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
| /*===- TableGen'erated file -------------------------------------*- C++ -*-===*\
|* *|
|* Subtarget Enumeration Source Fragment *|
|* *|
|* Automatically generated file, do not edit! *|
|* *|
\*===----------------------------------------------------------------------===*/
#ifdef GET_SUBTARGETINFO_ENUM
#undef GET_SUBTARGETINFO_ENUM
namespace llvm {
namespace AMDGPU {
enum {
Feature16BitInsts = 0,
FeatureAddNoCarryInsts = 1,
FeatureApertureRegs = 2,
FeatureAtomicFaddInsts = 3,
FeatureAutoWaitcntBeforeBarrier = 4,
FeatureCIInsts = 5,
FeatureCodeObjectV3 = 6,
FeatureCuMode = 7,
FeatureDLInsts = 8,
FeatureDPP = 9,
FeatureDPP8 = 10,
FeatureDisable = 11,
FeatureDoesNotSupportSRAMECC = 12,
FeatureDoesNotSupportXNACK = 13,
FeatureDot1Insts = 14,
FeatureDot2Insts = 15,
FeatureDot3Insts = 16,
FeatureDot4Insts = 17,
FeatureDot5Insts = 18,
FeatureDot6Insts = 19,
FeatureDumpCode = 20,
FeatureDumpCodeLower = 21,
FeatureEnableDS128 = 22,
FeatureEnableLoadStoreOpt = 23,
FeatureEnablePRTStrictNull = 24,
FeatureEnableSIScheduler = 25,
FeatureEnableUnsafeDSOffsetFolding = 26,
FeatureFMA = 27,
FeatureFP16Denormals = 28,
FeatureFP32Denormals = 29,
FeatureFP64 = 30,
FeatureFP64Denormals = 31,
FeatureFP64FP16Denormals = 32,
FeatureFPExceptions = 33,
FeatureFastFMAF32 = 34,
FeatureFlatAddressSpace = 35,
FeatureFlatForGlobal = 36,
FeatureFlatGlobalInsts = 37,
FeatureFlatInstOffsets = 38,
FeatureFlatScratchInsts = 39,
FeatureFlatSegmentOffsetBug = 40,
FeatureFmaMixInsts = 41,
FeatureGCN3Encoding = 42,
FeatureGFX7GFX8GFX9Insts = 43,
FeatureGFX8Insts = 44,
FeatureGFX9 = 45,
FeatureGFX9Insts = 46,
FeatureGFX10 = 47,
FeatureGFX10Insts = 48,
FeatureInstFwdPrefetchBug = 49,
FeatureIntClamp = 50,
FeatureInv2PiInlineImm = 51,
FeatureLDSBankCount16 = 52,
FeatureLDSBankCount32 = 53,
FeatureLdsBranchVmemWARHazard = 54,
FeatureLdsMisalignedBug = 55,
FeatureLocalMemorySize0 = 56,
FeatureLocalMemorySize32768 = 57,
FeatureLocalMemorySize65536 = 58,
FeatureMAIInsts = 59,
FeatureMFMAInlineLiteralBug = 60,
FeatureMIMG_R128 = 61,
FeatureMadMixInsts = 62,
FeatureMaxPrivateElementSize4 = 63,
FeatureMaxPrivateElementSize8 = 64,
FeatureMaxPrivateElementSize16 = 65,
FeatureMovrel = 66,
FeatureNSAEncoding = 67,
FeatureNSAtoVMEMBug = 68,
FeatureNoDataDepHazard = 69,
FeatureNoSdstCMPX = 70,
FeatureOffset3fBug = 71,
FeaturePkFmacF16Inst = 72,
FeaturePromoteAlloca = 73,
FeatureR128A16 = 74,
FeatureRegisterBanking = 75,
FeatureSDWA = 76,
FeatureSDWAMac = 77,
FeatureSDWAOmod = 78,
FeatureSDWAOutModsVOPC = 79,
FeatureSDWAScalar = 80,
FeatureSDWASdst = 81,
FeatureSGPRInitBug = 82,
FeatureSMEMtoVectorWriteHazard = 83,
FeatureSMemRealTime = 84,
FeatureSRAMECC = 85,
FeatureScalarAtomics = 86,
FeatureScalarFlatScratchInsts = 87,
FeatureScalarStores = 88,
FeatureSeaIslands = 89,
FeatureSouthernIslands = 90,
FeatureTrapHandler = 91,
FeatureTrigReducedRange = 92,
FeatureUnalignedBufferAccess = 93,
FeatureUnalignedScratchAccess = 94,
FeatureUnpackedD16VMem = 95,
FeatureVGPRIndexMode = 96,
FeatureVMEMtoScalarWriteHazard = 97,
FeatureVOP3Literal = 98,
FeatureVOP3P = 99,
FeatureVcmpxExecWARHazard = 100,
FeatureVcmpxPermlaneHazard = 101,
FeatureVolcanicIslands = 102,
FeatureVscnt = 103,
FeatureWavefrontSize16 = 104,
FeatureWavefrontSize32 = 105,
FeatureWavefrontSize64 = 106,
FeatureXNACK = 107,
HalfRate64Ops = 108,
NumSubtargetFeatures = 109
};
} // end namespace AMDGPU
} // end namespace llvm
#endif // GET_SUBTARGETINFO_ENUM
#ifdef GET_SUBTARGETINFO_MC_DESC
#undef GET_SUBTARGETINFO_MC_DESC
namespace llvm {
// Sorted (by key) array of values for CPU features.
extern const llvm::SubtargetFeatureKV AMDGPUFeatureKV[] = {
{ "16-bit-insts", "Has i16/f16 instructions", AMDGPU::Feature16BitInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "DumpCode", "Dump MachineInstrs in the CodeEmitter", AMDGPU::FeatureDumpCode, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "add-no-carry-insts", "Have VALU add/sub instructions without carry out", AMDGPU::FeatureAddNoCarryInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "aperture-regs", "Has Memory Aperture Base and Size Registers", AMDGPU::FeatureApertureRegs, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "atomic-fadd-insts", "Has buffer_atomic_add_f32, buffer_atomic_pk_add_f16, global_atomic_add_f32, global_atomic_pk_add_f16 instructions", AMDGPU::FeatureAtomicFaddInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "auto-waitcnt-before-barrier", "Hardware automatically inserts waitcnt before barrier", AMDGPU::FeatureAutoWaitcntBeforeBarrier, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "ci-insts", "Additional instructions for CI+", AMDGPU::FeatureCIInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "code-object-v3", "Generate code object version 3", AMDGPU::FeatureCodeObjectV3, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "cumode", "Enable CU wavefront execution mode", AMDGPU::FeatureCuMode, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dl-insts", "Has v_fmac_f32 and v_xnor_b32 instructions", AMDGPU::FeatureDLInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot1-insts", "Has v_dot4_i32_i8 and v_dot8_i32_i4 instructions", AMDGPU::FeatureDot1Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot2-insts", "Has v_dot2_f32_f16, v_dot2_i32_i16, v_dot2_u32_u16, v_dot4_u32_u8, v_dot8_u32_u4 instructions", AMDGPU::FeatureDot2Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot3-insts", "Has v_dot8c_i32_i4 instruction", AMDGPU::FeatureDot3Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot4-insts", "Has v_dot2c_i32_i16 instruction", AMDGPU::FeatureDot4Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot5-insts", "Has v_dot2c_f32_f16 instruction", AMDGPU::FeatureDot5Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dot6-insts", "Has v_dot4c_i32_i8 instruction", AMDGPU::FeatureDot6Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dpp", "Support DPP (Data Parallel Primitives) extension", AMDGPU::FeatureDPP, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dpp8", "Support DPP8 (Data Parallel Primitives) extension", AMDGPU::FeatureDPP8, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "dumpcode", "Dump MachineInstrs in the CodeEmitter", AMDGPU::FeatureDumpCodeLower, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "enable-ds128", "Use ds_{read|write}_b128", AMDGPU::FeatureEnableDS128, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "enable-prt-strict-null", "Enable zeroing of result registers for sparse texture fetches", AMDGPU::FeatureEnablePRTStrictNull, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fast-fmaf", "Assuming f32 fma is at least as fast as mul + add", AMDGPU::FeatureFastFMAF32, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-address-space", "Support flat address space", AMDGPU::FeatureFlatAddressSpace, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-for-global", "Force to generate flat instruction for global", AMDGPU::FeatureFlatForGlobal, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-global-insts", "Have global_* flat memory instructions", AMDGPU::FeatureFlatGlobalInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-inst-offsets", "Flat instructions have immediate offset addressing mode", AMDGPU::FeatureFlatInstOffsets, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-scratch-insts", "Have scratch_* flat memory instructions", AMDGPU::FeatureFlatScratchInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "flat-segment-offset-bug", "GFX10 bug, inst_offset ignored in flat segment", AMDGPU::FeatureFlatSegmentOffsetBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fma-mix-insts", "Has v_fma_mix_f32, v_fma_mixlo_f16, v_fma_mixhi_f16 instructions", AMDGPU::FeatureFmaMixInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fmaf", "Enable single precision FMA (not as fast as mul+add, but fused)", AMDGPU::FeatureFMA, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp-exceptions", "Enable floating point exceptions", AMDGPU::FeatureFPExceptions, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp16-denormals", "Enable half precision denormal handling", AMDGPU::FeatureFP16Denormals, { { { 0x100000000ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp32-denormals", "Enable single precision denormal handling", AMDGPU::FeatureFP32Denormals, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp64", "Enable double precision operations", AMDGPU::FeatureFP64, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp64-denormals", "Enable double and half precision denormal handling", AMDGPU::FeatureFP64Denormals, { { { 0x140000000ULL, 0x0ULL, 0x0ULL, } } } },
{ "fp64-fp16-denormals", "Enable double and half precision denormal handling", AMDGPU::FeatureFP64FP16Denormals, { { { 0x40000000ULL, 0x0ULL, 0x0ULL, } } } },
{ "gcn3-encoding", "Encoding format for VI", AMDGPU::FeatureGCN3Encoding, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "gfx10", "GFX10 GPU generation", AMDGPU::FeatureGFX10, { { { 0x240d52ec40001627ULL, 0x8c00135964ULL, 0x0ULL, } } } },
{ "gfx10-insts", "Additional instructions for GFX10+", AMDGPU::FeatureGFX10Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "gfx7-gfx8-gfx9-insts", "Instructions shared in GFX7, GFX8, GFX9", AMDGPU::FeatureGFX7GFX8GFX9Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "gfx8-insts", "Additional instructions for GFX8+", AMDGPU::FeatureGFX8Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "gfx9", "GFX9 GPU generation", AMDGPU::FeatureGFX9, { { { 0x40c5cec40000227ULL, 0x40901d35400ULL, 0x0ULL, } } } },
