DOT8_"#Type#"32_"#Type#4) (i32 8), $src0, (i32 8), $src1, (i32 8), $src2, (i1 0))>;
def ADst_32 : VOPDstOperand<AGPR_32>;
def ADst_64 : VOPDstOperand<AReg_64>;
def ADst_128 : VOPDstOperand<AReg_128>;
def ADst_256 : VOPDstOperand<AReg_256>;
def ADst_512 : VOPDstOperand<AReg_512>;
def ADst_1024 : VOPDstOperand<AReg_1024>;
def VDst_64 : VOPDstOperand<VReg_64>;
def VDst_128 : VOPDstOperand<VReg_128>;
def VDst_256 : VOPDstOperand<VReg_256>;
def VDst_512 : VOPDstOperand<VReg_512>;
def VDst_1024 : VOPDstOperand<VReg_1024>;
def VOPProfileAccRead : VOP3P_Profile<VOP_I32_I32, VOP3_MAI> {
let Src0RC64 = ARegSrc_32;
}
def VOPProfileAccWrite : VOP3P_Profile<VOP_I32_I32, VOP3_MAI> {
let DstRC = ADst_32;
let Src0RC64 = VCSrc_b32;
}
class VOPProfileMAI<VOPProfile P, RegisterOperand _SrcRC, RegisterOperand _DstRC,
RegisterOperand SrcABRC = AVSrc_32>
: VOP3P_Profile<P, VOP3_MAI> {
let DstRC = _DstRC;
let Src0RC64 = SrcABRC;
let Src1RC64 = SrcABRC;
let Src2RC64 = _SrcRC;
let HasOpSel = 0;
let HasClamp = 0;
let HasIntClamp = 0;
let HasOMod = 0;
let HasModifiers = 0;
let AsmVOP3Base = "$vdst, $src0, $src1, $src2$cbsz$abid$blgp";
let Ins64 = (ins Src0RC64:$src0, Src1RC64:$src1, Src2RC64:$src2, cbsz:$cbsz, abid:$abid, blgp:$blgp);
let InsVOP3Base = Ins64;
// Dst and SrcC cannot partially overlap if SrcC/Dst is bigger than 4 VGPRs.
// We then create two versions of the instruction: with tied dst and src2
// and with the earlyclobber flag on the dst. This is stricter than the
// actual HW restriction. In particular earlyclobber also affects src0 and
// src1 allocation which is not required.
bit NoDstOverlap = !gt(DstVT.Size, 128);
}
class VOPProfileSMFMAC<VOPProfile P, RegisterOperand _DstRC,
RegisterOperand _SrcARC, RegisterOperand _SrcBRC>
: VOPProfileMAI<P, _DstRC, _DstRC, _SrcARC> {
let Src1RC64 = _SrcBRC;
let Src2VT = DstVT;
let Asm64 = " $vdst, $src0, $src1, $idx$cbsz$abid";
let Outs64 = (outs DstRC:$vdst);
let Ins64 = (ins Src0RC64:$src0, Src1RC64:$src1, VRegSrc_32:$idx, cbsz:$cbsz, abid:$abid, Src2RC64:$src2);
}
def VOPProfileMAI_F32_F32_X4 : VOPProfileMAI<VOP_V4F32_F32_F32_V4F32, AISrc_128_f32, ADst_128>;
def VOPProfileMAI_F32_F32_X16 : VOPProfileMAI<VOP_V16F32_F32_F32_V16F32, AISrc_512_f32, ADst_512>;
def VOPProfileMAI_F32_F32_X32 : VOPProfileMAI<VOP_V32F32_F32_F32_V32F32, AISrc_1024_f32, ADst_1024>;
def VOPProfileMAI_I32_I32_X4 : VOPProfileMAI<VOP_V4I32_I32_I32_V4I32, AISrc_128_b32, ADst_128>;
def VOPProfileMAI_I32_I32_X16 : VOPProfileMAI<VOP_V16I32_I32_I32_V16I32, AISrc_512_b32, ADst_512>;
def VOPProfileMAI_I32_I32_X32 : VOPProfileMAI<VOP_V32I32_I32_I32_V32I32, AISrc_1024_b32, ADst_1024>;
def VOPProfileMAI_F32_V2I16_X4 : VOPProfileMAI<VOP_V4F32_V2I16_V2I16_V4F32, AISrc_128_b32, ADst_128>;
def VOPProfileMAI_F32_V2I16_X16 : VOPProfileMAI<VOP_V16F32_V2I16_V2I16_V16F32, AISrc_512_b32, ADst_512>;
def VOPProfileMAI_F32_V2I16_X32 : VOPProfileMAI<VOP_V32F32_V2I16_V2I16_V32F32, AISrc_1024_b32, ADst_1024>;
def VOPProfileMAI_F32_V4F16_X4 : VOPProfileMAI<VOP_V4F32_V4F16_V4F16_V4F32, AISrc_128_b32, ADst_128, AVSrc_64>;
def VOPProfileMAI_F32_V4F16_X16 : VOPProfileMAI<VOP_V16F32_V4F16_V4F16_V16F32, AISrc_512_b32, ADst_512, AVSrc_64>;
def VOPProfileMAI_F32_V4F16_X32 : VOPProfileMAI<VOP_V32F32_V4F16_V4F16_V32F32, AISrc_1024_b32, ADst_1024, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X4 : VOPProfileMAI<VOP_V4F32_V4I16_V4I16_V4F32, AISrc_128_b32, ADst_128, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X16 : VOPProfileMAI<VOP_V16F32_V4I16_V4I16_V16F32, AISrc_512_b32, ADst_512, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X32 : VOPProfileMAI<VOP_V32F32_V4I16_V4I16_V32F32, AISrc_1024_b32, ADst_1024, AVSrc_64>;
def VOPProfileMAI_F64_16X16X4F64 : VOPProfileMAI<VOP_V4F64_F64_F64_V4F64, AISrc_256_f64, ADst_256, AVSrc_64>;
def VOPProfileMAI_F64_4X4X4F64 : VOPProfileMAI<VOP_F64_F64_F64_F64, AISrc_64_f64, ADst_64, AVSrc_64>;
def VOPProfileMAI_I32_I64_X16 : VOPProfileMAI<VOP_V4I32_I64_I64_V4I32, AISrc_128_b32, ADst_128, AVSrc_64>;
def VOPProfileMAI_I32_I64_X32 : VOPProfileMAI<VOP_V16I32_I64_I64_V16I32, AISrc_512_b32, ADst_512, AVSrc_64>;
def VOPProfileMAI_F32_V2F32_X16 : VOPProfileMAI<VOP_V4F32_V2F32_V2F32_V4F32, AISrc_128_b32, ADst_128, AVSrc_64>;
def VOPProfileMAI_F32_V2F32_X32 : VOPProfileMAI<VOP_V16F32_V2F32_V2F32_V16F32, AISrc_512_b32, ADst_512, AVSrc_64>;
def VOPProfileMAI_F32_I64_X32 : VOPProfileMAI<VOP_V4F32_I64_I64_V4F32, AISrc_128_b32, ADst_128, AVSrc_64>;
def VOPProfileMAI_F32_I64_X16 : VOPProfileMAI<VOP_V16F32_I64_I64_V16F32, AISrc_512_b32, ADst_512, AVSrc_64>;
def VOPProfileMAI_F32_F32_X4_VCD : VOPProfileMAI<VOP_V4F32_F32_F32_V4F32, VISrc_128_f32, VDst_128>;
def VOPProfileMAI_F32_F32_X16_VCD : VOPProfileMAI<VOP_V16F32_F32_F32_V16F32, VISrc_512_f32, VDst_512>;
def VOPProfileMAI_F32_F32_X32_VCD : VOPProfileMAI<VOP_V32F32_F32_F32_V32F32, VISrc_1024_f32, VDst_1024>;
