llvm.org GIT mirror llvm / bb1734f lib / Target / AMDGPU / Disassembler / AMDGPUDisassembler.cpp
bb1734f

Tree @bb1734f (Download .tar.gz)

AMDGPUDisassembler.cpp @bb1734fraw · history · blame

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
//===- AMDGPUDisassembler.cpp - Disassembler for AMDGPU ISA ---------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
//===----------------------------------------------------------------------===//
//
/// \file
///
/// This file contains definition for AMDGPU ISA disassembler
//
//===----------------------------------------------------------------------===//

// ToDo: What to do with instruction suffixes (v_mov_b32 vs v_mov_b32_e32)?

#include "Disassembler/AMDGPUDisassembler.h"
#include "AMDGPU.h"
#include "AMDGPURegisterInfo.h"
#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
#include "SIDefines.h"
#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
#include "Utils/AMDGPUBaseInfo.h"
#include "llvm-c/Disassembler.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/Twine.h"
#include "llvm/BinaryFormat/ELF.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCDisassembler/MCDisassembler.h"
#include "llvm/MC/MCExpr.h"
#include "llvm/MC/MCFixedLenDisassembler.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCSubtargetInfo.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include <iterator>
#include <tuple>
#include <vector>

using namespace llvm;

#define DEBUG_TYPE "amdgpu-disassembler"

using DecodeStatus = llvm::MCDisassembler::DecodeStatus;

inline static MCDisassembler::DecodeStatus
addOperand(MCInst &Inst, const MCOperand& Opnd) {
  Inst.addOperand(Opnd);
  return Opnd.isValid() ?
    MCDisassembler::Success :
    MCDisassembler::SoftFail;
}

static int insertNamedMCOperand(MCInst &MI, const MCOperand &Op,
                                uint16_t NameIdx) {
  int OpIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(), NameIdx);
  if (OpIdx != -1) {
    auto I = MI.begin();
    std::advance(I, OpIdx);
    MI.insert(I, Op);
  }
  return OpIdx;
}

static DecodeStatus decodeSoppBrTarget(MCInst &Inst, unsigned Imm,
                                       uint64_t Addr, const void *Decoder) {
  auto DAsm = static_cast<const AMDGPUDisassembler*>(Decoder);

  // Our branches take a simm16, but we need two extra bits to account for the
  // factor of 4.
  APInt SignedOffset(18, Imm * 4, true);
  int64_t Offset = (SignedOffset.sext(64) + 4 + Addr).getSExtValue();

  if (DAsm->tryAddingSymbolicOperand(Inst, Offset, Addr, true, 2, 2))
    return MCDisassembler::Success;
  return addOperand(Inst, MCOperand::createImm(Imm));
}

#define DECODE_OPERAND(StaticDecoderName, DecoderName) \
static DecodeStatus StaticDecoderName(MCInst &Inst, \
                                       unsigned Imm, \
                                       uint64_t /*Addr*/, \
                                       const void *Decoder) { \
  auto DAsm = static_cast<const AMDGPUDisassembler*>(Decoder); \
  return addOperand(Inst, DAsm->DecoderName(Imm)); \
}

#define DECODE_OPERAND_REG(RegClass) \
DECODE_OPERAND(Decode##RegClass##RegisterClass, decodeOperand_##RegClass)

DECODE_OPERAND_REG(VGPR_32)
DECODE_OPERAND_REG(VRegOrLds_32)
DECODE_OPERAND_REG(VS_32)
DECODE_OPERAND_REG(VS_64)
DECODE_OPERAND_REG(VS_128)

DECODE_OPERAND_REG(VReg_64)
DECODE_OPERAND_REG(VReg_96)
DECODE_OPERAND_REG(VReg_128)

DECODE_OPERAND_REG(SReg_32)
DECODE_OPERAND_REG(SReg_32_XM0_XEXEC)
DECODE_OPERAND_REG(SReg_32_XEXEC_HI)
DECODE_OPERAND_REG(SRegOrLds_32)
DECODE_OPERAND_REG(SReg_64)
DECODE_OPERAND_REG(SReg_64_XEXEC)
DECODE_OPERAND_REG(SReg_128)
DECODE_OPERAND_REG(SReg_256)
DECODE_OPERAND_REG(SReg_512)

static DecodeStatus decodeOperand_VSrc16(MCInst &Inst,
                                         unsigned Imm,
                                         uint64_t Addr,
                                         const void *Decoder) {
  auto DAsm = static_cast<const AMDGPUDisassembler*>(Decoder);
  return addOperand(Inst, DAsm->decodeOperand_VSrc16(Imm));
}

static DecodeStatus decodeOperand_VSrcV216(MCInst &Inst,
                                         unsigned Imm,
                                         uint64_t Addr,
                                         const void *Decoder) {
  auto DAsm = static_cast<const AMDGPUDisassembler*>(Decoder);
  return addOperand(Inst, DAsm->decodeOperand_VSrcV216(Imm));
}