{ "gfx9-insts", "Additional instructions for GFX9+", AMDGPU::FeatureGFX9Insts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "half-rate-64-ops", "Most fp64 instructions are half rate instead of quarter", AMDGPU::HalfRate64Ops, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "inst-fwd-prefetch-bug", "S_INST_PREFETCH instruction causes shader to hang", AMDGPU::FeatureInstFwdPrefetchBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "int-clamp-insts", "Support clamp for integer destination", AMDGPU::FeatureIntClamp, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "inv-2pi-inline-imm", "Has 1 / (2 * pi) as inline immediate", AMDGPU::FeatureInv2PiInlineImm, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "lds-branch-vmem-war-hazard", "Switching between LDS and VMEM-tex not waiting VM_VSRC=0", AMDGPU::FeatureLdsBranchVmemWARHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "lds-misaligned-bug", "Some GFX10 bug with misaligned multi-dword LDS access in WGP mode", AMDGPU::FeatureLdsMisalignedBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "ldsbankcount16", "The number of LDS banks per compute unit.", AMDGPU::FeatureLDSBankCount16, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "ldsbankcount32", "The number of LDS banks per compute unit.", AMDGPU::FeatureLDSBankCount32, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "load-store-opt", "Enable SI load/store optimizer pass", AMDGPU::FeatureEnableLoadStoreOpt, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "localmemorysize0", "The size of local memory in bytes", AMDGPU::FeatureLocalMemorySize0, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "localmemorysize32768", "The size of local memory in bytes", AMDGPU::FeatureLocalMemorySize32768, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "localmemorysize65536", "The size of local memory in bytes", AMDGPU::FeatureLocalMemorySize65536, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "mad-mix-insts", "Has v_mad_mix_f32, v_mad_mixlo_f16, v_mad_mixhi_f16 instructions", AMDGPU::FeatureMadMixInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "mai-insts", "Has mAI instructions", AMDGPU::FeatureMAIInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "max-private-element-size-16", "Maximum private access size may be 16", AMDGPU::FeatureMaxPrivateElementSize16, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "max-private-element-size-4", "Maximum private access size may be 4", AMDGPU::FeatureMaxPrivateElementSize4, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "max-private-element-size-8", "Maximum private access size may be 8", AMDGPU::FeatureMaxPrivateElementSize8, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "mfma-inline-literal-bug", "MFMA cannot use inline literal as SrcC", AMDGPU::FeatureMFMAInlineLiteralBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "mimg-r128", "Support 128-bit texture resources", AMDGPU::FeatureMIMG_R128, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "movrel", "Has v_movrel*_b32 instructions", AMDGPU::FeatureMovrel, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "no-data-dep-hazard", "Does not need SW waitstates", AMDGPU::FeatureNoDataDepHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "no-sdst-cmpx", "V_CMPX does not write VCC/SGPR in addition to EXEC", AMDGPU::FeatureNoSdstCMPX, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "no-sram-ecc-support", "Hardware does not support SRAM ECC", AMDGPU::FeatureDoesNotSupportSRAMECC, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "no-xnack-support", "Hardware does not support XNACK", AMDGPU::FeatureDoesNotSupportXNACK, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "nsa-encoding", "Support NSA encoding for image instructions", AMDGPU::FeatureNSAEncoding, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "nsa-to-vmem-bug", "MIMG-NSA followed by VMEM fail if EXEC_LO or EXEC_HI equals zero", AMDGPU::FeatureNSAtoVMEMBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "offset-3f-bug", "Branch offset of 3f hardware bug", AMDGPU::FeatureOffset3fBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "pk-fmac-f16-inst", "Has v_pk_fmac_f16 instruction", AMDGPU::FeaturePkFmacF16Inst, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "promote-alloca", "Enable promote alloca pass", AMDGPU::FeaturePromoteAlloca, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "r128-a16", "Support 16 bit coordindates/gradients/lod/clamp/mip types on gfx9", AMDGPU::FeatureR128A16, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "register-banking", "Has register banking", AMDGPU::FeatureRegisterBanking, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "s-memrealtime", "Has s_memrealtime instruction", AMDGPU::FeatureSMemRealTime, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "scalar-atomics", "Has atomic scalar memory instructions", AMDGPU::FeatureScalarAtomics, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "scalar-flat-scratch-insts", "Have s_scratch_* flat memory instructions", AMDGPU::FeatureScalarFlatScratchInsts, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "scalar-stores", "Has store scalar memory instructions", AMDGPU::FeatureScalarStores, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa", "Support SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWA, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa-mav", "Support v_mac_f32/f16 with SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWAMac, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa-omod", "Support OMod with SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWAOmod, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa-out-mods-vopc", "Support clamp for VOPC with SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWAOutModsVOPC, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa-scalar", "Support scalar register with SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWAScalar, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sdwa-sdst", "Support scalar dst for VOPC with SDWA (Sub-DWORD Addressing) extension", AMDGPU::FeatureSDWASdst, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "sea-islands", "SEA_ISLANDS GPU generation", AMDGPU::FeatureSeaIslands, { { { 0x2400080840001020ULL, 0x40010000004ULL, 0x0ULL, } } } },
{ "sgpr-init-bug", "VI SGPR initialization bug requiring a fixed SGPR allocation size", AMDGPU::FeatureSGPRInitBug, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "si-scheduler", "Enable SI Machine Scheduler", AMDGPU::FeatureEnableSIScheduler, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "smem-to-vector-write-hazard", "s_load_dword followed by v_cmp page faults", AMDGPU::FeatureSMEMtoVectorWriteHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "southern-islands", "SOUTHERN_ISLANDS GPU generation", AMDGPU::FeatureSouthernIslands, { { { 0x2220000040003000ULL, 0x40010000004ULL, 0x0ULL, } } } },
{ "sram-ecc", "Enable SRAM ECC", AMDGPU::FeatureSRAMECC, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "trap-handler", "Trap handler support", AMDGPU::FeatureTrapHandler, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "trig-reduced-range", "Requires use of fract on arguments to trig instructions", AMDGPU::FeatureTrigReducedRange, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "unaligned-buffer-access", "Support unaligned global loads and stores", AMDGPU::FeatureUnalignedBufferAccess, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "unaligned-scratch-access", "Support unaligned scratch loads and stores", AMDGPU::FeatureUnalignedScratchAccess, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "unpacked-d16-vmem", "Has unpacked d16 vmem instructions", AMDGPU::FeatureUnpackedD16VMem, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "unsafe-ds-offset-folding", "Force using DS instruction immediate offsets on SI", AMDGPU::FeatureEnableUnsafeDSOffsetFolding, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vcmpx-exec-war-hazard", "V_CMPX WAR hazard on EXEC (V_CMPX issue ONLY)", AMDGPU::FeatureVcmpxExecWARHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vcmpx-permlane-hazard", "TODO: describe me", AMDGPU::FeatureVcmpxPermlaneHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vgpr-index-mode", "Has VGPR mode register indexing", AMDGPU::FeatureVGPRIndexMode, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vmem-to-scalar-write-hazard", "VMEM instruction followed by scalar writing to EXEC mask, M0 or SGPR leads to incorrect execution.", AMDGPU::FeatureVMEMtoScalarWriteHazard, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "volcanic-islands", "VOLCANIC_ISLANDS GPU generation", AMDGPU::FeatureVolcanicIslands, { { { 0x240c1c0840001221ULL, 0x4011110b004ULL, 0x0ULL, } } } },
{ "vop3-literal", "Can use one literal in VOP3", AMDGPU::FeatureVOP3Literal, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vop3p", "Has VOP3P packed instructions", AMDGPU::FeatureVOP3P, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "vscnt", "Has separate store vscnt counter", AMDGPU::FeatureVscnt, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "wavefrontsize16", "The number of threads per wavefront", AMDGPU::FeatureWavefrontSize16, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "wavefrontsize32", "The number of threads per wavefront", AMDGPU::FeatureWavefrontSize32, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "wavefrontsize64", "The number of threads per wavefront", AMDGPU::FeatureWavefrontSize64, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
{ "xnack", "Enable XNACK support", AMDGPU::FeatureXNACK, { { { 0x0ULL, 0x0ULL, 0x0ULL, } } } },
};
#ifdef DBGFIELD
#error "<target>GenSubtargetInfo.inc requires a DBGFIELD macro"
#endif
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
#define DBGFIELD(x) x,
#else
#define DBGFIELD(x)
#endif
// ===============================================================
// Data tables for the new per-operand machine model.