def VOPProfileMAI_I32_I32_X4_VCD : VOPProfileMAI<VOP_V4I32_I32_I32_V4I32, VISrc_128_b32, VDst_128>;
def VOPProfileMAI_I32_I32_X16_VCD : VOPProfileMAI<VOP_V16I32_I32_I32_V16I32, VISrc_512_b32, VDst_512>;
def VOPProfileMAI_I32_I32_X32_VCD : VOPProfileMAI<VOP_V32I32_I32_I32_V32I32, VISrc_1024_b32, VDst_1024>;
def VOPProfileMAI_F32_V2I16_X4_VCD : VOPProfileMAI<VOP_V4F32_V2I16_V2I16_V4F32, VISrc_128_b32, VDst_128>;
def VOPProfileMAI_F32_V2I16_X16_VCD : VOPProfileMAI<VOP_V16F32_V2I16_V2I16_V16F32, VISrc_512_b32, VDst_512>;
def VOPProfileMAI_F32_V2I16_X32_VCD : VOPProfileMAI<VOP_V32F32_V2I16_V2I16_V32F32, VISrc_1024_b32, VDst_1024>;
def VOPProfileMAI_F32_V4F16_X4_VCD : VOPProfileMAI<VOP_V4F32_V4F16_V4F16_V4F32, VISrc_128_b32, VDst_128, AVSrc_64>;
def VOPProfileMAI_F32_V4F16_X16_VCD : VOPProfileMAI<VOP_V16F32_V4F16_V4F16_V16F32, VISrc_512_b32, VDst_512, AVSrc_64>;
def VOPProfileMAI_F32_V4F16_X32_VCD : VOPProfileMAI<VOP_V32F32_V4F16_V4F16_V32F32, VISrc_1024_b32, VDst_1024, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X4_VCD : VOPProfileMAI<VOP_V4F32_V4I16_V4I16_V4F32, VISrc_128_b32, VDst_128, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X16_VCD : VOPProfileMAI<VOP_V16F32_V4I16_V4I16_V16F32, VISrc_512_b32, VDst_512, AVSrc_64>;
def VOPProfileMAI_F32_V4I16_X32_VCD : VOPProfileMAI<VOP_V32F32_V4I16_V4I16_V32F32, VISrc_1024_b32, VDst_1024, AVSrc_64>;
def VOPProfileMAI_F64_16X16X4F64_VCD : VOPProfileMAI<VOP_V4F64_F64_F64_V4F64, VISrc_256_f64, VDst_256, AVSrc_64>;
def VOPProfileMAI_F64_4X4X4F64_VCD : VOPProfileMAI<VOP_F64_F64_F64_F64, VISrc_64_f64, VDst_64, AVSrc_64>;
def VOPProfileMAI_I32_I64_X16_VCD : VOPProfileMAI<VOP_V4I32_I64_I64_V4I32, VISrc_128_b32, VDst_128, AVSrc_64>;
def VOPProfileMAI_I32_I64_X32_VCD : VOPProfileMAI<VOP_V16I32_I64_I64_V16I32, VISrc_512_b32, VDst_512, AVSrc_64>;
def VOPProfileMAI_F32_V2F32_X16_VCD : VOPProfileMAI<VOP_V4F32_V2F32_V2F32_V4F32, VISrc_128_b32, VDst_128, AVSrc_64>;
def VOPProfileMAI_F32_V2F32_X32_VCD : VOPProfileMAI<VOP_V16F32_V2F32_V2F32_V16F32, VISrc_512_b32, VDst_512, AVSrc_64>;
def VOPProfileMAI_F32_I64_X32_VCD : VOPProfileMAI<VOP_V4F32_I64_I64_V4F32, VISrc_128_b32, VDst_128, AVSrc_64>;
def VOPProfileMAI_F32_I64_X16_VCD : VOPProfileMAI<VOP_V16F32_I64_I64_V16F32, VISrc_512_b32, VDst_512, AVSrc_64>;
def VOPProfileSMFMAC_F32_16X16X32_F16 : VOPProfileSMFMAC<VOP_V4F32_V4F16_V8F16_I32, AVDst_128, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_F32_32X32X16_F16 : VOPProfileSMFMAC<VOP_V16F32_V4F16_V8F16_I32, AVDst_512, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_F32_16X16X32_I16 : VOPProfileSMFMAC<VOP_V4F32_V4I16_V8I16_I32, AVDst_128, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_F32_32X32X16_I16 : VOPProfileSMFMAC<VOP_V16F32_V4I16_V8I16_I32, AVDst_512, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_I32_16X16X64_I8 : VOPProfileSMFMAC<VOP_V4I32_V2I32_V4I32_I32, AVDst_128, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_I32_32X32X32_I8 : VOPProfileSMFMAC<VOP_V16I32_V2I32_V4I32_I32, AVDst_512, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_F32_16X16X64_F8 : VOPProfileSMFMAC<VOP_V4F32_V2I32_V4I32_I32, AVDst_128, AVSrc_64, AVSrc_128>;
def VOPProfileSMFMAC_F32_32X32X32_F8 : VOPProfileSMFMAC<VOP_V16F32_V2I32_V4I32_I32, AVDst_512, AVSrc_64, AVSrc_128>;
class MFMATable <bit is_mac, string Name> {
bit IsMac = is_mac;
string FMAOp = Name;
}
class MAIFrag<SDPatternOperator Op, code pred> : PatFrag <
(ops node:$src0, node:$src1, node:$src2, node:$cbsz, node:$abid, node:$blgp),
(Op $src0, $src1, $src2, $cbsz, $abid, $blgp),
pred
>;
let GISelPredicateCode = [{ return MF.getInfo<SIMachineFunctionInfo>()->mayNeedAGPRs(); }] in
class AgprMAIFrag<SDPatternOperator Op> :
MAIFrag<Op, [{ return MF->getInfo<SIMachineFunctionInfo>()->mayNeedAGPRs(); }]>;
let GISelPredicateCode = [{ return !MF.getInfo<SIMachineFunctionInfo>()->mayNeedAGPRs(); }] in
class VgprMAIFrag<SDPatternOperator Op> :
MAIFrag<Op, [{ return !MF->getInfo<SIMachineFunctionInfo>()->mayNeedAGPRs(); }]>;
let Predicates = [HasMAIInsts] in {
let isAsCheapAsAMove = 1, isReMaterializable = 1 in {
defm V_ACCVGPR_READ_B32 : VOP3Inst<"v_accvgpr_read_b32", VOPProfileAccRead>;
let isMoveImm = 1 in {
defm V_ACCVGPR_WRITE_B32 : VOP3Inst<"v_accvgpr_write_b32", VOPProfileAccWrite>;
} // End isMoveImm = 1
} // End isAsCheapAsAMove = 1, isReMaterializable = 1
class MAIInst<string OpName, VOPProfile P, SDPatternOperator node>
: VOP3InstBase<OpName, P, node> {
Instruction Opcode = !cast<Instruction>(NAME);
bit is_dgemm = 0;
bit is_gfx940_xdl = 0;
}
multiclass MAIInst<string OpName, string P, SDPatternOperator node,
bit NoDstOverlap = !cast<VOPProfileMAI>("VOPProfileMAI_" # P).NoDstOverlap> {
let isConvergent = 1, mayRaiseFPException = 0, ReadsModeReg = 1 in {
// FP32 denorm mode is respected, rounding mode is not. Exceptions are not supported.