#define DECODE_SDWA(DecName) \
DECODE_OPERAND(decodeSDWA##DecName, decodeSDWA##DecName)

DECODE_SDWA(Src32)
DECODE_SDWA(Src16)
DECODE_SDWA(VopcDst)

#include "AMDGPUGenDisassemblerTables.inc"

//===----------------------------------------------------------------------===//
//
//===----------------------------------------------------------------------===//

template <typename T> static inline T eatBytes(ArrayRef<uint8_t>& Bytes) {
  assert(Bytes.size() >= sizeof(T));
  const auto Res = support::endian::read<T, support::endianness::little>(Bytes.data());
  Bytes = Bytes.slice(sizeof(T));
  return Res;
}

DecodeStatus AMDGPUDisassembler::tryDecodeInst(const uint8_t* Table,
                                               MCInst &MI,
                                               uint64_t Inst,
                                               uint64_t Address) const {
  assert(MI.getOpcode() == 0);
  assert(MI.getNumOperands() == 0);
  MCInst TmpInst;
  HasLiteral = false;
  const auto SavedBytes = Bytes;
  if (decodeInstruction(Table, TmpInst, Inst, Address, this, STI)) {
    MI = TmpInst;
    return MCDisassembler::Success;
  }
  Bytes = SavedBytes;
  return MCDisassembler::Fail;
}

DecodeStatus AMDGPUDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
                                                ArrayRef<uint8_t> Bytes_,
                                                uint64_t Address,
                                                raw_ostream &WS,
                                                raw_ostream &CS) const {
  CommentStream = &CS;
  bool IsSDWA = false;

  // ToDo: AMDGPUDisassembler supports only VI ISA.
  if (!STI.getFeatureBits()[AMDGPU::FeatureGCN3Encoding])
    report_fatal_error("Disassembly not yet supported for subtarget");

  const unsigned MaxInstBytesNum = (std::min)((size_t)8, Bytes_.size());
  Bytes = Bytes_.slice(0, MaxInstBytesNum);

  DecodeStatus Res = MCDisassembler::Fail;
  do {
    // ToDo: better to switch encoding length using some bit predicate
    // but it is unknown yet, so try all we can

    // Try to decode DPP and SDWA first to solve conflict with VOP1 and VOP2
    // encodings
    if (Bytes.size() >= 8) {
      const uint64_t QW = eatBytes<uint64_t>(Bytes);
      Res = tryDecodeInst(DecoderTableDPP64, MI, QW, Address);
      if (Res) break;

      Res = tryDecodeInst(DecoderTableSDWA64, MI, QW, Address);
      if (Res) { IsSDWA = true;  break; }

      Res = tryDecodeInst(DecoderTableSDWA964, MI, QW, Address);
      if (Res) { IsSDWA = true;  break; }

      if (STI.getFeatureBits()[AMDGPU::FeatureUnpackedD16VMem]) {
        Res = tryDecodeInst(DecoderTableGFX80_UNPACKED64, MI, QW, Address);
        if (Res)
          break;
      }

      // Some GFX9 subtargets repurposed the v_mad_mix_f32, v_mad_mixlo_f16 and
      // v_mad_mixhi_f16 for FMA variants. Try to decode using this special
      // table first so we print the correct name.
      if (STI.getFeatureBits()[AMDGPU::FeatureFmaMixInsts]) {
        Res = tryDecodeInst(DecoderTableGFX9_DL64, MI, QW, Address);
        if (Res)
          break;
      }
    }

    // Reinitialize Bytes as DPP64 could have eaten too much
    Bytes = Bytes_.slice(0, MaxInstBytesNum);

    // Try decode 32-bit instruction
    if (Bytes.size() < 4) break;
    const uint32_t DW = eatBytes<uint32_t>(Bytes);
    Res = tryDecodeInst(DecoderTableVI32, MI, DW, Address);
    if (Res) break;

    Res = tryDecodeInst(DecoderTableAMDGPU32, MI, DW, Address);
    if (Res) break;

    Res = tryDecodeInst(DecoderTableGFX932, MI, DW, Address);
    if (Res) break;

    if (Bytes.size() < 4) break;
    const uint64_t QW = ((uint64_t)eatBytes<uint32_t>(Bytes) << 32) | DW;
    Res = tryDecodeInst(DecoderTableVI64, MI, QW, Address);
    if (Res) break;