// {ProcResourceIdx, Cycles}
extern const llvm::MCWriteProcResEntry AMDGPUWriteProcResTable[] = {
{ 0, 0}, // Invalid
{ 4, 1}, // #1
{ 5, 1}, // #2
{ 6, 1}, // #3
{ 3, 1}, // #4
{ 2, 1}, // #5
{ 1, 1}, // #6
{ 5, 2}, // #7
{ 4, 1}, // #8
{ 6, 1}, // #9
{ 4, 1}, // #10
{ 7, 1}, // #11
{ 3, 1}, // #12
{ 4, 1}, // #13
{ 2, 1}, // #14
{ 4, 1}, // #15
{ 4, 2}, // #16
{ 5, 1}, // #17
{ 6, 1}, // #18
{ 4, 2}, // #19
{ 6, 2} // #20
}; // AMDGPUWriteProcResTable
// {Cycles, WriteResourceID}
extern const llvm::MCWriteLatencyEntry AMDGPUWriteLatencyTable[] = {
{ 0, 0}, // Invalid
{ 1, 0}, // #1 WriteSALU_Write32Bit_WriteFloatFMA
{80, 0}, // #2 WriteVMEM
{ 5, 0}, // #3 WriteLDS_WriteSMEM_WriteSALU_Write32Bit_WriteFloatFMA
{ 4, 0}, // #4 WriteExport_WriteQuarterRate32_WriteDoubleCvt
{ 8, 0}, // #5 WriteBranch_WriteDoubleAdd_Write8PassMAI
{ 1, 0}, // #6 Write32Bit_WriteFloatFMA
{ 1, 0}, // #7 WriteSALU
{ 2, 0}, // #8 Write64Bit_Write2PassMAI_WriteDoubleAdd
{16, 0}, // #9 WriteFloatFMA_WriteDouble_Write16PassMAI_WriteExport
{16, 0}, // #10 WriteFloatFMA_WriteDouble
{ 1, 0}, // #11 WriteSALU
{ 4, 0}, // #12 WriteQuarterRate32_WriteDouble
{ 1, 0}, // #13 WriteSALU
{ 2, 0}, // #14 Write64Bit
{ 2, 0}, // #15 Write64Bit
{500, 0}, // #16 WriteBarrier
{320, 0}, // #17 WriteVMEM
{20, 0}, // #18 WriteLDS_WriteSMEM
{32, 0}, // #19 WriteBranch
{ 5, 0}, // #20 Write32Bit_WriteFloatFMA
{ 5, 0}, // #21 WriteSALU
{17, 0}, // #22 WriteDoubleAdd_WriteQuarterRate32_WriteDoubleCvt
{ 9, 0}, // #23 Write64Bit
{17, 0}, // #24 WriteDouble_WriteQuarterRate32
{ 5, 0}, // #25 WriteSALU
{ 9, 0}, // #26 Write64Bit
{ 9, 0}, // #27 Write64Bit
{2000, 0} // #28 WriteBarrier
}; // AMDGPUWriteLatencyTable
// {UseIdx, WriteResourceID, Cycles}
extern const llvm::MCReadAdvanceEntry AMDGPUReadAdvanceTable[] = {
{0, 0, 0}, // Invalid
{0, 0, -4}, // #1
{0, 0, -2} // #2
}; // AMDGPUReadAdvanceTable
// {Name, NumMicroOps, BeginGroup, EndGroup, WriteProcResIdx,#, WriteLatencyIdx,#, ReadAdvanceIdx,#}
static const llvm::MCSchedClassDesc SIQuarterSpeedModelSchedClasses[] = {
{DBGFIELD("InvalidSchedClass") 16383, false, false, 0, 0, 0, 0, 0, 0},
{DBGFIELD("NullALU_WriteSALU") 1, false, false, 1, 1, 1, 1, 0, 0}, // #1
{DBGFIELD("NullALU_Write32Bit") 1, false, false, 2, 1, 1, 1, 0, 0}, // #2
{DBGFIELD("NullALU_WriteVMEM") 1, false, false, 3, 1, 2, 1, 0, 0}, // #3
{DBGFIELD("NullALU_WriteLDS") 1, false, false, 4, 1, 3, 1, 0, 0}, // #4
{DBGFIELD("NullALU_WriteExport") 1, false, false, 5, 1, 4, 1, 0, 0}, // #5
{DBGFIELD("NullALU_WriteBranch") 1, false, false, 6, 1, 5, 1, 0, 0}, // #6
{DBGFIELD("NullALU") 16383, false, false, 0, 0, 0, 0, 0, 0}, // #7
{DBGFIELD("NullALU_WriteSMEM") 1, false, false, 4, 1, 3, 1, 0, 0}, // #8
{DBGFIELD("NullALU_Write32Bit_WriteSALU") 2, false, false, 1, 2, 6, 2, 0, 0}, // #9
{DBGFIELD("NullALU_WriteDoubleAdd") 1, false, false, 2, 1, 5, 1, 0, 0}, // #10
{DBGFIELD("NullALU_Write64Bit") 1, false, false, 2, 1, 8, 1, 0, 0}, // #11
{DBGFIELD("NullALU_WriteQuarterRate32") 1, false, false, 2, 1, 4, 1, 0, 0}, // #12
{DBGFIELD("NullALU_WriteDoubleCvt") 1, false, false, 2, 1, 4, 1, 0, 0}, // #13
{DBGFIELD("NullALU_WriteFloatFMA") 1, false, false, 2, 1, 9, 1, 0, 0}, // #14
{DBGFIELD("NullALU_WriteDouble") 1, false, false, 2, 1, 9, 1, 0, 0}, // #15
{DBGFIELD("NullALU_WriteFloatFMA_WriteSALU") 2, false, false, 1, 2, 10, 2, 0, 0}, // #16
{DBGFIELD("NullALU_WriteDouble_WriteSALU") 2, false, false, 1, 2, 10, 2, 0, 0}, // #17
{DBGFIELD("NullALU_WriteQuarterRate32_WriteSALU") 2, false, false, 1, 2, 12, 2, 0, 0}, // #18
{DBGFIELD("NullALU_Write64Bit_Write64Bit") 2, false, false, 7, 1, 14, 2, 0, 0}, // #19
{DBGFIELD("NullALU_WriteBarrier") 1, false, false, 6, 1, 16, 1, 0, 0}, // #20
{DBGFIELD("V_ACCVGPR_WRITE_B32") 16382, false, false, 0, 0, 0, 0, 0, 0}, // #21
{DBGFIELD("V_MFMA_F32_4X4X1F32_V_MFMA_F32_4X4X2BF16_V_MFMA_F32_4X4X4F16_V_MFMA_I32_4X4X4I8_V_MFMA_F32_4X4X1F32_vi_V_MFMA_F32_4X4X2BF16_vi_V_MFMA_F32_4X4X4F16_vi_V_MFMA_I32_4X4X4I8_vi") 1, false, false, 2, 1, 8, 1, 1, 1}, // #22
{DBGFIELD("V_MFMA_F32_16X16X16F16_V_MFMA_F32_16X16X1F32_V_MFMA_F32_16X16X2BF16_V_MFMA_F32_16X16X4F16_V_MFMA_F32_16X16X4F32_V_MFMA_F32_16X16X8BF16_V_MFMA_I32_16X16X16I8_V_MFMA_I32_16X16X4I8_V_MFMA_F32_16X16X16F16_vi_V_MFMA_F32_16X16X1F32_vi_V_MFMA_F32_16X16X2BF16_vi_V_MFMA_F32_16X16X4F16_vi_V_MFMA_F32_16X16X4F32_vi_V_MFMA_F32_16X16X8BF16_vi_V_MFMA_I32_16X16X16I8_vi_V_MFMA_I32_16X16X4I8_vi") 1, false, false, 2, 1, 5, 1, 1, 1}, // #23