let Constraints = !if(NoDstOverlap, "@earlyclobber $vdst", "") in {
def _e64 : MAIInst<OpName, !cast<VOPProfileMAI>("VOPProfileMAI_" # P),
!if(NoDstOverlap, null_frag, AgprMAIFrag<node>)>,
MFMATable<0, NAME # "_e64">;
let SubtargetPredicate = isGFX90APlus, Mnemonic = OpName in
def _vgprcd_e64 : MAIInst<OpName # "_vgprcd", !cast<VOPProfileMAI>("VOPProfileMAI_" # P # "_VCD"),
!if(NoDstOverlap, null_frag, VgprMAIFrag<node>)>,
MFMATable<0, NAME # "_vgprcd_e64">;
}
if NoDstOverlap then {
let Constraints = !if(NoDstOverlap, "$vdst = $src2", ""),
isConvertibleToThreeAddress = NoDstOverlap,
Mnemonic = OpName in {
def "_mac_e64" : MAIInst<OpName # "_mac", !cast<VOPProfileMAI>("VOPProfileMAI_" # P), AgprMAIFrag<node>>,
MFMATable<1, NAME # "_e64">;
let SubtargetPredicate = isGFX90APlus in
def _mac_vgprcd_e64 : MAIInst<OpName # "_mac_vgprcd", !cast<VOPProfileMAI>("VOPProfileMAI_" # P # "_VCD"),
VgprMAIFrag<node>>,
MFMATable<1, NAME # "_vgprcd_e64">;
}
}
} // End isConvergent = 1, mayRaiseFPException = 0, ReadsModeReg = 1
}
defm V_MFMA_F32_4X4X1F32 : MAIInst<"v_mfma_f32_4x4x1f32", "F32_F32_X4", int_amdgcn_mfma_f32_4x4x1f32>;
defm V_MFMA_F32_16X16X1F32 : MAIInst<"v_mfma_f32_16x16x1f32", "F32_F32_X16", int_amdgcn_mfma_f32_16x16x1f32>;
defm V_MFMA_F32_16X16X4F32 : MAIInst<"v_mfma_f32_16x16x4f32", "F32_F32_X4", int_amdgcn_mfma_f32_16x16x4f32>;
defm V_MFMA_F32_32X32X1F32 : MAIInst<"v_mfma_f32_32x32x1f32", "F32_F32_X32", int_amdgcn_mfma_f32_32x32x1f32>;
defm V_MFMA_F32_32X32X2F32 : MAIInst<"v_mfma_f32_32x32x2f32", "F32_F32_X16", int_amdgcn_mfma_f32_32x32x2f32>;
let is_gfx940_xdl = 1 in {
defm V_MFMA_F32_4X4X4F16 : MAIInst<"v_mfma_f32_4x4x4f16", "F32_V4F16_X4", int_amdgcn_mfma_f32_4x4x4f16>;
defm V_MFMA_I32_4X4X4I8 : MAIInst<"v_mfma_i32_4x4x4i8", "I32_I32_X4", int_amdgcn_mfma_i32_4x4x4i8>;
defm V_MFMA_F32_16X16X4F16 : MAIInst<"v_mfma_f32_16x16x4f16", "F32_V4F16_X16", int_amdgcn_mfma_f32_16x16x4f16>;
defm V_MFMA_F32_16X16X16F16 : MAIInst<"v_mfma_f32_16x16x16f16", "F32_V4F16_X4", int_amdgcn_mfma_f32_16x16x16f16>;
defm V_MFMA_I32_16X16X4I8 : MAIInst<"v_mfma_i32_16x16x4i8", "I32_I32_X16", int_amdgcn_mfma_i32_16x16x4i8>;
defm V_MFMA_F32_32X32X4F16 : MAIInst<"v_mfma_f32_32x32x4f16", "F32_V4F16_X32", int_amdgcn_mfma_f32_32x32x4f16>;
defm V_MFMA_F32_32X32X8F16 : MAIInst<"v_mfma_f32_32x32x8f16", "F32_V4F16_X16", int_amdgcn_mfma_f32_32x32x8f16>;
defm V_MFMA_I32_32X32X4I8 : MAIInst<"v_mfma_i32_32x32x4i8", "I32_I32_X32", int_amdgcn_mfma_i32_32x32x4i8>;
}
let Predicates = [isGFX908orGFX90A] in {
defm V_MFMA_I32_16X16X16I8 : MAIInst<"v_mfma_i32_16x16x16i8", "I32_I32_X4", int_amdgcn_mfma_i32_16x16x16i8>;
defm V_MFMA_I32_32X32X8I8 : MAIInst<"v_mfma_i32_32x32x8i8", "I32_I32_X16", int_amdgcn_mfma_i32_32x32x8i8>;
defm V_MFMA_F32_4X4X2BF16 : MAIInst<"v_mfma_f32_4x4x2bf16", "F32_V2I16_X4", int_amdgcn_mfma_f32_4x4x2bf16>;
defm V_MFMA_F32_16X16X2BF16 : MAIInst<"v_mfma_f32_16x16x2bf16", "F32_V2I16_X16", int_amdgcn_mfma_f32_16x16x2bf16>;
defm V_MFMA_F32_16X16X8BF16 : MAIInst<"v_mfma_f32_16x16x8bf16", "F32_V2I16_X4", int_amdgcn_mfma_f32_16x16x8bf16>;
defm V_MFMA_F32_32X32X2BF16 : MAIInst<"v_mfma_f32_32x32x2bf16", "F32_V2I16_X32", int_amdgcn_mfma_f32_32x32x2bf16>;
defm V_MFMA_F32_32X32X4BF16 : MAIInst<"v_mfma_f32_32x32x4bf16", "F32_V2I16_X16", int_amdgcn_mfma_f32_32x32x4bf16>;
}
} // End SubtargetPredicate = HasMAIInsts
let Predicates = [isGFX90APlus] in {
let is_gfx940_xdl = 1 in {
defm V_MFMA_F32_32X32X4BF16_1K : MAIInst<"v_mfma_f32_32x32x4bf16_1k", "F32_V4I16_X32", int_amdgcn_mfma_f32_32x32x4bf16_1k>;
defm V_MFMA_F32_16X16X4BF16_1K : MAIInst<"v_mfma_f32_16x16x4bf16_1k", "F32_V4I16_X16", int_amdgcn_mfma_f32_16x16x4bf16_1k>;
defm V_MFMA_F32_4X4X4BF16_1K : MAIInst<"v_mfma_f32_4x4x4bf16_1k", "F32_V4I16_X4", int_amdgcn_mfma_f32_4x4x4bf16_1k>;
defm V_MFMA_F32_32X32X8BF16_1K : MAIInst<"v_mfma_f32_32x32x8bf16_1k", "F32_V4I16_X16", int_amdgcn_mfma_f32_32x32x8bf16_1k>;
defm V_MFMA_F32_16X16X16BF16_1K : MAIInst<"v_mfma_f32_16x16x16bf16_1k", "F32_V4I16_X4", int_amdgcn_mfma_f32_16x16x16bf16_1k>;
}
let is_dgemm = 1 in {
defm V_MFMA_F64_16X16X4F64 : MAIInst<"v_mfma_f64_16x16x4f64", "F64_16X16X4F64", int_amdgcn_mfma_f64_16x16x4f64>;
defm V_MFMA_F64_4X4X4F64 : MAIInst<"v_mfma_f64_4x4x4f64", "F64_4X4X4F64", int_amdgcn_mfma_f64_4x4x4f64>;
}
} // End Predicates = [isGFX90APlus]
let Predicates = [isGFX940Plus], is_gfx940_xdl = 1 in {
defm V_MFMA_I32_32X32X16I8 : MAIInst<"v_mfma_i32_32x32x16i8", "I32_I64_X32", int_amdgcn_mfma_i32_32x32x16_i8>;
defm V_MFMA_I32_16X16X32I8 : MAIInst<"v_mfma_i32_16x16x32i8", "I32_I64_X16", int_amdgcn_mfma_i32_16x16x32_i8>;
defm V_MFMA_F32_16X16X8XF32 : MAIInst<"v_mfma_f32_16x16x8xf32", "F32_V2F32_X16", int_amdgcn_mfma_f32_16x16x8_xf32>;
defm V_MFMA_F32_32X32X4XF32 : MAIInst<"v_mfma_f32_32x32x4xf32", "F32_V2F32_X32", int_amdgcn_mfma_f32_32x32x4_xf32>;
defm V_MFMA_F32_16X16X32_BF8_BF8 : MAIInst<"v_mfma_f32_16x16x32_bf8_bf8", "F32_I64_X32", int_amdgcn_mfma_f32_16x16x32_bf8_bf8>;
defm V_MFMA_F32_16X16X32_BF8_FP8 : MAIInst<"v_mfma_f32_16x16x32_bf8_fp8", "F32_I64_X32", int_amdgcn_mfma_f32_16x16x32_bf8_fp8>;
defm V_MFMA_F32_16X16X32_FP8_BF8 : MAIInst<"v_mfma_f32_16x16x32_fp8_bf8", "F32_I64_X32", int_amdgcn_mfma_f32_16x16x32_fp8_bf8>;
defm V_MFMA_F32_16X16X32_FP8_FP8 : MAIInst<"v_mfma_f32_16x16x32_fp8_fp8", "F32_I64_X32", int_amdgcn_mfma_f32_16x16x32_fp8_fp8>;
defm V_MFMA_F32_32X32X16_BF8_BF8 : MAIInst<"v_mfma_f32_32x32x16_bf8_bf8", "F32_I64_X16", int_amdgcn_mfma_f32_32x32x16_bf8_bf8>;
defm V_MFMA_F32_32X32X16_BF8_FP8 : MAIInst<"v_mfma_f32_32x32x16_bf8_fp8", "F32_I64_X16", int_amdgcn_mfma_f32_32x32x16_bf8_fp8>;
defm V_MFMA_F32_32X32X16_FP8_BF8 : MAIInst<"v_mfma_f32_32x32x16_fp8_bf8", "F32_I64_X16", int_amdgcn_mfma_f32_32x32x16_fp8_bf8>;
defm V_MFMA_F32_32X32X16_FP8_FP8 : MAIInst<"v_mfma_f32_32x32x16_fp8_fp8", "F32_I64_X16", int_amdgcn_mfma_f32_32x32x16_fp8_fp8>;
} // End Predicates = [isGFX940Plus], is_gfx940_xdl = 1
multiclass SMFMACInst<string OpName, string P, SDPatternOperator node> {
let Constraints = "$vdst = $src2", DisableEncoding = "$src2",
isConvergent = 1, mayRaiseFPException = 0, ReadsModeReg = 1, is_gfx940_xdl = 1 in {
def _e64 : MAIInst<OpName, !cast<VOPProfileSMFMAC>("VOPProfileSMFMAC_" # P), node>;
}
}
let SubtargetPredicate = isGFX940Plus in {
defm V_SMFMAC_F32_16X16X32_F16 : SMFMACInst<"v_smfmac_f32_16x16x32_f16", "F32_16X16X32_F16", int_amdgcn_smfmac_f32_16x16x32_f16>;