    Res = tryDecodeInst(DecoderTableAMDGPU64, MI, QW, Address);
    if (Res) break;

    Res = tryDecodeInst(DecoderTableGFX964, MI, QW, Address);
  } while (false);

  if (Res && (MI.getOpcode() == AMDGPU::V_MAC_F32_e64_vi ||
              MI.getOpcode() == AMDGPU::V_MAC_F32_e64_si ||
              MI.getOpcode() == AMDGPU::V_MAC_F16_e64_vi ||
              MI.getOpcode() == AMDGPU::V_FMAC_F32_e64_vi)) {
    // Insert dummy unused src2_modifiers.
    insertNamedMCOperand(MI, MCOperand::createImm(0),
                         AMDGPU::OpName::src2_modifiers);
  }

  if (Res && (MCII->get(MI.getOpcode()).TSFlags & SIInstrFlags::MIMG)) {
    Res = convertMIMGInst(MI);
  }

  if (Res && IsSDWA)
    Res = convertSDWAInst(MI);

  // if the opcode was not recognized we'll assume a Size of 4 bytes
  // (unless there are fewer bytes left)
  Size = Res ? (MaxInstBytesNum - Bytes.size())
             : std::min((size_t)4, Bytes_.size());
  return Res;
}

DecodeStatus AMDGPUDisassembler::convertSDWAInst(MCInst &MI) const {
  if (STI.getFeatureBits()[AMDGPU::FeatureGFX9]) {
    if (AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::sdst) != -1)
      // VOPC - insert clamp
      insertNamedMCOperand(MI, MCOperand::createImm(0), AMDGPU::OpName::clamp);
  } else if (STI.getFeatureBits()[AMDGPU::FeatureVolcanicIslands]) {
    int SDst = AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::sdst);
    if (SDst != -1) {
      // VOPC - insert VCC register as sdst
      insertNamedMCOperand(MI, createRegOperand(AMDGPU::VCC),
                           AMDGPU::OpName::sdst);
    } else {
      // VOP1/2 - insert omod if present in instruction
      insertNamedMCOperand(MI, MCOperand::createImm(0), AMDGPU::OpName::omod);
    }
  }
  return MCDisassembler::Success;
}

// Note that MIMG format provides no information about VADDR size.
// Consequently, decoded instructions always show address
// as if it has 1 dword, which could be not really so.
DecodeStatus AMDGPUDisassembler::convertMIMGInst(MCInst &MI) const {

  int VDstIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                           AMDGPU::OpName::vdst);

  int VDataIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                            AMDGPU::OpName::vdata);

  int DMaskIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                            AMDGPU::OpName::dmask);

  int TFEIdx   = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                            AMDGPU::OpName::tfe);
  int D16Idx   = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                            AMDGPU::OpName::d16);

  assert(VDataIdx != -1);
  assert(DMaskIdx != -1);
  assert(TFEIdx != -1);

  bool IsAtomic = (VDstIdx != -1);
  bool IsGather4 = MCII->get(MI.getOpcode()).TSFlags & SIInstrFlags::Gather4;

  unsigned DMask = MI.getOperand(DMaskIdx).getImm() & 0xf;
  if (DMask == 0)
    return MCDisassembler::Success;

  unsigned DstSize = IsGather4 ? 4 : countPopulation(DMask);
  if (DstSize == 1)
    return MCDisassembler::Success;

  bool D16 = D16Idx >= 0 && MI.getOperand(D16Idx).getImm();
  if (D16 && AMDGPU::hasPackedD16(STI)) {
    DstSize = (DstSize + 1) / 2;
  }

  // FIXME: Add tfe support
  if (MI.getOperand(TFEIdx).getImm())
    return MCDisassembler::Success;

  int NewOpcode = -1;

  if (IsGather4) {
    if (D16 && AMDGPU::hasPackedD16(STI))
      NewOpcode = AMDGPU::getMaskedMIMGOp(MI.getOpcode(), 2);
    else
      return MCDisassembler::Success;
  } else {
    NewOpcode = AMDGPU::getMaskedMIMGOp(MI.getOpcode(), DstSize);
    if (NewOpcode == -1)
      return MCDisassembler::Success;
  }

  auto RCID = MCII->get(NewOpcode).OpInfo[VDataIdx].RegClass;

  // Get first subregister of VData
  unsigned Vdata0 = MI.getOperand(VDataIdx).getReg();
  unsigned VdataSub0 = MRI.getSubReg(Vdata0, AMDGPU::sub0);
  Vdata0 = (VdataSub0 != 0)? VdataSub0 : Vdata0;