{DBGFIELD("V_MFMA_F32_32X32X1F32_V_MFMA_F32_32X32X2BF16_V_MFMA_F32_32X32X2F32_V_MFMA_F32_32X32X4BF16_V_MFMA_F32_32X32X4F16_V_MFMA_F32_32X32X8F16_V_MFMA_I32_32X32X4I8_V_MFMA_I32_32X32X8I8_V_MFMA_F32_32X32X1F32_vi_V_MFMA_F32_32X32X2BF16_vi_V_MFMA_F32_32X32X2F32_vi_V_MFMA_F32_32X32X4BF16_vi_V_MFMA_F32_32X32X4F16_vi_V_MFMA_F32_32X32X8F16_vi_V_MFMA_I32_32X32X4I8_vi_V_MFMA_I32_32X32X8I8_vi") 1, false, false, 2, 1, 9, 1, 1, 1}, // #24
{DBGFIELD("COPY") 16382, false, false, 0, 0, 0, 0, 0, 0}, // #25
{DBGFIELD("Write64Bit_MIVGPRRead") 1, false, false, 2, 1, 8, 1, 2, 1}, // #26
{DBGFIELD("Write64Bit_ReadDefault") 1, false, false, 2, 1, 8, 1, 0, 0}, // #27
{DBGFIELD("Write32Bit") 1, false, false, 2, 1, 1, 1, 0, 0}, // #28
{DBGFIELD("Write64Bit") 1, false, false, 2, 1, 8, 1, 0, 0}, // #29
{DBGFIELD("WriteSALU") 1, false, false, 1, 1, 1, 1, 0, 0}, // #30
}; // SIQuarterSpeedModelSchedClasses
// {Name, NumMicroOps, BeginGroup, EndGroup, WriteProcResIdx,#, WriteLatencyIdx,#, ReadAdvanceIdx,#}
static const llvm::MCSchedClassDesc GFX10SpeedModelSchedClasses[] = {
{DBGFIELD("InvalidSchedClass") 16383, false, false, 0, 0, 0, 0, 0, 0},
{DBGFIELD("NullALU_WriteSALU") 1, false, false, 1, 2, 3, 1, 0, 0}, // #1
{DBGFIELD("NullALU_Write32Bit") 1, false, false, 8, 2, 3, 1, 0, 0}, // #2
{DBGFIELD("NullALU_WriteVMEM") 1, false, false, 10, 2, 17, 1, 0, 0}, // #3
{DBGFIELD("NullALU_WriteLDS") 1, false, false, 12, 2, 18, 1, 0, 0}, // #4
{DBGFIELD("NullALU_WriteExport") 1, false, false, 14, 2, 9, 1, 0, 0}, // #5
{DBGFIELD("NullALU_WriteBranch") 1, false, false, 6, 1, 19, 1, 0, 0}, // #6
{DBGFIELD("NullALU") 16383, false, false, 0, 0, 0, 0, 0, 0}, // #7
{DBGFIELD("NullALU_WriteSMEM") 1, false, false, 12, 2, 18, 1, 0, 0}, // #8
{DBGFIELD("NullALU_Write32Bit_WriteSALU") 2, false, false, 16, 3, 20, 2, 0, 0}, // #9
{DBGFIELD("NullALU_WriteDoubleAdd") 1, false, false, 8, 2, 22, 1, 0, 0}, // #10
{DBGFIELD("NullALU_Write64Bit") 1, false, false, 8, 2, 23, 1, 0, 0}, // #11
{DBGFIELD("NullALU_WriteQuarterRate32") 1, false, false, 8, 2, 22, 1, 0, 0}, // #12
{DBGFIELD("NullALU_WriteDoubleCvt") 1, false, false, 8, 2, 22, 1, 0, 0}, // #13
{DBGFIELD("NullALU_WriteFloatFMA") 1, false, false, 8, 2, 3, 1, 0, 0}, // #14
{DBGFIELD("NullALU_WriteDouble") 1, false, false, 8, 2, 22, 1, 0, 0}, // #15
{DBGFIELD("NullALU_WriteFloatFMA_WriteSALU") 2, false, false, 16, 3, 20, 2, 0, 0}, // #16
{DBGFIELD("NullALU_WriteDouble_WriteSALU") 2, false, false, 16, 3, 24, 2, 0, 0}, // #17
{DBGFIELD("NullALU_WriteQuarterRate32_WriteSALU") 2, false, false, 16, 3, 24, 2, 0, 0}, // #18
{DBGFIELD("NullALU_Write64Bit_Write64Bit") 2, false, false, 19, 2, 26, 2, 0, 0}, // #19
{DBGFIELD("NullALU_WriteBarrier") 1, false, false, 6, 1, 28, 1, 0, 0}, // #20
{DBGFIELD("V_ACCVGPR_WRITE_B32") 1, false, false, 8, 2, 3, 1, 0, 0}, // #21
{DBGFIELD("V_MFMA_F32_4X4X1F32_V_MFMA_F32_4X4X2BF16_V_MFMA_F32_4X4X4F16_V_MFMA_I32_4X4X4I8_V_MFMA_F32_4X4X1F32_vi_V_MFMA_F32_4X4X2BF16_vi_V_MFMA_F32_4X4X4F16_vi_V_MFMA_I32_4X4X4I8_vi") 1, false, false, 8, 2, 3, 1, 0, 0}, // #22
{DBGFIELD("V_MFMA_F32_16X16X16F16_V_MFMA_F32_16X16X1F32_V_MFMA_F32_16X16X2BF16_V_MFMA_F32_16X16X4F16_V_MFMA_F32_16X16X4F32_V_MFMA_F32_16X16X8BF16_V_MFMA_I32_16X16X16I8_V_MFMA_I32_16X16X4I8_V_MFMA_F32_16X16X16F16_vi_V_MFMA_F32_16X16X1F32_vi_V_MFMA_F32_16X16X2BF16_vi_V_MFMA_F32_16X16X4F16_vi_V_MFMA_F32_16X16X4F32_vi_V_MFMA_F32_16X16X8BF16_vi_V_MFMA_I32_16X16X16I8_vi_V_MFMA_I32_16X16X4I8_vi") 1, false, false, 8, 2, 3, 1, 0, 0}, // #23
{DBGFIELD("V_MFMA_F32_32X32X1F32_V_MFMA_F32_32X32X2BF16_V_MFMA_F32_32X32X2F32_V_MFMA_F32_32X32X4BF16_V_MFMA_F32_32X32X4F16_V_MFMA_F32_32X32X8F16_V_MFMA_I32_32X32X4I8_V_MFMA_I32_32X32X8I8_V_MFMA_F32_32X32X1F32_vi_V_MFMA_F32_32X32X2BF16_vi_V_MFMA_F32_32X32X2F32_vi_V_MFMA_F32_32X32X4BF16_vi_V_MFMA_F32_32X32X4F16_vi_V_MFMA_F32_32X32X8F16_vi_V_MFMA_I32_32X32X4I8_vi_V_MFMA_I32_32X32X8I8_vi") 1, false, false, 8, 2, 3, 1, 0, 0}, // #24
{DBGFIELD("COPY") 16382, false, false, 0, 0, 0, 0, 0, 0}, // #25
{DBGFIELD("Write64Bit_MIVGPRRead") 0, false, false, 0, 0, 0, 0, 0, 0}, // #26