defm V_SMFMAC_F32_32X32X16_F16 : SMFMACInst<"v_smfmac_f32_32x32x16_f16", "F32_32X32X16_F16", int_amdgcn_smfmac_f32_32x32x16_f16>;
defm V_SMFMAC_F32_16X16X32_BF16 : SMFMACInst<"v_smfmac_f32_16x16x32_bf16", "F32_16X16X32_I16", int_amdgcn_smfmac_f32_16x16x32_bf16>;
defm V_SMFMAC_F32_32X32X16_BF16 : SMFMACInst<"v_smfmac_f32_32x32x16_bf16", "F32_32X32X16_I16", int_amdgcn_smfmac_f32_32x32x16_bf16>;
defm V_SMFMAC_I32_16X16X64_I8 : SMFMACInst<"v_smfmac_i32_16x16x64_i8", "I32_16X16X64_I8", int_amdgcn_smfmac_i32_16x16x64_i8>;
defm V_SMFMAC_I32_32X32X32_I8 : SMFMACInst<"v_smfmac_i32_32x32x32_i8", "I32_32X32X32_I8", int_amdgcn_smfmac_i32_32x32x32_i8>;
defm V_SMFMAC_F32_16X16X64_BF8_BF8 : SMFMACInst<"v_smfmac_f32_16x16x64_bf8_bf8", "F32_16X16X64_F8", int_amdgcn_smfmac_f32_16x16x64_bf8_bf8>;
defm V_SMFMAC_F32_16X16X64_BF8_FP8 : SMFMACInst<"v_smfmac_f32_16x16x64_bf8_fp8", "F32_16X16X64_F8", int_amdgcn_smfmac_f32_16x16x64_bf8_fp8>;
defm V_SMFMAC_F32_16X16X64_FP8_BF8 : SMFMACInst<"v_smfmac_f32_16x16x64_fp8_bf8", "F32_16X16X64_F8", int_amdgcn_smfmac_f32_16x16x64_fp8_bf8>;
defm V_SMFMAC_F32_16X16X64_FP8_FP8 : SMFMACInst<"v_smfmac_f32_16x16x64_fp8_fp8", "F32_16X16X64_F8", int_amdgcn_smfmac_f32_16x16x64_fp8_fp8>;
defm V_SMFMAC_F32_32X32X32_BF8_BF8 : SMFMACInst<"v_smfmac_f32_32x32x32_bf8_bf8", "F32_32X32X32_F8", int_amdgcn_smfmac_f32_32x32x32_bf8_bf8>;
defm V_SMFMAC_F32_32X32X32_BF8_FP8 : SMFMACInst<"v_smfmac_f32_32x32x32_bf8_fp8", "F32_32X32X32_F8", int_amdgcn_smfmac_f32_32x32x32_bf8_fp8>;
defm V_SMFMAC_F32_32X32X32_FP8_BF8 : SMFMACInst<"v_smfmac_f32_32x32x32_fp8_bf8", "F32_32X32X32_F8", int_amdgcn_smfmac_f32_32x32x32_fp8_bf8>;
defm V_SMFMAC_F32_32X32X32_FP8_FP8 : SMFMACInst<"v_smfmac_f32_32x32x32_fp8_fp8", "F32_32X32X32_F8", int_amdgcn_smfmac_f32_32x32x32_fp8_fp8>;
}
def MAIInstInfoTable : GenericTable {
let FilterClass = "MAIInst";
let CppTypeName = "MAIInstInfo";
let Fields = [
"Opcode", "is_dgemm", "is_gfx940_xdl"
];
let PrimaryKey = ["Opcode"];
let PrimaryKeyName = "getMAIInstInfoHelper";
}
let SubtargetPredicate = HasPackedFP32Ops, isCommutable = 1, isReMaterializable = 1 in {
defm V_PK_FMA_F32 : VOP3PInst<"v_pk_fma_f32", VOP3P_Profile<VOP_V2F32_V2F32_V2F32_V2F32, VOP3_PACKED>, any_fma>;
defm V_PK_MUL_F32 : VOP3PInst<"v_pk_mul_f32", VOP3P_Profile<VOP_V2F32_V2F32_V2F32, VOP3_PACKED>, any_fmul>;
defm V_PK_ADD_F32 : VOP3PInst<"v_pk_add_f32", VOP3P_Profile<VOP_V2F32_V2F32_V2F32, VOP3_PACKED>, any_fadd>;
defm V_PK_MOV_B32 : VOP3PInst<"v_pk_mov_b32", VOP3P_Profile<VOP_V2I32_V2I32_V2I32, VOP3_PACKED>>;
} // End SubtargetPredicate = HasPackedFP32Ops, isCommutable = 1
def : MnemonicAlias<"v_accvgpr_read", "v_accvgpr_read_b32">;
def : MnemonicAlias<"v_accvgpr_write", "v_accvgpr_write_b32">;
class VOPProfileWMMA<VOPProfile P, string Suffix, RegisterOperand _Src01RC64, bit _HasClamp, bit _HasOpSel> : VOP3P_Profile<P> {
let DstRC = !if(!eq(Suffix, "_w32"), VDst_256, VDst_128);
let Src0RC64 = _Src01RC64;
let Src1RC64 = _Src01RC64;
let Src2RC64 = !if(!eq(Suffix, "_w32"), VISrc_256_f64, VISrc_128_f32);
let HasClamp = _HasClamp;
let HasOpSel = _HasOpSel;
let IsPacked = 1;
let IsWMMA = 1;
}
def VOP_V8F32_V16F16_V16F16_V8F32 : VOPProfile <[v8f32, v16f16, v16f16, v8f32]>;
def VOP_V8F32_V16I16_V16I16_V8F32 : VOPProfile <[v8f32, v16i16, v16i16, v8f32]>;
def VOP_V16F16_V16F16_V16F16_V16F16 : VOPProfile <[v16f16, v16f16, v16f16, v16f16]>;
def VOP_V16I16_V16I16_V16I16_V16I16 : VOPProfile <[v16i16, v16i16, v16i16, v16i16]>;
def VOP_V8I32_V4I32_V4I32_V8I32 : VOPProfile <[v8i32, v4i32, v4i32, v8i32]>;
def VOP_V8I32_V2I32_V2I32_V8I32 : VOPProfile <[v8i32, v2i32, v2i32, v8i32]>;
def VOP_V4F32_V16F16_V16F16_V4F32 : VOPProfile <[v4f32, v16f16, v16f16, v4f32]>;
def VOP_V4F32_V16I16_V16I16_V4F32 : VOPProfile <[v4f32, v16i16, v16i16, v4f32]>;
def VOP_V8F16_V16F16_V16F16_V8F16 : VOPProfile <[v8f16, v16f16, v16f16, v8f16]>;
def VOP_V8I16_V16I16_V16I16_V8I16 : VOPProfile <[v8i16, v16i16, v16i16, v8i16]>;
def VOP_V4I32_V4I32_V4I32_V4I32 : VOPProfile <[v4i32, v4i32, v4i32, v4i32]>;
def VOP_V4I32_V2I32_V2I32_V4I32 : VOPProfile <[v4i32, v2i32, v2i32, v4i32]>;
class WMMAType <bits<2> val> {
bit hasClamp = val{0};
bit hasOpsel = val{1};
}
def WMMARegular : WMMAType<0b00>;
def WMMAUIClamp : WMMAType<0b01>;
def WMMAOpSel : WMMAType<0b10>;
class WMMARegularPat<Instruction Inst, SDPatternOperator node, VOPProfile P> :
GCNPat < (P.DstVT (node
(P.Src0VT (VOP3PMods P.Src0VT:$src0, i32:$src0_modifiers)),
(P.Src1VT (VOP3PMods P.Src1VT:$src1, i32:$src1_modifiers)),
(P.Src2VT (VOP3PMods P.Src2VT:$src2, i32:$src2_modifiers))
)),
(P.DstVT (Inst i32:$src0_modifiers, P.Src0VT:$src0, i32:$src1_modifiers, P.Src1VT:$src1, $src2_modifiers, P.Src2VT:$src2))
>;
class WMMAOpSelPat<Instruction Inst, SDPatternOperator node, VOPProfile P> :
GCNPat < (P.DstVT (node
(P.Src0VT P.Src0VT:$src0),
(P.Src1VT P.Src1VT:$src1),
(P.Src2VT P.Src2VT:$src2), (WMMAOpSelVOP3PMods i32:$src2_modifiers)
)),
(P.DstVT (Inst (i32 8), P.Src0VT:$src0, (i32 8), P.Src1VT:$src1, i32:$src2_modifiers, P.Src2VT:$src2))
>;
class WMMAUIClampPat<Instruction Inst, SDPatternOperator node, VOPProfile P> :
GCNPat < (P.DstVT (node
(DotIUVOP3PMods i32:$src0_modifiers), (P.Src0VT P.Src0VT:$src0),
(DotIUVOP3PMods i32:$src1_modifiers), (P.Src1VT P.Src1VT:$src1),
(P.Src2VT P.Src2VT:$src2), (i1 timm:$clamp)
)),
(P.DstVT (Inst i32:$src0_modifiers, P.Src0VT:$src0, i32:$src1_modifiers, P.Src1VT:$src1, (i32 8), P.Src2VT:$src2, i1:$clamp))
>;
class WMMAOpcodeMapping<Instruction TwoAddr, Instruction ThreeAddr> {
Instruction Opcode2Addr = TwoAddr;
Instruction Opcode3Addr = ThreeAddr;
Predicate WaveSizePredicate;
}
def WMMAOpcode : GenericEnum {
let FilterClass = "VOP3P_Pseudo";
}
class WMMAMappingTable : GenericTable {
let FilterClass = "WMMAOpcodeMapping";
let CppTypeName = "WMMAOpcodeMappingInfo";
let Fields = ["Opcode2Addr", "Opcode3Addr"];
string TypeOf_Opcode2Addr = "WMMAOpcode";
string TypeOf_Opcode3Addr = "WMMAOpcode";
}
def WMMAOpcode2AddrMappingTable : WMMAMappingTable {
let PrimaryKey = ["Opcode2Addr"];
let PrimaryKeyName = "getWMMAMappingInfoFrom2AddrOpcode";
}
def WMMAOpcode3AddrMappingTable : WMMAMappingTable {
let PrimaryKey = ["Opcode3Addr"];
let PrimaryKeyName = "getWMMAMappingInfoFrom3AddrOpcode";
}
// The WMMA instruction has extra constraints:
// Matrices A and B cannot overlap with D. C cannot partially overlap with D,
// but it is OK for them to be the same (which is a typical case).