  // Widen the register to the correct number of enabled channels.
  auto NewVdata = MRI.getMatchingSuperReg(Vdata0, AMDGPU::sub0,
                                          &MRI.getRegClass(RCID));
  if (NewVdata == AMDGPU::NoRegister) {
    // It's possible to encode this such that the low register + enabled
    // components exceeds the register count.
    return MCDisassembler::Success;
  }

  MI.setOpcode(NewOpcode);
  // vaddr will be always appear as a single VGPR. This will look different than
  // how it is usually emitted because the number of register components is not
  // in the instruction encoding.
  MI.getOperand(VDataIdx) = MCOperand::createReg(NewVdata);

  if (IsAtomic) {
    // Atomic operations have an additional operand (a copy of data)
    MI.getOperand(VDstIdx) = MCOperand::createReg(NewVdata);
  }

  return MCDisassembler::Success;
}

const char* AMDGPUDisassembler::getRegClassName(unsigned RegClassID) const {
  return getContext().getRegisterInfo()->
    getRegClassName(&AMDGPUMCRegisterClasses[RegClassID]);
}

inline
MCOperand AMDGPUDisassembler::errOperand(unsigned V,
                                         const Twine& ErrMsg) const {
  *CommentStream << "Error: " + ErrMsg;

  // ToDo: add support for error operands to MCInst.h
  // return MCOperand::createError(V);
  return MCOperand();
}

inline
MCOperand AMDGPUDisassembler::createRegOperand(unsigned int RegId) const {
  return MCOperand::createReg(AMDGPU::getMCReg(RegId, STI));
}

inline
MCOperand AMDGPUDisassembler::createRegOperand(unsigned RegClassID,
                                               unsigned Val) const {
  const auto& RegCl = AMDGPUMCRegisterClasses[RegClassID];
  if (Val >= RegCl.getNumRegs())
    return errOperand(Val, Twine(getRegClassName(RegClassID)) +
                           ": unknown register " + Twine(Val));
  return createRegOperand(RegCl.getRegister(Val));
}

inline
MCOperand AMDGPUDisassembler::createSRegOperand(unsigned SRegClassID,
                                                unsigned Val) const {
  // ToDo: SI/CI have 104 SGPRs, VI - 102
  // Valery: here we accepting as much as we can, let assembler sort it out
  int shift = 0;
  switch (SRegClassID) {
  case AMDGPU::SGPR_32RegClassID:
  case AMDGPU::TTMP_32RegClassID:
    break;
  case AMDGPU::SGPR_64RegClassID:
  case AMDGPU::TTMP_64RegClassID:
    shift = 1;
    break;
  case AMDGPU::SGPR_128RegClassID:
  case AMDGPU::TTMP_128RegClassID:
  // ToDo: unclear if s[100:104] is available on VI. Can we use VCC as SGPR in
  // this bundle?
  case AMDGPU::SGPR_256RegClassID:
  case AMDGPU::TTMP_256RegClassID:
    // ToDo: unclear if s[96:104] is available on VI. Can we use VCC as SGPR in
  // this bundle?
  case AMDGPU::SGPR_512RegClassID:
  case AMDGPU::TTMP_512RegClassID:
    shift = 2;
    break;
  // ToDo: unclear if s[88:104] is available on VI. Can we use VCC as SGPR in
  // this bundle?
  default:
    llvm_unreachable("unhandled register class");
  }

  if (Val % (1 << shift)) {
    *CommentStream << "Warning: " << getRegClassName(SRegClassID)
                   << ": scalar reg isn't aligned " << Val;
  }

  return createRegOperand(SRegClassID, Val >> shift);
}

MCOperand AMDGPUDisassembler::decodeOperand_VS_32(unsigned Val) const {
  return decodeSrcOp(OPW32, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VS_64(unsigned Val) const {
  return decodeSrcOp(OPW64, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VS_128(unsigned Val) const {
  return decodeSrcOp(OPW128, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VSrc16(unsigned Val) const {
  return decodeSrcOp(OPW16, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VSrcV216(unsigned Val) const {
  return decodeSrcOp(OPWV216, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VGPR_32(unsigned Val) const {
  // Some instructions have operand restrictions beyond what the encoding
  // allows. Some ordinarily VSrc_32 operands are VGPR_32, so clear the extra
  // high bit.
  Val &= 255;