{DBGFIELD("Write64Bit_ReadDefault") 0, false, false, 0, 0, 0, 0, 0, 0}, // #27
{DBGFIELD("Write32Bit") 1, false, false, 8, 2, 3, 1, 0, 0}, // #28
{DBGFIELD("Write64Bit") 1, false, false, 8, 2, 23, 1, 0, 0}, // #29
{DBGFIELD("WriteSALU") 1, false, false, 1, 2, 3, 1, 0, 0}, // #30
}; // GFX10SpeedModelSchedClasses
// {Name, NumMicroOps, BeginGroup, EndGroup, WriteProcResIdx,#, WriteLatencyIdx,#, ReadAdvanceIdx,#}
static const llvm::MCSchedClassDesc SIFullSpeedModelSchedClasses[] = {
{DBGFIELD("InvalidSchedClass") 16383, false, false, 0, 0, 0, 0, 0, 0},
{DBGFIELD("NullALU_WriteSALU") 1, false, false, 1, 1, 1, 1, 0, 0}, // #1
{DBGFIELD("NullALU_Write32Bit") 1, false, false, 2, 1, 1, 1, 0, 0}, // #2
{DBGFIELD("NullALU_WriteVMEM") 1, false, false, 3, 1, 2, 1, 0, 0}, // #3
{DBGFIELD("NullALU_WriteLDS") 1, false, false, 4, 1, 3, 1, 0, 0}, // #4
{DBGFIELD("NullALU_WriteExport") 1, false, false, 5, 1, 4, 1, 0, 0}, // #5
{DBGFIELD("NullALU_WriteBranch") 1, false, false, 6, 1, 5, 1, 0, 0}, // #6
{DBGFIELD("NullALU") 16383, false, false, 0, 0, 0, 0, 0, 0}, // #7
{DBGFIELD("NullALU_WriteSMEM") 1, false, false, 4, 1, 3, 1, 0, 0}, // #8
{DBGFIELD("NullALU_Write32Bit_WriteSALU") 2, false, false, 1, 2, 6, 2, 0, 0}, // #9
{DBGFIELD("NullALU_WriteDoubleAdd") 1, false, false, 2, 1, 8, 1, 0, 0}, // #10
{DBGFIELD("NullALU_Write64Bit") 1, false, false, 2, 1, 8, 1, 0, 0}, // #11
{DBGFIELD("NullALU_WriteQuarterRate32") 1, false, false, 2, 1, 4, 1, 0, 0}, // #12
{DBGFIELD("NullALU_WriteDoubleCvt") 1, false, false, 2, 1, 4, 1, 0, 0}, // #13
{DBGFIELD("NullALU_WriteFloatFMA") 1, false, false, 2, 1, 1, 1, 0, 0}, // #14
{DBGFIELD("NullALU_WriteDouble") 1, false, false, 2, 1, 4, 1, 0, 0}, // #15
{DBGFIELD("NullALU_WriteFloatFMA_WriteSALU") 2, false, false, 1, 2, 6, 2, 0, 0}, // #16
{DBGFIELD("NullALU_WriteDouble_WriteSALU") 2, false, false, 1, 2, 12, 2, 0, 0}, // #17
{DBGFIELD("NullALU_WriteQuarterRate32_WriteSALU") 2, false, false, 1, 2, 12, 2, 0, 0}, // #18
{DBGFIELD("NullALU_Write64Bit_Write64Bit") 2, false, false, 7, 1, 14, 2, 0, 0}, // #19
{DBGFIELD("NullALU_WriteBarrier") 1, false, false, 6, 1, 16, 1, 0, 0}, // #20
{DBGFIELD("V_ACCVGPR_WRITE_B32") 16382, false, false, 0, 0, 0, 0, 0, 0}, // #21
{DBGFIELD("V_MFMA_F32_4X4X1F32_V_MFMA_F32_4X4X2BF16_V_MFMA_F32_4X4X4F16_V_MFMA_I32_4X4X4I8_V_MFMA_F32_4X4X1F32_vi_V_MFMA_F32_4X4X2BF16_vi_V_MFMA_F32_4X4X4F16_vi_V_MFMA_I32_4X4X4I8_vi") 1, false, false, 2, 1, 8, 1, 1, 1}, // #22
{DBGFIELD("V_MFMA_F32_16X16X16F16_V_MFMA_F32_16X16X1F32_V_MFMA_F32_16X16X2BF16_V_MFMA_F32_16X16X4F16_V_MFMA_F32_16X16X4F32_V_MFMA_F32_16X16X8BF16_V_MFMA_I32_16X16X16I8_V_MFMA_I32_16X16X4I8_V_MFMA_F32_16X16X16F16_vi_V_MFMA_F32_16X16X1F32_vi_V_MFMA_F32_16X16X2BF16_vi_V_MFMA_F32_16X16X4F16_vi_V_MFMA_F32_16X16X4F32_vi_V_MFMA_F32_16X16X8BF16_vi_V_MFMA_I32_16X16X16I8_vi_V_MFMA_I32_16X16X4I8_vi") 1, false, false, 2, 1, 5, 1, 1, 1}, // #23
{DBGFIELD("V_MFMA_F32_32X32X1F32_V_MFMA_F32_32X32X2BF16_V_MFMA_F32_32X32X2F32_V_MFMA_F32_32X32X4BF16_V_MFMA_F32_32X32X4F16_V_MFMA_F32_32X32X8F16_V_MFMA_I32_32X32X4I8_V_MFMA_I32_32X32X8I8_V_MFMA_F32_32X32X1F32_vi_V_MFMA_F32_32X32X2BF16_vi_V_MFMA_F32_32X32X2F32_vi_V_MFMA_F32_32X32X4BF16_vi_V_MFMA_F32_32X32X4F16_vi_V_MFMA_F32_32X32X8F16_vi_V_MFMA_I32_32X32X4I8_vi_V_MFMA_I32_32X32X8I8_vi") 1, false, false, 2, 1, 9, 1, 1, 1}, // #24
{DBGFIELD("COPY") 16382, false, false, 0, 0, 0, 0, 0, 0}, // #25
{DBGFIELD("Write64Bit_MIVGPRRead") 1, false, false, 2, 1, 8, 1, 2, 1}, // #26
{DBGFIELD("Write64Bit_ReadDefault") 1, false, false, 2, 1, 8, 1, 0, 0}, // #27
{DBGFIELD("Write32Bit") 1, false, false, 2, 1, 1, 1, 0, 0}, // #28
{DBGFIELD("Write64Bit") 1, false, false, 2, 1, 8, 1, 0, 0}, // #29
{DBGFIELD("WriteSALU") 1, false, false, 1, 1, 1, 1, 0, 0}, // #30
}; // SIFullSpeedModelSchedClasses
#undef DBGFIELD
static const llvm::MCSchedModel NoSchedModel = {
MCSchedModel::DefaultIssueWidth,
MCSchedModel::DefaultMicroOpBufferSize,
MCSchedModel::DefaultLoopMicroOpBufferSize,
MCSchedModel::DefaultLoadLatency,
MCSchedModel::DefaultHighLatency,
MCSchedModel::DefaultMispredictPenalty,
false, // PostRAScheduler
false, // CompleteModel
0, // Processor ID
nullptr, nullptr, 0, 0, // No instruction-level machine model.