//
// We implement it as follows:
// 1) Map the intrinsic to the pseudo where D is tied to C ($vdst = $src2).
// 2) The pass twoaddressinstruction checks if src2 is live and if that is the case
// it converts the default pseudo to the pseudo where src2 is not the same as vdst.
// 3) @earlyclobber on the destination satisfies the constraint during RA.
multiclass WMMAInst<string Suffix, string Instr, VOPProfile P, SDPatternOperator node = null_frag, RegisterOperand _Src01RC64 = VRegSrc_256, WMMAType Type> {
defvar WMMAConstraints2Addr = "@earlyclobber $vdst,$vdst = $src2";
defvar WMMAConstraints3Addr = "@earlyclobber $vdst";
defvar WMMAProfile = VOPProfileWMMA<P, Suffix, _Src01RC64, Type.hasClamp, Type.hasOpsel>;
if !eq(Suffix, "_w32") then {
let Mnemonic = Instr, mayRaiseFPException = 0, ReadsModeReg = 0 in {
let Constraints = WMMAConstraints2Addr, isConvertibleToThreeAddress = 1 in {
def _twoaddr_w32 : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
}
let Constraints = WMMAConstraints3Addr, SchedRW = [Write32Bit, Write32Bit] in {
def _threeaddr_w32 : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
}
}
def : WMMAOpcodeMapping<!cast<Instruction>(NAME # _twoaddr_w32),
!cast<Instruction>(NAME # _threeaddr_w32)>;
} else if !eq(Suffix, "_w64") then {
let Mnemonic = Instr, mayRaiseFPException = 0, ReadsModeReg = 0 in {
let Constraints = WMMAConstraints2Addr, isConvertibleToThreeAddress = 1 in {
def _twoaddr_w64 : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
}
let Constraints = WMMAConstraints3Addr, SchedRW = [Write32Bit, Write32Bit] in {
def _threeaddr_w64 : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
}
}
def : WMMAOpcodeMapping<!cast<Instruction>(NAME # _twoaddr_w64),
!cast<Instruction>(NAME # _threeaddr_w64)>;
}
if !eq(Type, WMMAOpSel) then {
def : WMMAOpSelPat<!cast<Instruction>(NAME # _twoaddr # Suffix), node, P>;
} else if !eq(Type, WMMAUIClamp) then {
def : WMMAUIClampPat<!cast<Instruction>(NAME # _twoaddr # Suffix), node, P>;
} else {
def : WMMARegularPat<!cast<Instruction>(NAME # _twoaddr # Suffix), node, P>;
}
}
let WaveSizePredicate = isWave32 in {
defm V_WMMA_F32_16X16X16_F16 : WMMAInst<"_w32", "v_wmma_f32_16x16x16_f16", VOP_V8F32_V16F16_V16F16_V8F32, int_amdgcn_wmma_f32_16x16x16_f16, VRegSrc_256, WMMARegular>;
defm V_WMMA_F32_16X16X16_BF16 : WMMAInst<"_w32", "v_wmma_f32_16x16x16_bf16", VOP_V8F32_V16I16_V16I16_V8F32, int_amdgcn_wmma_f32_16x16x16_bf16, VRegSrc_256, WMMARegular>;
defm V_WMMA_F16_16X16X16_F16 : WMMAInst<"_w32", "v_wmma_f16_16x16x16_f16", VOP_V16F16_V16F16_V16F16_V16F16, int_amdgcn_wmma_f16_16x16x16_f16, VRegSrc_256, WMMAOpSel>;
defm V_WMMA_BF16_16X16X16_BF16 : WMMAInst<"_w32", "v_wmma_bf16_16x16x16_bf16", VOP_V16I16_V16I16_V16I16_V16I16, int_amdgcn_wmma_bf16_16x16x16_bf16, VRegSrc_256, WMMAOpSel>;
defm V_WMMA_I32_16X16X16_IU8 : WMMAInst<"_w32", "v_wmma_i32_16x16x16_iu8", VOP_V8I32_V4I32_V4I32_V8I32, int_amdgcn_wmma_i32_16x16x16_iu8, VRegSrc_128, WMMAUIClamp>;
defm V_WMMA_I32_16X16X16_IU4 : WMMAInst<"_w32", "v_wmma_i32_16x16x16_iu4", VOP_V8I32_V2I32_V2I32_V8I32, int_amdgcn_wmma_i32_16x16x16_iu4, VRegSrc_64, WMMAUIClamp>;
}
let WaveSizePredicate = isWave64 in {
defm V_WMMA_F32_16X16X16_F16 : WMMAInst<"_w64", "v_wmma_f32_16x16x16_f16", VOP_V4F32_V16F16_V16F16_V4F32, int_amdgcn_wmma_f32_16x16x16_f16, VRegSrc_256, WMMARegular>;
defm V_WMMA_F32_16X16X16_BF16 : WMMAInst<"_w64", "v_wmma_f32_16x16x16_bf16", VOP_V4F32_V16I16_V16I16_V4F32, int_amdgcn_wmma_f32_16x16x16_bf16, VRegSrc_256, WMMARegular>;
defm V_WMMA_F16_16X16X16_F16 : WMMAInst<"_w64", "v_wmma_f16_16x16x16_f16", VOP_V8F16_V16F16_V16F16_V8F16, int_amdgcn_wmma_f16_16x16x16_f16, VRegSrc_256, WMMAOpSel>;
defm V_WMMA_BF16_16X16X16_BF16 : WMMAInst<"_w64", "v_wmma_bf16_16x16x16_bf16", VOP_V8I16_V16I16_V16I16_V8I16, int_amdgcn_wmma_bf16_16x16x16_bf16, VRegSrc_256, WMMAOpSel>;
defm V_WMMA_I32_16X16X16_IU8 : WMMAInst<"_w64", "v_wmma_i32_16x16x16_iu8", VOP_V4I32_V4I32_V4I32_V4I32, int_amdgcn_wmma_i32_16x16x16_iu8, VRegSrc_128, WMMAUIClamp>;
defm V_WMMA_I32_16X16X16_IU4 : WMMAInst<"_w64", "v_wmma_i32_16x16x16_iu4", VOP_V4I32_V2I32_V2I32_V4I32, int_amdgcn_wmma_i32_16x16x16_iu4, VRegSrc_64, WMMAUIClamp>;
}
//===----------------------------------------------------------------------===//
// Begin Real Encodings
//===----------------------------------------------------------------------===//
class VOP3P_DPP16<bits<7> op, VOP_DPP_Pseudo ps, int subtarget,
string opName = ps.OpName>
: VOP3P_DPP<op, opName, ps.Pfl, 1>, SIMCInstr<ps.PseudoInstr, subtarget> {
let hasSideEffects = ps.hasSideEffects;
let Defs = ps.Defs;
let SchedRW = ps.SchedRW;
let Uses = ps.Uses;
let AssemblerPredicate = HasDPP16;
let SubtargetPredicate = HasDPP16;
let OtherPredicates = ps.OtherPredicates;
}
class VOP3P_DPP8_Base<bits<7> op, VOP_Pseudo ps, string opName = ps.OpName>
: VOP3P_DPP8<op, opName, ps.Pfl> {
let hasSideEffects = ps.hasSideEffects;
let Defs = ps.Defs;
let SchedRW = ps.SchedRW;
let Uses = ps.Uses;
let OtherPredicates = ps.OtherPredicates;
}
//===----------------------------------------------------------------------===//
// GFX11.