  return createRegOperand(AMDGPU::VGPR_32RegClassID, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VRegOrLds_32(unsigned Val) const {
  return decodeSrcOp(OPW32, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VReg_64(unsigned Val) const {
  return createRegOperand(AMDGPU::VReg_64RegClassID, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VReg_96(unsigned Val) const {
  return createRegOperand(AMDGPU::VReg_96RegClassID, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_VReg_128(unsigned Val) const {
  return createRegOperand(AMDGPU::VReg_128RegClassID, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_32(unsigned Val) const {
  // table-gen generated disassembler doesn't care about operand types
  // leaving only registry class so SSrc_32 operand turns into SReg_32
  // and therefore we accept immediates and literals here as well
  return decodeSrcOp(OPW32, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_32_XM0_XEXEC(
  unsigned Val) const {
  // SReg_32_XM0 is SReg_32 without M0 or EXEC_LO/EXEC_HI
  return decodeOperand_SReg_32(Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_32_XEXEC_HI(
  unsigned Val) const {
  // SReg_32_XM0 is SReg_32 without EXEC_HI
  return decodeOperand_SReg_32(Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SRegOrLds_32(unsigned Val) const {
  // table-gen generated disassembler doesn't care about operand types
  // leaving only registry class so SSrc_32 operand turns into SReg_32
  // and therefore we accept immediates and literals here as well
  return decodeSrcOp(OPW32, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_64(unsigned Val) const {
  return decodeSrcOp(OPW64, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_64_XEXEC(unsigned Val) const {
  return decodeSrcOp(OPW64, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_128(unsigned Val) const {
  return decodeSrcOp(OPW128, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_256(unsigned Val) const {
  return decodeDstOp(OPW256, Val);
}

MCOperand AMDGPUDisassembler::decodeOperand_SReg_512(unsigned Val) const {
  return decodeDstOp(OPW512, Val);
}

MCOperand AMDGPUDisassembler::decodeLiteralConstant() const {
  // For now all literal constants are supposed to be unsigned integer
  // ToDo: deal with signed/unsigned 64-bit integer constants
  // ToDo: deal with float/double constants
  if (!HasLiteral) {
    if (Bytes.size() < 4) {
      return errOperand(0, "cannot read literal, inst bytes left " +
                        Twine(Bytes.size()));
    }
    HasLiteral = true;
    Literal = eatBytes<uint32_t>(Bytes);
  }
  return MCOperand::createImm(Literal);
}

MCOperand AMDGPUDisassembler::decodeIntImmed(unsigned Imm) {
  using namespace AMDGPU::EncValues;

  assert(Imm >= INLINE_INTEGER_C_MIN && Imm <= INLINE_INTEGER_C_MAX);
  return MCOperand::createImm((Imm <= INLINE_INTEGER_C_POSITIVE_MAX) ?
    (static_cast<int64_t>(Imm) - INLINE_INTEGER_C_MIN) :
    (INLINE_INTEGER_C_POSITIVE_MAX - static_cast<int64_t>(Imm)));
      // Cast prevents negative overflow.
}

static int64_t getInlineImmVal32(unsigned Imm) {
  switch (Imm) {
  case 240:
    return FloatToBits(0.5f);
  case 241:
    return FloatToBits(-0.5f);
  case 242:
    return FloatToBits(1.0f);
  case 243:
    return FloatToBits(-1.0f);
  case 244:
    return FloatToBits(2.0f);
  case 245:
    return FloatToBits(-2.0f);
  case 246:
    return FloatToBits(4.0f);
  case 247:
    return FloatToBits(-4.0f);
  case 248: // 1 / (2 * PI)
    return 0x3e22f983;
  default:
    llvm_unreachable("invalid fp inline imm");
  }
}

static int64_t getInlineImmVal64(unsigned Imm) {
  switch (Imm) {
  case 240:
    return DoubleToBits(0.5);
  case 241:
    return DoubleToBits(-0.5);
  case 242:
    return DoubleToBits(1.0);
  case 243:
    return DoubleToBits(-1.0);
  case 244:
    return DoubleToBits(2.0);
  case 245:
    return DoubleToBits(-2.0);
  case 246:
    return DoubleToBits(4.0);
  case 247:
    return DoubleToBits(-4.0);
  case 248: // 1 / (2 * PI)
    return 0x3fc45f306dc9c882;
  default:
    llvm_unreachable("invalid fp inline imm");
  }
}

static int64_t getInlineImmVal16(unsigned Imm) {
  switch (Imm) {
  case 240:
    return 0x3800;
  case 241:
    return 0xB800;
  case 242:
    return 0x3C00;
  case 243:
    return 0xBC00;
  case 244:
    return 0x4000;
  case 245:
    return 0xC000;
  case 246:
    return 0x4400;
  case 247:
    return 0xC400;
  case 248: // 1 / (2 * PI)
    return 0x3118;
  default:
    llvm_unreachable("invalid fp inline imm");
  }
}