nullptr, // No Itinerary
nullptr // No extra processor descriptor
};
static const unsigned SIQuarterSpeedModelProcResourceSubUnits[] = {
0, // Invalid
};
// {Name, NumUnits, SuperIdx, BufferSize, SubUnitsIdxBegin}
static const llvm::MCProcResourceDesc SIQuarterSpeedModelProcResources[] = {
{"InvalidUnit", 0, 0, 0, 0},
{"HWBranch", 1, 0, 1, nullptr}, // #1
{"HWExport", 1, 0, 7, nullptr}, // #2
{"HWLGKM", 1, 0, 31, nullptr}, // #3
{"HWSALU", 1, 0, 1, nullptr}, // #4
{"HWVALU", 1, 0, 1, nullptr}, // #5
{"HWVMEM", 1, 0, 15, nullptr}, // #6
};
static const llvm::MCSchedModel SIQuarterSpeedModel = {
1, // IssueWidth
1, // MicroOpBufferSize
MCSchedModel::DefaultLoopMicroOpBufferSize,
MCSchedModel::DefaultLoadLatency,
MCSchedModel::DefaultHighLatency,
20, // MispredictPenalty
true, // PostRAScheduler
false, // CompleteModel
1, // Processor ID
SIQuarterSpeedModelProcResources,
SIQuarterSpeedModelSchedClasses,
7,
31,
nullptr, // No Itinerary
nullptr // No extra processor descriptor
};
static const unsigned GFX10SpeedModelProcResourceSubUnits[] = {
0, // Invalid
};
// {Name, NumUnits, SuperIdx, BufferSize, SubUnitsIdxBegin}
static const llvm::MCProcResourceDesc GFX10SpeedModelProcResources[] = {
{"InvalidUnit", 0, 0, 0, 0},
{"HWBranch", 1, 0, 1, nullptr}, // #1
{"HWExport", 1, 0, 7, nullptr}, // #2
{"HWLGKM", 1, 0, 31, nullptr}, // #3
{"HWRC", 1, 0, 1, nullptr}, // #4
{"HWSALU", 1, 0, 1, nullptr}, // #5
{"HWVALU", 1, 0, 1, nullptr}, // #6
{"HWVMEM", 1, 0, 15, nullptr}, // #7
};
static const llvm::MCSchedModel GFX10SpeedModel = {
1, // IssueWidth
1, // MicroOpBufferSize
MCSchedModel::DefaultLoopMicroOpBufferSize,
MCSchedModel::DefaultLoadLatency,
MCSchedModel::DefaultHighLatency,
20, // MispredictPenalty
true, // PostRAScheduler
false, // CompleteModel
2, // Processor ID
GFX10SpeedModelProcResources,
GFX10SpeedModelSchedClasses,
8,
31,
nullptr, // No Itinerary
nullptr // No extra processor descriptor
};
static const unsigned SIFullSpeedModelProcResourceSubUnits[] = {
0, // Invalid
};
// {Name, NumUnits, SuperIdx, BufferSize, SubUnitsIdxBegin}
static const llvm::MCProcResourceDesc SIFullSpeedModelProcResources[] = {
{"InvalidUnit", 0, 0, 0, 0},
{"HWBranch", 1, 0, 1, nullptr}, // #1
{"HWExport", 1, 0, 7, nullptr}, // #2
{"HWLGKM", 1, 0, 31, nullptr}, // #3
{"HWSALU", 1, 0, 1, nullptr}, // #4
{"HWVALU", 1, 0, 1, nullptr}, // #5
{"HWVMEM", 1, 0, 15, nullptr}, // #6
};
static const llvm::MCSchedModel SIFullSpeedModel = {
1, // IssueWidth
1, // MicroOpBufferSize
MCSchedModel::DefaultLoopMicroOpBufferSize,
MCSchedModel::DefaultLoadLatency,
MCSchedModel::DefaultHighLatency,
20, // MispredictPenalty
true, // PostRAScheduler
false, // CompleteModel
3, // Processor ID
SIFullSpeedModelProcResources,
SIFullSpeedModelSchedClasses,
7,
31,
nullptr, // No Itinerary
nullptr // No extra processor descriptor
};
// Sorted (by key) array of values for CPU subtype.
extern const llvm::SubtargetSubTypeKV AMDGPUSubTypeKV[] = {
{ "bonaire", { { { 0x20000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "carrizo", { { { 0x20000400000040ULL, 0x184080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "fiji", { { { 0x20000000002040ULL, 0x4080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "generic", { { { 0x0ULL, 0x40000000000ULL, 0x0ULL, } } }, &NoSchedModel },
{ "generic-hsa", { { { 0x800000000ULL, 0x40000000000ULL, 0x0ULL, } } }, &NoSchedModel },
{ "gfx1010", { { { 0xe2810000002140ULL, 0x23201c80098ULL, 0x0ULL, } } }, &GFX10SpeedModel },
{ "gfx1011", { { { 0x628100000ce140ULL, 0x23201c80098ULL, 0x0ULL, } } }, &GFX10SpeedModel },
{ "gfx1012", { { { 0xe28100000ce140ULL, 0x23201c80098ULL, 0x0ULL, } } }, &GFX10SpeedModel },
{ "gfx600", { { { 0x20000400002040ULL, 0x100004000000ULL, 0x0ULL, } } }, &SIFullSpeedModel },
{ "gfx601", { { { 0x20000000002040ULL, 0x4000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx700", { { { 0x20000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx701", { { { 0x20000400002040ULL, 0x100002000000ULL, 0x0ULL, } } }, &SIFullSpeedModel },
{ "gfx702", { { { 0x10000400002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx703", { { { 0x10000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx704", { { { 0x20000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx801", { { { 0x20000400000040ULL, 0x184080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx802", { { { 0x20000000002040ULL, 0x4080040000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx803", { { { 0x20000000002040ULL, 0x4080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx810", { { { 0x10000000000040ULL, 0x84000000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx900", { { { 0x4020200000003040ULL, 0x0ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx902", { { { 0x4020200000001040ULL, 0x80000000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx904", { { { 0x20220000003040ULL, 0x0ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx906", { { { 0x2022000000e140ULL, 0x100000000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx908", { { { 0x18202200000fc148ULL, 0x100000200100ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "gfx909", { { { 0x4020200000000040ULL, 0x80000000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "hainan", { { { 0x20000000002040ULL, 0x4000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "hawaii", { { { 0x20000400002040ULL, 0x100002000000ULL, 0x0ULL, } } }, &SIFullSpeedModel },
{ "iceland", { { { 0x20000000002040ULL, 0x4080040000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "kabini", { { { 0x10000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "kaveri", { { { 0x20000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "mullins", { { { 0x10000000002040ULL, 0x2000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "oland", { { { 0x20000000002040ULL, 0x4000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "pitcairn", { { { 0x20000000002040ULL, 0x4000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "polaris10", { { { 0x20000000002040ULL, 0x4080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "polaris11", { { { 0x20000000002040ULL, 0x4080000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "stoney", { { { 0x10000000000040ULL, 0x84000000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "tahiti", { { { 0x20000400002040ULL, 0x100004000000ULL, 0x0ULL, } } }, &SIFullSpeedModel },
{ "tonga", { { { 0x20000000002040ULL, 0x4080040000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
{ "verde", { { { 0x20000000002040ULL, 0x4000000ULL, 0x0ULL, } } }, &SIQuarterSpeedModel },
};
namespace AMDGPU_MC {
unsigned resolveVariantSchedClassImpl(unsigned SchedClass,
const MCInst *MI, unsigned CPUID) {
switch (SchedClass) {
case 21: // V_ACCVGPR_WRITE_B32
if (CPUID == 1) { // SIQuarterSpeedModel
return 27; // Write64Bit_ReadDefault
}
if (CPUID == 3) { // SIFullSpeedModel
return 27; // Write64Bit_ReadDefault
}
break;
case 25: // COPY
if (CPUID == 1) { // SIQuarterSpeedModel
return 30; // WriteSALU
}
if (CPUID == 2) { // GFX10SpeedModel
return 30; // WriteSALU
}
if (CPUID == 3) { // SIFullSpeedModel
return 30; // WriteSALU
}
break;
};
// Don't know how to resolve this scheduling class.
return 0;
}
} // end namespace AMDGPU_MC
struct AMDGPUGenMCSubtargetInfo : public MCSubtargetInfo {
AMDGPUGenMCSubtargetInfo(const Triple &TT,
StringRef CPU, StringRef FS, ArrayRef<SubtargetFeatureKV> PF,
ArrayRef<SubtargetSubTypeKV> PD,
const MCWriteProcResEntry *WPR,
const MCWriteLatencyEntry *WL,
const MCReadAdvanceEntry *RA, const InstrStage *IS,
const unsigned *OC, const unsigned *FP) :
MCSubtargetInfo(TT, CPU, FS, PF, PD,
WPR, WL, RA, IS, OC, FP) { }
unsigned resolveVariantSchedClass(unsigned SchedClass,
const MCInst *MI, unsigned CPUID) const override {
return AMDGPU_MC::resolveVariantSchedClassImpl(SchedClass, MI, CPUID);
}
};
static inline MCSubtargetInfo *createAMDGPUMCSubtargetInfoImpl(const Triple &TT, StringRef CPU, StringRef FS) {
return new AMDGPUGenMCSubtargetInfo(TT, CPU, FS, AMDGPUFeatureKV, AMDGPUSubTypeKV,
AMDGPUWriteProcResTable, AMDGPUWriteLatencyTable, AMDGPUReadAdvanceTable,
nullptr, nullptr, nullptr);
}
} // end namespace llvm
#endif // GET_SUBTARGETINFO_MC_DESC
#ifdef GET_SUBTARGETINFO_TARGET_DESC
#undef GET_SUBTARGETINFO_TARGET_DESC
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
// ParseSubtargetFeatures - Parses features string setting specified
// subtarget options.