//===----------------------------------------------------------------------===//
let AssemblerPredicate = isGFX11Plus,
DecoderNamespace = "GFX11" in {
multiclass VOP3P_Real_gfx11<bits<7> op, string backing_ps_name = NAME,
string asmName = !cast<VOP3P_Pseudo>(NAME).Mnemonic> {
def _gfx11 : VOP3P_Real<!cast<VOP3P_Pseudo>(backing_ps_name),
SIEncodingFamily.GFX11, asmName>,
VOP3Pe_gfx11<op, !cast<VOP3P_Pseudo>(backing_ps_name).Pfl>;
}
multiclass VOP3P_Real_dpp_gfx11<bits<7> op, string backing_ps_name = NAME,
string asmName = !cast<VOP3P_Pseudo>(NAME).Mnemonic> {
defvar ps = !cast<VOP3P_Pseudo>(backing_ps_name);
def _dpp_gfx11
: VOP3P_DPP16<op, !cast<VOP_DPP_Pseudo>(backing_ps_name #"_dpp"),
SIEncodingFamily.GFX11> {
let AsmString = asmName #ps.Pfl.AsmVOP3DPP16;
let DecoderNamespace = "DPPGFX11";
}
}
multiclass VOP3P_Real_dpp8_gfx11<bits<7> op, string backing_ps_name = NAME,
string asmName = !cast<VOP3P_Pseudo>(NAME).Mnemonic> {
defvar ps = !cast<VOP3P_Pseudo>(backing_ps_name);
def _dpp8_gfx11 : VOP3P_DPP8_Base<op, ps> {
let AsmString = asmName #ps.Pfl.AsmVOP3DPP8;
let DecoderNamespace = "DPP8GFX11";
}
}
multiclass VOP3P_Realtriple_gfx11<bits<7> op, string backing_ps_name = NAME,
string asmName = !cast<VOP3P_Pseudo>(NAME).Mnemonic>
: VOP3P_Real_gfx11<op, backing_ps_name, asmName>,
VOP3P_Real_dpp_gfx11<op, backing_ps_name, asmName>,
VOP3P_Real_dpp8_gfx11<op, backing_ps_name, asmName>;
} // End AssemblerPredicate = isGFX11Plus, DecoderNamespace = "GFX11"
defm V_DOT4_I32_IU8 : VOP3P_Real_gfx11 <0x16>;
defm V_DOT8_I32_IU4 : VOP3P_Real_gfx11 <0x18>;
defm V_DOT2_F32_BF16 : VOP3P_Real_gfx11 <0x1a>;
multiclass VOP3P_Real_WMMA <bits<7> op> {
let WaveSizePredicate = isWave32, DecoderNamespace = "GFX11" in {
defm _twoaddr_w32 : VOP3P_Real_gfx11 <op>;
}
let WaveSizePredicate = isWave64, DecoderNamespace = "WMMAGFX11" in {
defm _twoaddr_w64 : VOP3P_Real_gfx11 <op>;
}
}
defm V_WMMA_F32_16X16X16_F16 : VOP3P_Real_WMMA <0x040>;
defm V_WMMA_F32_16X16X16_BF16 : VOP3P_Real_WMMA <0x041>;
defm V_WMMA_F16_16X16X16_F16 : VOP3P_Real_WMMA <0x042>;
defm V_WMMA_BF16_16X16X16_BF16 : VOP3P_Real_WMMA <0x043>;
defm V_WMMA_I32_16X16X16_IU8 : VOP3P_Real_WMMA <0x044>;
defm V_WMMA_I32_16X16X16_IU4 : VOP3P_Real_WMMA <0x045>;
//===----------------------------------------------------------------------===//
// GFX8 (VI)
//===----------------------------------------------------------------------===//
multiclass VOP3P_Real_vi<bits<7> op> {
def _vi : VOP3P_Real<!cast<VOP3_Pseudo>(NAME), SIEncodingFamily.VI>,
VOP3Pe <op, !cast<VOP3_Pseudo>(NAME).Pfl> {
let AssemblerPredicate = HasVOP3PInsts;
let DecoderNamespace = "GFX8";
let VOP3P = 1;
}
}
multiclass VOP3P_Real_MAI<bits<7> op> {
def _vi : VOP3P_Real<!cast<VOP3_Pseudo>(NAME#"_e64"), SIEncodingFamily.VI>,
VOP3Pe_MAI <op, !cast<VOP3_Pseudo>(NAME#"_e64").Pfl, ?> {
let AssemblerPredicate = HasMAIInsts;
let DecoderNamespace = "GFX8";
let Inst{14} = ?; // op_sel_hi(2)
let Inst{59} = ?; // op_sel_hi(0)
let Inst{60} = ?; // op_sel_hi(1)
}
}
let Constraints = "" in {
multiclass VOP3P_Real_MFMA_gfx90a<bits<7> op> {
let SubtargetPredicate = isGFX90AOnly,
AssemblerPredicate = isGFX90AOnly, DecoderNamespace = "GFX90A" in {
def _gfx90a_acd : VOP3P_Real<!cast<VOP3_Pseudo>(NAME#"_e64"), SIEncodingFamily.GFX90A>,
VOP3Pe_MAI <op, !cast<VOP3_Pseudo>(NAME#"_e64").Pfl, 1>;
def _gfx90a_vcd : VOP3P_Real<!cast<VOP3_Pseudo>(NAME # "_vgprcd" # "_e64"), SIEncodingFamily.GFX90A>,
VOP3Pe_MAI <op, !cast<VOP3_Pseudo>(NAME # "_vgprcd" # "_e64").Pfl, 0>;
} // End AssemblerPredicate = isGFX90AOnly, DecoderNamespace = "GFX90A"
}
}
multiclass VOP3P_Real_MFMA_gfx940_aliases<string NameFrom, string NameTo, string Op,
VOP3_Pseudo PS_ACD = !cast<VOP3_Pseudo>(Op # "_e64"),
VOP3_Pseudo PS_VCD = !cast<VOP3_Pseudo>(Op # "_vgprcd" # "_e64"),
VOPProfile Pfl_ACD = PS_ACD.Pfl,
VOPProfile Pfl_VCD = PS_VCD.Pfl> {
let Predicates = [isGFX940Plus] in {
if !ne(NameFrom, NameTo) then {
def : InstAlias <NameTo # " " # PS_ACD.AsmOperands,
(!cast<VOP3P_Real>(Op # "_gfx940_acd") Pfl_ACD.DstRC:$vdst,
Pfl_ACD.Src0RC64:$src0, Pfl_ACD.Src1RC64:$src1, Pfl_ACD.Src2RC64:$src2,
cbsz:$cbsz, abid:$abid, blgp:$blgp)>, PredicateControl;
def : InstAlias <NameTo # " " # PS_VCD.AsmOperands,
(!cast<VOP3P_Real>(Op # "_gfx940_vcd") Pfl_VCD.DstRC:$vdst,
Pfl_VCD.Src0RC64:$src0, Pfl_VCD.Src1RC64:$src1, Pfl_VCD.Src2RC64:$src2,
cbsz:$cbsz, abid:$abid, blgp:$blgp)>, PredicateControl;
}
} // End Predicates = [isGFX940Plus]
}
multiclass VOP3P_Real_MFMA_gfx940<bits<7> op, string Name = !cast<VOP3_Pseudo>(NAME#"_e64").Mnemonic,
VOP3_Pseudo PS_ACD = !cast<VOP3_Pseudo>(NAME # "_e64"),
VOP3_Pseudo PS_VCD = !cast<VOP3_Pseudo>(NAME # "_vgprcd" # "_e64")> {
let SubtargetPredicate = isGFX940Plus,
AssemblerPredicate = isGFX940Plus, DecoderNamespace = "GFX940",
AsmString = Name # PS_ACD.AsmOperands, Constraints = "" in {
def _gfx940_acd : VOP3P_Real<PS_ACD, SIEncodingFamily.GFX940>,
VOP3Pe_MAI <op, PS_ACD.Pfl, 1>;
def _gfx940_vcd : VOP3P_Real<PS_VCD, SIEncodingFamily.GFX940>,
VOP3Pe_MAI <op, PS_VCD.Pfl, 0>;