MCOperand AMDGPUDisassembler::decodeFPImmed(OpWidthTy Width, unsigned Imm) {
  assert(Imm >= AMDGPU::EncValues::INLINE_FLOATING_C_MIN
      && Imm <= AMDGPU::EncValues::INLINE_FLOATING_C_MAX);

  // ToDo: case 248: 1/(2*PI) - is allowed only on VI
  switch (Width) {
  case OPW32:
    return MCOperand::createImm(getInlineImmVal32(Imm));
  case OPW64:
    return MCOperand::createImm(getInlineImmVal64(Imm));
  case OPW16:
  case OPWV216:
    return MCOperand::createImm(getInlineImmVal16(Imm));
  default:
    llvm_unreachable("implement me");
  }
}

unsigned AMDGPUDisassembler::getVgprClassId(const OpWidthTy Width) const {
  using namespace AMDGPU;

  assert(OPW_FIRST_ <= Width && Width < OPW_LAST_);
  switch (Width) {
  default: // fall
  case OPW32:
  case OPW16:
  case OPWV216:
    return VGPR_32RegClassID;
  case OPW64: return VReg_64RegClassID;
  case OPW128: return VReg_128RegClassID;
  }
}

unsigned AMDGPUDisassembler::getSgprClassId(const OpWidthTy Width) const {
  using namespace AMDGPU;

  assert(OPW_FIRST_ <= Width && Width < OPW_LAST_);
  switch (Width) {
  default: // fall
  case OPW32:
  case OPW16:
  case OPWV216:
    return SGPR_32RegClassID;
  case OPW64: return SGPR_64RegClassID;
  case OPW128: return SGPR_128RegClassID;
  case OPW256: return SGPR_256RegClassID;
  case OPW512: return SGPR_512RegClassID;
  }
}

unsigned AMDGPUDisassembler::getTtmpClassId(const OpWidthTy Width) const {
  using namespace AMDGPU;

  assert(OPW_FIRST_ <= Width && Width < OPW_LAST_);
  switch (Width) {
  default: // fall
  case OPW32:
  case OPW16:
  case OPWV216:
    return TTMP_32RegClassID;
  case OPW64: return TTMP_64RegClassID;
  case OPW128: return TTMP_128RegClassID;
  case OPW256: return TTMP_256RegClassID;
  case OPW512: return TTMP_512RegClassID;
  }
}

int AMDGPUDisassembler::getTTmpIdx(unsigned Val) const {
  using namespace AMDGPU::EncValues;

  unsigned TTmpMin = isGFX9() ? TTMP_GFX9_MIN : TTMP_VI_MIN;
  unsigned TTmpMax = isGFX9() ? TTMP_GFX9_MAX : TTMP_VI_MAX;

  return (TTmpMin <= Val && Val <= TTmpMax)? Val - TTmpMin : -1;
}

MCOperand AMDGPUDisassembler::decodeSrcOp(const OpWidthTy Width, unsigned Val) const {
  using namespace AMDGPU::EncValues;

  assert(Val < 512); // enum9

  if (VGPR_MIN <= Val && Val <= VGPR_MAX) {
    return createRegOperand(getVgprClassId(Width), Val - VGPR_MIN);
  }
  if (Val <= SGPR_MAX) {
    assert(SGPR_MIN == 0); // "SGPR_MIN <= Val" is always true and causes compilation warning.
    return createSRegOperand(getSgprClassId(Width), Val - SGPR_MIN);
  }

  int TTmpIdx = getTTmpIdx(Val);
  if (TTmpIdx >= 0) {
    return createSRegOperand(getTtmpClassId(Width), TTmpIdx);
  }

  if (INLINE_INTEGER_C_MIN <= Val && Val <= INLINE_INTEGER_C_MAX)
    return decodeIntImmed(Val);

  if (INLINE_FLOATING_C_MIN <= Val && Val <= INLINE_FLOATING_C_MAX)
    return decodeFPImmed(Width, Val);

  if (Val == LITERAL_CONST)
    return decodeLiteralConstant();

  switch (Width) {
  case OPW32:
  case OPW16:
  case OPWV216:
    return decodeSpecialReg32(Val);
  case OPW64:
    return decodeSpecialReg64(Val);
  default:
    llvm_unreachable("unexpected immediate type");
  }
}