void llvm::AMDGPUSubtarget::ParseSubtargetFeatures(StringRef CPU, StringRef FS) {
LLVM_DEBUG(dbgs() << "\nFeatures:" << FS);
LLVM_DEBUG(dbgs() << "\nCPU:" << CPU << "\n\n");
InitMCProcessorInfo(CPU, FS);
const FeatureBitset& Bits = getFeatureBits();
if (Bits[AMDGPU::Feature16BitInsts]) Has16BitInsts = true;
if (Bits[AMDGPU::FeatureAddNoCarryInsts]) AddNoCarryInsts = true;
if (Bits[AMDGPU::FeatureApertureRegs]) HasApertureRegs = true;
if (Bits[AMDGPU::FeatureAtomicFaddInsts]) HasAtomicFaddInsts = true;
if (Bits[AMDGPU::FeatureAutoWaitcntBeforeBarrier]) AutoWaitcntBeforeBarrier = true;
if (Bits[AMDGPU::FeatureCIInsts]) CIInsts = true;
if (Bits[AMDGPU::FeatureCodeObjectV3]) CodeObjectV3 = true;
if (Bits[AMDGPU::FeatureCuMode]) EnableCuMode = true;
if (Bits[AMDGPU::FeatureDLInsts]) HasDLInsts = true;
if (Bits[AMDGPU::FeatureDPP]) HasDPP = true;
if (Bits[AMDGPU::FeatureDPP8]) HasDPP8 = true;
if (Bits[AMDGPU::FeatureDisable]) FeatureDisable = true;
if (Bits[AMDGPU::FeatureDoesNotSupportSRAMECC]) DoesNotSupportSRAMECC = true;
if (Bits[AMDGPU::FeatureDoesNotSupportXNACK]) DoesNotSupportXNACK = true;
if (Bits[AMDGPU::FeatureDot1Insts]) HasDot1Insts = true;
if (Bits[AMDGPU::FeatureDot2Insts]) HasDot2Insts = true;
if (Bits[AMDGPU::FeatureDot3Insts]) HasDot3Insts = true;
if (Bits[AMDGPU::FeatureDot4Insts]) HasDot4Insts = true;
if (Bits[AMDGPU::FeatureDot5Insts]) HasDot5Insts = true;
if (Bits[AMDGPU::FeatureDot6Insts]) HasDot6Insts = true;
if (Bits[AMDGPU::FeatureDumpCode]) DumpCode = true;
if (Bits[AMDGPU::FeatureDumpCodeLower]) DumpCode = true;
if (Bits[AMDGPU::FeatureEnableDS128]) EnableDS128 = true;
if (Bits[AMDGPU::FeatureEnableLoadStoreOpt]) EnableLoadStoreOpt = true;
if (Bits[AMDGPU::FeatureEnablePRTStrictNull]) EnablePRTStrictNull = true;
if (Bits[AMDGPU::FeatureEnableSIScheduler]) EnableSIScheduler = true;
if (Bits[AMDGPU::FeatureEnableUnsafeDSOffsetFolding]) EnableUnsafeDSOffsetFolding = true;
if (Bits[AMDGPU::FeatureFMA]) FMA = true;
if (Bits[AMDGPU::FeatureFP16Denormals]) FP64FP16Denormals = true;
if (Bits[AMDGPU::FeatureFP32Denormals]) FP32Denormals = true;
if (Bits[AMDGPU::FeatureFP64]) FP64 = true;
if (Bits[AMDGPU::FeatureFP64Denormals]) FP64FP16Denormals = true;
if (Bits[AMDGPU::FeatureFP64FP16Denormals]) FP64FP16Denormals = true;
if (Bits[AMDGPU::FeatureFPExceptions]) FPExceptions = true;
if (Bits[AMDGPU::FeatureFastFMAF32]) FastFMAF32 = true;
if (Bits[AMDGPU::FeatureFlatAddressSpace]) FlatAddressSpace = true;
if (Bits[AMDGPU::FeatureFlatForGlobal]) FlatForGlobal = true;
if (Bits[AMDGPU::FeatureFlatGlobalInsts]) FlatGlobalInsts = true;
if (Bits[AMDGPU::FeatureFlatInstOffsets]) FlatInstOffsets = true;
if (Bits[AMDGPU::FeatureFlatScratchInsts]) FlatScratchInsts = true;
if (Bits[AMDGPU::FeatureFlatSegmentOffsetBug]) HasFlatSegmentOffsetBug = true;
if (Bits[AMDGPU::FeatureFmaMixInsts]) HasFmaMixInsts = true;
if (Bits[AMDGPU::FeatureGCN3Encoding]) GCN3Encoding = true;
if (Bits[AMDGPU::FeatureGFX7GFX8GFX9Insts]) GFX7GFX8GFX9Insts = true;
if (Bits[AMDGPU::FeatureGFX8Insts]) GFX8Insts = true;
if (Bits[AMDGPU::FeatureGFX9] && Gen < GCNSubtarget::GFX9) Gen = GCNSubtarget::GFX9;
if (Bits[AMDGPU::FeatureGFX9Insts]) GFX9Insts = true;
if (Bits[AMDGPU::FeatureGFX10] && Gen < GCNSubtarget::GFX10) Gen = GCNSubtarget::GFX10;
if (Bits[AMDGPU::FeatureGFX10Insts]) GFX10Insts = true;
if (Bits[AMDGPU::FeatureInstFwdPrefetchBug]) HasInstFwdPrefetchBug = true;
if (Bits[AMDGPU::FeatureIntClamp]) HasIntClamp = true;
if (Bits[AMDGPU::FeatureInv2PiInlineImm]) HasInv2PiInlineImm = true;
if (Bits[AMDGPU::FeatureLDSBankCount16] && LDSBankCount < 16) LDSBankCount = 16;
if (Bits[AMDGPU::FeatureLDSBankCount32] && LDSBankCount < 32) LDSBankCount = 32;
if (Bits[AMDGPU::FeatureLdsBranchVmemWARHazard]) HasLdsBranchVmemWARHazard = true;
if (Bits[AMDGPU::FeatureLdsMisalignedBug]) LDSMisalignedBug = true;
if (Bits[AMDGPU::FeatureLocalMemorySize0] && LocalMemorySize < 0) LocalMemorySize = 0;
if (Bits[AMDGPU::FeatureLocalMemorySize32768] && LocalMemorySize < 32768) LocalMemorySize = 32768;
if (Bits[AMDGPU::FeatureLocalMemorySize65536] && LocalMemorySize < 65536) LocalMemorySize = 65536;
if (Bits[AMDGPU::FeatureMAIInsts]) HasMAIInsts = true;
if (Bits[AMDGPU::FeatureMFMAInlineLiteralBug]) HasMFMAInlineLiteralBug = true;
if (Bits[AMDGPU::FeatureMIMG_R128]) MIMG_R128 = true;
if (Bits[AMDGPU::FeatureMadMixInsts]) HasMadMixInsts = true;
if (Bits[AMDGPU::FeatureMaxPrivateElementSize4] && MaxPrivateElementSize < 4) MaxPrivateElementSize = 4;
if (Bits[AMDGPU::FeatureMaxPrivateElementSize8] && MaxPrivateElementSize < 8) MaxPrivateElementSize = 8;
if (Bits[AMDGPU::FeatureMaxPrivateElementSize16] && MaxPrivateElementSize < 16) MaxPrivateElementSize = 16;
if (Bits[AMDGPU::FeatureMovrel]) HasMovrel = true;
if (Bits[AMDGPU::FeatureNSAEncoding]) HasNSAEncoding = true;
if (Bits[AMDGPU::FeatureNSAtoVMEMBug]) HasNSAtoVMEMBug = true;
if (Bits[AMDGPU::FeatureNoDataDepHazard]) HasNoDataDepHazard = true;
if (Bits[AMDGPU::FeatureNoSdstCMPX]) HasNoSdstCMPX = true;
if (Bits[AMDGPU::FeatureOffset3fBug]) HasOffset3fBug = true;
if (Bits[AMDGPU::FeaturePkFmacF16Inst]) HasPkFmacF16Inst = true;
if (Bits[AMDGPU::FeaturePromoteAlloca]) EnablePromoteAlloca = true;
if (Bits[AMDGPU::FeatureR128A16]) HasR128A16 = true;
if (Bits[AMDGPU::FeatureRegisterBanking]) HasRegisterBanking = true;
if (Bits[AMDGPU::FeatureSDWA]) HasSDWA = true;
if (Bits[AMDGPU::FeatureSDWAMac]) HasSDWAMac = true;
if (Bits[AMDGPU::FeatureSDWAOmod]) HasSDWAOmod = true;
if (Bits[AMDGPU::FeatureSDWAOutModsVOPC]) HasSDWAOutModsVOPC = true;
if (Bits[AMDGPU::FeatureSDWAScalar]) HasSDWAScalar = true;
if (Bits[AMDGPU::FeatureSDWASdst]) HasSDWASdst = true;