} // End AssemblerPredicate = isGFX940Plus, DecoderNamespace = "GFX940"
defm : VOP3P_Real_MFMA_gfx940_aliases<Name, PS_ACD.Mnemonic, NAME>;
if !ne(!subst("_1k", "", PS_ACD.Mnemonic), PS_ACD.Mnemonic) then
defm : VOP3P_Real_MFMA_gfx940_aliases<Name, !subst("_1k", "", PS_ACD.Mnemonic), NAME>;
}
multiclass VOP3P_Real_MFMA<bits<7> op, string GFX940Name = !cast<VOP3_Pseudo>(NAME#"_e64").Mnemonic> :
VOP3P_Real_MFMA_gfx90a <op>,
VOP3P_Real_MFMA_gfx940 <op, GFX940Name> {
def _vi : VOP3P_Real<!cast<VOP3_Pseudo>(NAME#"_e64"), SIEncodingFamily.VI>,
VOP3Pe_MAI <op, !cast<VOP3_Pseudo>(NAME#"_e64").Pfl, ?> {
let AssemblerPredicate = HasMAIInsts;
let DecoderNamespace = "GFX8";
let Constraints = "";
}
}
multiclass VOP3P_Real_SMFMAC<bits<7> op, string alias> {
def _gfx940 : VOP3P_Real<!cast<VOP3_Pseudo>(NAME#"_e64"), SIEncodingFamily.VI>,
VOP3Pe_SMFMAC <op> {
let AssemblerPredicate = isGFX940Plus;
let DecoderNamespace = "GFX8";
}
def : MnemonicAlias<alias, !cast<VOP3_Pseudo>(NAME#"_e64").Mnemonic>;
}
defm V_PK_MAD_I16 : VOP3P_Real_vi <0x00>;
defm V_PK_MUL_LO_U16 : VOP3P_Real_vi <0x01>;
defm V_PK_ADD_I16 : VOP3P_Real_vi <0x02>;
defm V_PK_SUB_I16 : VOP3P_Real_vi <0x03>;
defm V_PK_LSHLREV_B16 : VOP3P_Real_vi <0x04>;
defm V_PK_LSHRREV_B16 : VOP3P_Real_vi <0x05>;
defm V_PK_ASHRREV_I16 : VOP3P_Real_vi <0x06>;
defm V_PK_MAX_I16 : VOP3P_Real_vi <0x07>;
defm V_PK_MIN_I16 : VOP3P_Real_vi <0x08>;
defm V_PK_MAD_U16 : VOP3P_Real_vi <0x09>;
defm V_PK_ADD_U16 : VOP3P_Real_vi <0x0a>;
defm V_PK_SUB_U16 : VOP3P_Real_vi <0x0b>;
defm V_PK_MAX_U16 : VOP3P_Real_vi <0x0c>;
defm V_PK_MIN_U16 : VOP3P_Real_vi <0x0d>;
defm V_PK_FMA_F16 : VOP3P_Real_vi <0x0e>;
defm V_PK_ADD_F16 : VOP3P_Real_vi <0x0f>;
defm V_PK_MUL_F16 : VOP3P_Real_vi <0x10>;
defm V_PK_MIN_F16 : VOP3P_Real_vi <0x11>;
defm V_PK_MAX_F16 : VOP3P_Real_vi <0x12>;
let SubtargetPredicate = HasMadMixInsts in {
defm V_MAD_MIX_F32 : VOP3P_Real_vi <0x20>;
defm V_MAD_MIXLO_F16 : VOP3P_Real_vi <0x21>;
defm V_MAD_MIXHI_F16 : VOP3P_Real_vi <0x22>;
}
let SubtargetPredicate = HasFmaMixInsts in {
let DecoderNamespace = "GFX9_DL" in {
// The mad_mix instructions were renamed and their behaviors changed,
// but the opcode stayed the same so we need to put these in a
// different DecoderNamespace to avoid the ambiguity.
defm V_FMA_MIX_F32 : VOP3P_Real_vi <0x20>;
defm V_FMA_MIXLO_F16 : VOP3P_Real_vi <0x21>;
defm V_FMA_MIXHI_F16 : VOP3P_Real_vi <0x22>;
}
}
defm V_DOT2_I32_I16 : VOP3P_Real_vi <0x26>;
defm V_DOT2_U32_U16 : VOP3P_Real_vi <0x27>;
defm V_DOT2_F32_F16 : VOP3P_Real_vi <0x23>;
defm V_DOT4_U32_U8 : VOP3P_Real_vi <0x29>;
defm V_DOT8_U32_U4 : VOP3P_Real_vi <0x2b>;
defm V_DOT4_I32_I8 : VOP3P_Real_vi <0x28>;
defm V_DOT8_I32_I4 : VOP3P_Real_vi <0x2a>;
let SubtargetPredicate = HasMAIInsts in {
defm V_ACCVGPR_READ_B32 : VOP3P_Real_MAI <0x58>;
defm V_ACCVGPR_WRITE_B32 : VOP3P_Real_MAI <0x59>;
defm V_MFMA_F32_32X32X1F32 : VOP3P_Real_MFMA <0x40, "v_mfma_f32_32x32x1_2b_f32">;
defm V_MFMA_F32_16X16X1F32 : VOP3P_Real_MFMA <0x41, "v_mfma_f32_16x16x1_4b_f32">;
defm V_MFMA_F32_4X4X1F32 : VOP3P_Real_MFMA <0x42, "v_mfma_f32_4x4x1_16b_f32">;
defm V_MFMA_F32_32X32X2F32 : VOP3P_Real_MFMA <0x44, "v_mfma_f32_32x32x2_f32">;
defm V_MFMA_F32_16X16X4F32 : VOP3P_Real_MFMA <0x45, "v_mfma_f32_16x16x4_f32">;
defm V_MFMA_F32_32X32X4F16 : VOP3P_Real_MFMA <0x48, "v_mfma_f32_32x32x4_2b_f16">;
defm V_MFMA_F32_16X16X4F16 : VOP3P_Real_MFMA <0x49, "v_mfma_f32_16x16x4_4b_f16">;
defm V_MFMA_F32_4X4X4F16 : VOP3P_Real_MFMA <0x4a, "v_mfma_f32_4x4x4_16b_f16">;
defm V_MFMA_F32_32X32X8F16 : VOP3P_Real_MFMA <0x4c, "v_mfma_f32_32x32x8_f16">;
defm V_MFMA_F32_16X16X16F16 : VOP3P_Real_MFMA <0x4d, "v_mfma_f32_16x16x16_f16">;
defm V_MFMA_I32_32X32X4I8 : VOP3P_Real_MFMA <0x50, "v_mfma_i32_32x32x4_2b_i8">;
defm V_MFMA_I32_16X16X4I8 : VOP3P_Real_MFMA <0x51, "v_mfma_i32_16x16x4_4b_i8">;
defm V_MFMA_I32_4X4X4I8 : VOP3P_Real_MFMA <0x52, "v_mfma_i32_4x4x4_16b_i8">;
let SubtargetPredicate = isGFX908orGFX90A in {
defm V_MFMA_I32_16X16X16I8 : VOP3P_Real_MFMA <0x55>;
defm V_MFMA_I32_32X32X8I8 : VOP3P_Real_MFMA <0x54>;
defm V_MFMA_F32_32X32X2BF16 : VOP3P_Real_MFMA <0x68>;
defm V_MFMA_F32_16X16X2BF16 : VOP3P_Real_MFMA <0x69>;
defm V_MFMA_F32_4X4X2BF16 : VOP3P_Real_MFMA <0x6b>;
defm V_MFMA_F32_32X32X4BF16 : VOP3P_Real_MFMA <0x6c>;
defm V_MFMA_F32_16X16X8BF16 : VOP3P_Real_MFMA <0x6d>;
}
} // End SubtargetPredicate = HasMAIInsts
defm V_MFMA_F32_32X32X4BF16_1K : VOP3P_Real_MFMA_gfx90a <0x63>;
defm V_MFMA_F32_16X16X4BF16_1K : VOP3P_Real_MFMA_gfx90a <0x64>;
defm V_MFMA_F32_4X4X4BF16_1K : VOP3P_Real_MFMA_gfx90a <0x65>;
defm V_MFMA_F32_32X32X8BF16_1K : VOP3P_Real_MFMA_gfx90a <0x66>;
defm V_MFMA_F32_16X16X16BF16_1K : VOP3P_Real_MFMA_gfx90a <0x67>;
defm V_MFMA_F64_16X16X4F64 : VOP3P_Real_MFMA_gfx90a <0x6e>;
defm V_MFMA_F64_4X4X4F64 : VOP3P_Real_MFMA_gfx90a <0x6f>;
defm V_MFMA_I32_32X32X16I8 : VOP3P_Real_MFMA_gfx940 <0x56, "v_mfma_i32_32x32x16_i8">;
defm V_MFMA_I32_16X16X32I8 : VOP3P_Real_MFMA_gfx940 <0x57, "v_mfma_i32_16x16x32_i8">;
defm V_MFMA_F32_16X16X8XF32 : VOP3P_Real_MFMA_gfx940 <0x3e, "v_mfma_f32_16x16x8_xf32">;