MCOperand AMDGPUDisassembler::decodeDstOp(const OpWidthTy Width, unsigned Val) const {
  using namespace AMDGPU::EncValues;

  assert(Val < 128);
  assert(Width == OPW256 || Width == OPW512);

  if (Val <= SGPR_MAX) {
    assert(SGPR_MIN == 0); // "SGPR_MIN <= Val" is always true and causes compilation warning.
    return createSRegOperand(getSgprClassId(Width), Val - SGPR_MIN);
  }

  int TTmpIdx = getTTmpIdx(Val);
  if (TTmpIdx >= 0) {
    return createSRegOperand(getTtmpClassId(Width), TTmpIdx);
  }

  llvm_unreachable("unknown dst register");
}

MCOperand AMDGPUDisassembler::decodeSpecialReg32(unsigned Val) const {
  using namespace AMDGPU;

  switch (Val) {
  case 102: return createRegOperand(FLAT_SCR_LO);
  case 103: return createRegOperand(FLAT_SCR_HI);
  case 104: return createRegOperand(XNACK_MASK_LO);
  case 105: return createRegOperand(XNACK_MASK_HI);
  case 106: return createRegOperand(VCC_LO);
  case 107: return createRegOperand(VCC_HI);
  case 108: assert(!isGFX9()); return createRegOperand(TBA_LO);
  case 109: assert(!isGFX9()); return createRegOperand(TBA_HI);
  case 110: assert(!isGFX9()); return createRegOperand(TMA_LO);
  case 111: assert(!isGFX9()); return createRegOperand(TMA_HI);
  case 124: return createRegOperand(M0);
  case 126: return createRegOperand(EXEC_LO);
  case 127: return createRegOperand(EXEC_HI);
  case 235: return createRegOperand(SRC_SHARED_BASE);
  case 236: return createRegOperand(SRC_SHARED_LIMIT);
  case 237: return createRegOperand(SRC_PRIVATE_BASE);
  case 238: return createRegOperand(SRC_PRIVATE_LIMIT);
    // TODO: SRC_POPS_EXITING_WAVE_ID
    // ToDo: no support for vccz register
  case 251: break;
    // ToDo: no support for execz register
  case 252: break;
  case 253: return createRegOperand(SCC);
  case 254: return createRegOperand(LDS_DIRECT);
  default: break;
  }
  return errOperand(Val, "unknown operand encoding " + Twine(Val));
}

MCOperand AMDGPUDisassembler::decodeSpecialReg64(unsigned Val) const {
  using namespace AMDGPU;

  switch (Val) {
  case 102: return createRegOperand(FLAT_SCR);
  case 104: return createRegOperand(XNACK_MASK);
  case 106: return createRegOperand(VCC);
  case 108: assert(!isGFX9()); return createRegOperand(TBA);
  case 110: assert(!isGFX9()); return createRegOperand(TMA);
  case 126: return createRegOperand(EXEC);
  default: break;
  }
  return errOperand(Val, "unknown operand encoding " + Twine(Val));
}

MCOperand AMDGPUDisassembler::decodeSDWASrc(const OpWidthTy Width,
                                            const unsigned Val) const {
  using namespace AMDGPU::SDWA;
  using namespace AMDGPU::EncValues;

  if (STI.getFeatureBits()[AMDGPU::FeatureGFX9]) {
    // XXX: static_cast<int> is needed to avoid stupid warning:
    // compare with unsigned is always true
    if (SDWA9EncValues::SRC_VGPR_MIN <= static_cast<int>(Val) &&
        Val <= SDWA9EncValues::SRC_VGPR_MAX) {
      return createRegOperand(getVgprClassId(Width),
                              Val - SDWA9EncValues::SRC_VGPR_MIN);
    }
    if (SDWA9EncValues::SRC_SGPR_MIN <= Val &&
        Val <= SDWA9EncValues::SRC_SGPR_MAX) {
      return createSRegOperand(getSgprClassId(Width),
                               Val - SDWA9EncValues::SRC_SGPR_MIN);
    }
    if (SDWA9EncValues::SRC_TTMP_MIN <= Val &&
        Val <= SDWA9EncValues::SRC_TTMP_MAX) {
      return createSRegOperand(getTtmpClassId(Width),
                               Val - SDWA9EncValues::SRC_TTMP_MIN);
    }

    const unsigned SVal = Val - SDWA9EncValues::SRC_SGPR_MIN;

    if (INLINE_INTEGER_C_MIN <= SVal && SVal <= INLINE_INTEGER_C_MAX)
      return decodeIntImmed(SVal);

    if (INLINE_FLOATING_C_MIN <= SVal && SVal <= INLINE_FLOATING_C_MAX)
      return decodeFPImmed(Width, SVal);

    return decodeSpecialReg32(SVal);
  } else if (STI.getFeatureBits()[AMDGPU::FeatureVolcanicIslands]) {
    return createRegOperand(getVgprClassId(Width), Val);
  }
  llvm_unreachable("unsupported target");
}