if (Bits[AMDGPU::FeatureSGPRInitBug]) SGPRInitBug = true;
if (Bits[AMDGPU::FeatureSMEMtoVectorWriteHazard]) HasSMEMtoVectorWriteHazard = true;
if (Bits[AMDGPU::FeatureSMemRealTime]) HasSMemRealTime = true;
if (Bits[AMDGPU::FeatureSRAMECC]) EnableSRAMECC = true;
if (Bits[AMDGPU::FeatureScalarAtomics]) HasScalarAtomics = true;
if (Bits[AMDGPU::FeatureScalarFlatScratchInsts]) ScalarFlatScratchInsts = true;
if (Bits[AMDGPU::FeatureScalarStores]) HasScalarStores = true;
if (Bits[AMDGPU::FeatureSeaIslands] && Gen < GCNSubtarget::SEA_ISLANDS) Gen = GCNSubtarget::SEA_ISLANDS;
if (Bits[AMDGPU::FeatureSouthernIslands] && Gen < GCNSubtarget::SOUTHERN_ISLANDS) Gen = GCNSubtarget::SOUTHERN_ISLANDS;
if (Bits[AMDGPU::FeatureTrapHandler]) TrapHandler = true;
if (Bits[AMDGPU::FeatureTrigReducedRange]) HasTrigReducedRange = true;
if (Bits[AMDGPU::FeatureUnalignedBufferAccess]) UnalignedBufferAccess = true;
if (Bits[AMDGPU::FeatureUnalignedScratchAccess]) UnalignedScratchAccess = true;
if (Bits[AMDGPU::FeatureUnpackedD16VMem]) HasUnpackedD16VMem = true;
if (Bits[AMDGPU::FeatureVGPRIndexMode]) HasVGPRIndexMode = true;
if (Bits[AMDGPU::FeatureVMEMtoScalarWriteHazard]) HasVMEMtoScalarWriteHazard = true;
if (Bits[AMDGPU::FeatureVOP3Literal]) HasVOP3Literal = true;
if (Bits[AMDGPU::FeatureVOP3P]) HasVOP3PInsts = true;
if (Bits[AMDGPU::FeatureVcmpxExecWARHazard]) HasVcmpxExecWARHazard = true;
if (Bits[AMDGPU::FeatureVcmpxPermlaneHazard]) HasVcmpxPermlaneHazard = true;
if (Bits[AMDGPU::FeatureVolcanicIslands] && Gen < GCNSubtarget::VOLCANIC_ISLANDS) Gen = GCNSubtarget::VOLCANIC_ISLANDS;
if (Bits[AMDGPU::FeatureVscnt]) HasVscnt = true;
if (Bits[AMDGPU::FeatureWavefrontSize16] && WavefrontSize < 16) WavefrontSize = 16;
if (Bits[AMDGPU::FeatureWavefrontSize32] && WavefrontSize < 32) WavefrontSize = 32;
if (Bits[AMDGPU::FeatureWavefrontSize64] && WavefrontSize < 64) WavefrontSize = 64;
if (Bits[AMDGPU::FeatureXNACK]) EnableXNACK = true;
if (Bits[AMDGPU::HalfRate64Ops]) HalfRate64Ops = true;
}
#endif // GET_SUBTARGETINFO_TARGET_DESC
#ifdef GET_SUBTARGETINFO_HEADER
#undef GET_SUBTARGETINFO_HEADER
namespace llvm {
class DFAPacketizer;
namespace AMDGPU_MC {
unsigned resolveVariantSchedClassImpl(unsigned SchedClass, const MCInst *MI, unsigned CPUID);
} // end namespace AMDGPU_MC
struct AMDGPUGenSubtargetInfo : public TargetSubtargetInfo {
explicit AMDGPUGenSubtargetInfo(const Triple &TT, StringRef CPU, StringRef FS);
public:
unsigned resolveSchedClass(unsigned SchedClass, const MachineInstr *DefMI, const TargetSchedModel *SchedModel) const override;
unsigned resolveVariantSchedClass(unsigned SchedClass, const MCInst *MI, unsigned CPUID) const override;
DFAPacketizer *createDFAPacketizer(const InstrItineraryData *IID) const;
};
} // end namespace llvm
#endif // GET_SUBTARGETINFO_HEADER
#ifdef GET_SUBTARGETINFO_CTOR
#undef GET_SUBTARGETINFO_CTOR
#include "llvm/CodeGen/TargetSchedule.h"
namespace llvm {
extern const llvm::SubtargetFeatureKV AMDGPUFeatureKV[];
extern const llvm::SubtargetSubTypeKV AMDGPUSubTypeKV[];
extern const llvm::MCWriteProcResEntry AMDGPUWriteProcResTable[];
extern const llvm::MCWriteLatencyEntry AMDGPUWriteLatencyTable[];
extern const llvm::MCReadAdvanceEntry AMDGPUReadAdvanceTable[];
AMDGPUGenSubtargetInfo::AMDGPUGenSubtargetInfo(const Triple &TT, StringRef CPU, StringRef FS)
: TargetSubtargetInfo(TT, CPU, FS, makeArrayRef(AMDGPUFeatureKV, 108), makeArrayRef(AMDGPUSubTypeKV, 39),
AMDGPUWriteProcResTable, AMDGPUWriteLatencyTable, AMDGPUReadAdvanceTable,
nullptr, nullptr, nullptr) {}
unsigned AMDGPUGenSubtargetInfo
::resolveSchedClass(unsigned SchedClass, const MachineInstr *MI, const TargetSchedModel *SchedModel) const {
const SIInstrInfo *TII =
static_cast<const SIInstrInfo*>(SchedModel->getInstrInfo());
(void)TII;
switch (SchedClass) {
case 21: // V_ACCVGPR_WRITE_B32
if (SchedModel->getProcessorID() == 1) { // SIQuarterSpeedModel
if (TII->hasVGPRUses(*MI))
return 26; // Write64Bit_MIVGPRRead
return 27; // Write64Bit_ReadDefault
}
if (SchedModel->getProcessorID() == 3) { // SIFullSpeedModel
if (TII->hasVGPRUses(*MI))
return 26; // Write64Bit_MIVGPRRead
return 27; // Write64Bit_ReadDefault
}
break;
case 25: // COPY
if (SchedModel->getProcessorID() == 1) { // SIQuarterSpeedModel
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) <= 32)
return 28; // Write32Bit
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) > 32)
return 29; // Write64Bit
return 30; // WriteSALU
}
if (SchedModel->getProcessorID() == 2) { // GFX10SpeedModel
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) <= 32)
return 28; // Write32Bit
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) > 32)
return 29; // Write64Bit
return 30; // WriteSALU
}
if (SchedModel->getProcessorID() == 3) { // SIFullSpeedModel
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) <= 32)
return 28; // Write32Bit
if (TII->isVGPRCopy(*MI) && TII->getOpSize(*MI, 0) > 32)
return 29; // Write64Bit
return 30; // WriteSALU
}
break;
};
report_fatal_error("Expected a variant SchedClass");
} // AMDGPUGenSubtargetInfo::resolveSchedClass
unsigned AMDGPUGenSubtargetInfo
::resolveVariantSchedClass(unsigned SchedClass, const MCInst *MI, unsigned CPUID) const {
return AMDGPU_MC::resolveVariantSchedClassImpl(SchedClass, MI, CPUID);
} // AMDGPUGenSubtargetInfo::resolveVariantSchedClass
} // end namespace llvm
#endif // GET_SUBTARGETINFO_CTOR
#ifdef GET_STIPREDICATE_DECLS_FOR_MC_ANALYSIS
#undef GET_STIPREDICATE_DECLS_FOR_MC_ANALYSIS
#endif // GET_STIPREDICATE_DECLS_FOR_MC_ANALYSIS
#ifdef GET_STIPREDICATE_DEFS_FOR_MC_ANALYSIS
#undef GET_STIPREDICATE_DEFS_FOR_MC_ANALYSIS
#endif // GET_STIPREDICATE_DEFS_FOR_MC_ANALYSIS
|