defm V_MFMA_F32_32X32X4XF32 : VOP3P_Real_MFMA_gfx940 <0x3f, "v_mfma_f32_32x32x4_xf32">;
defm V_MFMA_F32_16X16X32_BF8_BF8 : VOP3P_Real_MFMA_gfx940 <0x70>;
defm V_MFMA_F32_16X16X32_BF8_FP8 : VOP3P_Real_MFMA_gfx940 <0x71>;
defm V_MFMA_F32_16X16X32_FP8_BF8 : VOP3P_Real_MFMA_gfx940 <0x72>;
defm V_MFMA_F32_16X16X32_FP8_FP8 : VOP3P_Real_MFMA_gfx940 <0x73>;
defm V_MFMA_F32_32X32X16_BF8_BF8 : VOP3P_Real_MFMA_gfx940 <0x74>;
defm V_MFMA_F32_32X32X16_BF8_FP8 : VOP3P_Real_MFMA_gfx940 <0x75>;
defm V_MFMA_F32_32X32X16_FP8_BF8 : VOP3P_Real_MFMA_gfx940 <0x76>;
defm V_MFMA_F32_32X32X16_FP8_FP8 : VOP3P_Real_MFMA_gfx940 <0x77>;
defm V_MFMA_F32_32X32X4BF16_1K : VOP3P_Real_MFMA_gfx940 <0x5d, "v_mfma_f32_32x32x4_2b_bf16">;
defm V_MFMA_F32_16X16X4BF16_1K : VOP3P_Real_MFMA_gfx940 <0x5e, "v_mfma_f32_16x16x4_4b_bf16">;
defm V_MFMA_F32_4X4X4BF16_1K : VOP3P_Real_MFMA_gfx940 <0x5f, "v_mfma_f32_4x4x4_16b_bf16">;
defm V_MFMA_F32_32X32X8BF16_1K : VOP3P_Real_MFMA_gfx940 <0x60, "v_mfma_f32_32x32x8_bf16">;
defm V_MFMA_F32_16X16X16BF16_1K : VOP3P_Real_MFMA_gfx940 <0x61, "v_mfma_f32_16x16x16_bf16">;
defm V_MFMA_F64_16X16X4F64 : VOP3P_Real_MFMA_gfx940 <0x6e, "v_mfma_f64_16x16x4_f64">;
defm V_MFMA_F64_4X4X4F64 : VOP3P_Real_MFMA_gfx940 <0x6f, "v_mfma_f64_4x4x4_4b_f64">;
defm V_SMFMAC_F32_16X16X32_F16 : VOP3P_Real_SMFMAC <0x62, "v_smfmac_f32_16x16x32f16">;
defm V_SMFMAC_F32_32X32X16_F16 : VOP3P_Real_SMFMAC <0x64, "v_smfmac_f32_32x32x16f16">;
defm V_SMFMAC_F32_16X16X32_BF16 : VOP3P_Real_SMFMAC <0x66, "v_smfmac_f32_16x16x32bf16">;
defm V_SMFMAC_F32_32X32X16_BF16 : VOP3P_Real_SMFMAC <0x68, "v_smfmac_f32_32x32x16bf16">;
defm V_SMFMAC_I32_16X16X64_I8 : VOP3P_Real_SMFMAC <0x6a, "v_smfmac_i32_16x16x64i8">;
defm V_SMFMAC_I32_32X32X32_I8 : VOP3P_Real_SMFMAC <0x6c, "v_smfmac_i32_32x32x32i8">;
defm V_SMFMAC_F32_16X16X64_BF8_BF8 : VOP3P_Real_SMFMAC <0x78, "v_smfmac_f32_16x16x64bf8bf8">;
defm V_SMFMAC_F32_16X16X64_BF8_FP8 : VOP3P_Real_SMFMAC <0x79, "v_smfmac_f32_16x16x64bf8fp8">;
defm V_SMFMAC_F32_16X16X64_FP8_BF8 : VOP3P_Real_SMFMAC <0x7a, "v_smfmac_f32_16x16x64fp8bf8">;
defm V_SMFMAC_F32_16X16X64_FP8_FP8 : VOP3P_Real_SMFMAC <0x7b, "v_smfmac_f32_16x16x64fp8fp8">;
defm V_SMFMAC_F32_32X32X32_BF8_BF8 : VOP3P_Real_SMFMAC <0x7c, "v_smfmac_f32_32x32x32bf8bf8">;
defm V_SMFMAC_F32_32X32X32_BF8_FP8 : VOP3P_Real_SMFMAC <0x7d, "v_smfmac_f32_32x32x32bf8fp8">;
defm V_SMFMAC_F32_32X32X32_FP8_BF8 : VOP3P_Real_SMFMAC <0x7e, "v_smfmac_f32_32x32x32fp8bf8">;
defm V_SMFMAC_F32_32X32X32_FP8_FP8 : VOP3P_Real_SMFMAC <0x7f, "v_smfmac_f32_32x32x32fp8fp8">;
let SubtargetPredicate = HasPackedFP32Ops in {
defm V_PK_FMA_F32 : VOP3P_Real_vi <0x30>;
defm V_PK_MUL_F32 : VOP3P_Real_vi <0x31>;
defm V_PK_ADD_F32 : VOP3P_Real_vi <0x32>;
defm V_PK_MOV_B32 : VOP3P_Real_vi <0x33>;
} // End SubtargetPredicate = HasPackedFP32Ops
//===----------------------------------------------------------------------===//
// GFX10.
//===----------------------------------------------------------------------===//
let AssemblerPredicate = isGFX10Only, DecoderNamespace = "GFX10", VOP3P = 1 in {
multiclass VOP3P_Real_gfx10<bits<7> op> {
def _gfx10 : VOP3P_Real<!cast<VOP3P_Pseudo>(NAME), SIEncodingFamily.GFX10>,
VOP3Pe_gfx10 <op, !cast<VOP3P_Pseudo>(NAME).Pfl>;
}
} // End AssemblerPredicate = isGFX10Only, DecoderNamespace = "GFX10", VOP3P = 1
multiclass VOP3P_Real_gfx10_gfx11<bits<7> op>
: VOP3P_Real_gfx10<op>, VOP3P_Real_gfx11<op>;
multiclass VOP3P_Real_gfx10_gfx11_Triple<bits<7> op>
: VOP3P_Real_gfx10<op>, VOP3P_Realtriple_gfx11<op>;
defm V_PK_MAD_I16 : VOP3P_Real_gfx10_gfx11<0x00>;
defm V_PK_MUL_LO_U16 : VOP3P_Real_gfx10_gfx11<0x01>;
defm V_PK_ADD_I16 : VOP3P_Real_gfx10_gfx11<0x02>;
defm V_PK_SUB_I16 : VOP3P_Real_gfx10_gfx11<0x03>;
defm V_PK_LSHLREV_B16 : VOP3P_Real_gfx10_gfx11<0x04>;
defm V_PK_LSHRREV_B16 : VOP3P_Real_gfx10_gfx11<0x05>;
defm V_PK_ASHRREV_I16 : VOP3P_Real_gfx10_gfx11<0x06>;
defm V_PK_MAX_I16 : VOP3P_Real_gfx10_gfx11<0x07>;
defm V_PK_MIN_I16 : VOP3P_Real_gfx10_gfx11<0x08>;
defm V_PK_MAD_U16 : VOP3P_Real_gfx10_gfx11<0x09>;
defm V_PK_ADD_U16 : VOP3P_Real_gfx10_gfx11<0x0a>;
defm V_PK_SUB_U16 : VOP3P_Real_gfx10_gfx11<0x0b>;
defm V_PK_MAX_U16 : VOP3P_Real_gfx10_gfx11<0x0c>;
defm V_PK_MIN_U16 : VOP3P_Real_gfx10_gfx11<0x0d>;
defm V_PK_FMA_F16 : VOP3P_Real_gfx10_gfx11<0x0e>;
defm V_PK_ADD_F16 : VOP3P_Real_gfx10_gfx11<0x0f>;
defm V_PK_MUL_F16 : VOP3P_Real_gfx10_gfx11<0x10>;
defm V_PK_MIN_F16 : VOP3P_Real_gfx10_gfx11<0x11>;
defm V_PK_MAX_F16 : VOP3P_Real_gfx10_gfx11<0x12>;
defm V_FMA_MIX_F32 : VOP3P_Real_gfx10_gfx11_Triple <0x20>;
defm V_FMA_MIXLO_F16 : VOP3P_Real_gfx10_gfx11_Triple <0x21>;
defm V_FMA_MIXHI_F16 : VOP3P_Real_gfx10_gfx11_Triple <0x22>;
defm V_DOT2_I32_I16 : VOP3P_Real_gfx10 <0x14>;
defm V_DOT2_U32_U16 : VOP3P_Real_gfx10 <0x15>;
defm V_DOT2_F32_F16 : VOP3P_Real_gfx10_gfx11_Triple <0x13>;
defm V_DOT4_U32_U8 : VOP3P_Real_gfx10_gfx11 <0x17>;
defm V_DOT8_U32_U4 : VOP3P_Real_gfx10_gfx11 <0x19>;
defm V_DOT4_I32_I8 : VOP3P_Real_gfx10 <0x16>;
defm V_DOT8_I32_I4 : VOP3P_Real_gfx10 <0x18>;
|