MCOperand AMDGPUDisassembler::decodeSDWASrc16(unsigned Val) const {
  return decodeSDWASrc(OPW16, Val);
}

MCOperand AMDGPUDisassembler::decodeSDWASrc32(unsigned Val) const {
  return decodeSDWASrc(OPW32, Val);
}

MCOperand AMDGPUDisassembler::decodeSDWAVopcDst(unsigned Val) const {
  using namespace AMDGPU::SDWA;

  assert(STI.getFeatureBits()[AMDGPU::FeatureGFX9] &&
         "SDWAVopcDst should be present only on GFX9");
  if (Val & SDWA9EncValues::VOPC_DST_VCC_MASK) {
    Val &= SDWA9EncValues::VOPC_DST_SGPR_MASK;

    int TTmpIdx = getTTmpIdx(Val);
    if (TTmpIdx >= 0) {
      return createSRegOperand(getTtmpClassId(OPW64), TTmpIdx);
    } else if (Val > AMDGPU::EncValues::SGPR_MAX) {
      return decodeSpecialReg64(Val);
    } else {
      return createSRegOperand(getSgprClassId(OPW64), Val);
    }
  } else {
    return createRegOperand(AMDGPU::VCC);
  }
}

bool AMDGPUDisassembler::isVI() const {
  return STI.getFeatureBits()[AMDGPU::FeatureVolcanicIslands];
}

bool AMDGPUDisassembler::isGFX9() const {
  return STI.getFeatureBits()[AMDGPU::FeatureGFX9];
}

//===----------------------------------------------------------------------===//
// AMDGPUSymbolizer
//===----------------------------------------------------------------------===//

// Try to find symbol name for specified label
bool AMDGPUSymbolizer::tryAddingSymbolicOperand(MCInst &Inst,
                                raw_ostream &/*cStream*/, int64_t Value,
                                uint64_t /*Address*/, bool IsBranch,
                                uint64_t /*Offset*/, uint64_t /*InstSize*/) {
  using SymbolInfoTy = std::tuple<uint64_t, StringRef, uint8_t>;
  using SectionSymbolsTy = std::vector<SymbolInfoTy>;

  if (!IsBranch) {
    return false;
  }

  auto *Symbols = static_cast<SectionSymbolsTy *>(DisInfo);
  if (!Symbols)
    return false;

  auto Result = std::find_if(Symbols->begin(), Symbols->end(),
                             [Value](const SymbolInfoTy& Val) {
                                return std::get<0>(Val) == static_cast<uint64_t>(Value)
                                    && std::get<2>(Val) == ELF::STT_NOTYPE;
                             });
  if (Result != Symbols->end()) {
    auto *Sym = Ctx.getOrCreateSymbol(std::get<1>(*Result));
    const auto *Add = MCSymbolRefExpr::create(Sym, Ctx);
    Inst.addOperand(MCOperand::createExpr(Add));
    return true;
  }
  return false;
}

void AMDGPUSymbolizer::tryAddingPcLoadReferenceComment(raw_ostream &cStream,
                                                       int64_t Value,
                                                       uint64_t Address) {
  llvm_unreachable("unimplemented");
}

//===----------------------------------------------------------------------===//
// Initialization
//===----------------------------------------------------------------------===//

static MCSymbolizer *createAMDGPUSymbolizer(const Triple &/*TT*/,
                              LLVMOpInfoCallback /*GetOpInfo*/,
                              LLVMSymbolLookupCallback /*SymbolLookUp*/,
                              void *DisInfo,
                              MCContext *Ctx,
                              std::unique_ptr<MCRelocationInfo> &&RelInfo) {
  return new AMDGPUSymbolizer(*Ctx, std::move(RelInfo), DisInfo);
}

static MCDisassembler *createAMDGPUDisassembler(const Target &T,
                                                const MCSubtargetInfo &STI,
                                                MCContext &Ctx) {
  return new AMDGPUDisassembler(STI, Ctx, T.createMCInstrInfo());
}

extern "C" void LLVMInitializeAMDGPUDisassembler() {
  TargetRegistry::RegisterMCDisassembler(getTheGCNTarget(),
                                         createAMDGPUDisassembler);
  TargetRegistry::RegisterMCSymbolizer(getTheGCNTarget(),
                                       createAMDGPUSymbolizer);
}