-
Notifications
You must be signed in to change notification settings - Fork 107
/
Copy pathdisassembler_riscv64.cc
2025 lines (1896 loc) · 71.8 KB
/
disassembler_riscv64.cc
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
/*
* Copyright (C) 2023 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "disassembler_riscv64.h"
#include "android-base/logging.h"
#include "android-base/stringprintf.h"
#include "base/bit_utils.h"
#include "base/casts.h"
using android::base::StringPrintf;
namespace art {
namespace riscv64 {
class DisassemblerRiscv64::Printer {
public:
Printer(DisassemblerRiscv64* disassembler, std::ostream& os)
: disassembler_(disassembler), os_(os) {}
void Dump32(const uint8_t* insn);
void Dump16(const uint8_t* insn);
void Dump2Byte(const uint8_t* data);
void DumpByte(const uint8_t* data);
private:
// This enumeration should mirror the declarations in runtime/arch/riscv64/registers_riscv64.h.
// We do not include that file to avoid a dependency on libart.
enum {
Zero = 0,
RA = 1,
FP = 8,
TR = 9,
};
enum class MemAddressMode : uint32_t {
kUnitStride = 0b00,
kIndexedUnordered = 0b01,
kStrided = 0b10,
kIndexedOrdered = 0b11,
};
enum class Nf : uint32_t {
k1 = 0b000,
k2 = 0b001,
k3 = 0b010,
k4 = 0b011,
k5 = 0b100,
k6 = 0b101,
k7 = 0b110,
k8 = 0b111,
};
enum class VAIEncodings : uint32_t {
kOpIVV = 0b000,
kOpFVV = 0b001,
kOpMVV = 0b010,
kOpIVI = 0b011,
kOpIVX = 0b100,
kOpFVF = 0b101,
kOpMVX = 0b110,
kOpCFG = 0b111,
};
class ScopedNewLinePrinter {
std::ostream& os_;
public:
explicit ScopedNewLinePrinter(std::ostream& os) : os_(os) {}
~ScopedNewLinePrinter() { os_ << '\n'; }
};
static const char* XRegName(uint32_t regno);
static const char* FRegName(uint32_t regno);
static const char* VRegName(uint32_t regno);
static const char* RoundingModeName(uint32_t rm);
// Regular instruction immediate utils
static int32_t Decode32Imm12(uint32_t insn32) {
uint32_t sign = (insn32 >> 31);
uint32_t imm12 = (insn32 >> 20);
return static_cast<int32_t>(imm12) - static_cast<int32_t>(sign << 12); // Sign-extend.
}
static uint32_t Decode32UImm7(uint32_t insn32) { return (insn32 >> 25) & 0x7Fu; }
static uint32_t Decode32UImm12(uint32_t insn32) { return (insn32 >> 20) & 0xFFFu; }
static int32_t Decode32StoreOffset(uint32_t insn32) {
uint32_t bit11 = insn32 >> 31;
uint32_t bits5_11 = insn32 >> 25;
uint32_t bits0_4 = (insn32 >> 7) & 0x1fu;
uint32_t imm = (bits5_11 << 5) + bits0_4;
return static_cast<int32_t>(imm) - static_cast<int32_t>(bit11 << 12); // Sign-extend.
}
// Compressed instruction immediate utils
// Extracts the offset from a compressed instruction
// where `offset[5:3]` is in bits `[12:10]` and `offset[2|6]` is in bits `[6:5]`
static uint32_t Decode16CMOffsetW(uint32_t insn16) {
DCHECK(IsUint<16>(insn16));
return BitFieldExtract(insn16, 5, 1) << 6 | BitFieldExtract(insn16, 10, 3) << 3 |
BitFieldExtract(insn16, 6, 1) << 2;
}
// Extracts the offset from a compressed instruction
// where `offset[5:3]` is in bits `[12:10]` and `offset[7:6]` is in bits `[6:5]`
static uint32_t Decode16CMOffsetD(uint32_t insn16) {
DCHECK(IsUint<16>(insn16));
return BitFieldExtract(insn16, 5, 2) << 6 | BitFieldExtract(insn16, 10, 3) << 3;
}
// Re-orders raw immediatate into real value
// where `imm[5:3]` is in bits `[5:3]` and `imm[8:6]` is in bits `[2:0]`
static uint32_t Uimm6ToOffsetD16(uint32_t uimm6) {
DCHECK(IsUint<6>(uimm6));
return (BitFieldExtract(uimm6, 3, 3) << 3) | (BitFieldExtract(uimm6, 0, 3) << 6);
}
// Re-orders raw immediatate to form real value
// where `imm[5:2]` is in bits `[5:2]` and `imm[7:6]` is in bits `[1:0]`
static uint32_t Uimm6ToOffsetW16(uint32_t uimm6) {
DCHECK(IsUint<6>(uimm6));
return (BitFieldExtract(uimm6, 2, 4) << 2) | (BitFieldExtract(uimm6, 0, 2) << 6);
}
// Re-orders raw immediatate to form real value
// where `imm[1]` is in bit `[0]` and `imm[0]` is in bit `[1]`
static uint32_t Uimm2ToOffset10(uint32_t uimm2) {
DCHECK(IsUint<2>(uimm2));
return (uimm2 >> 1) | (uimm2 & 0x1u) << 1;
}
// Re-orders raw immediatate to form real value
// where `imm[1]` is in bit `[0]` and `imm[0]` is `0`
static uint32_t Uimm2ToOffset1(uint32_t uimm2) {
DCHECK(IsUint<2>(uimm2));
return (uimm2 & 0x1u) << 1;
}
template <size_t kWidth>
static constexpr int32_t SignExtendBits(uint32_t bits) {
static_assert(kWidth < BitSizeOf<uint32_t>());
const uint32_t sign_bit = (bits >> kWidth) & 1u;
return static_cast<int32_t>(bits) - static_cast<int32_t>(sign_bit << kWidth);
}
// Extracts the immediate from a compressed instruction
// where `imm[5]` is in bit `[12]` and `imm[4:0]` is in bits `[6:2]`
// and performs sign-extension if required
template <typename T>
static T Decode16Imm6(uint32_t insn16) {
DCHECK(IsUint<16>(insn16));
static_assert(std::is_integral_v<T>, "T must be integral");
const T bits =
BitFieldInsert(BitFieldExtract(insn16, 2, 5), BitFieldExtract(insn16, 12, 1), 5, 1);
const T checked_bits = dchecked_integral_cast<T>(bits);
if (std::is_unsigned_v<T>) {
return checked_bits;
}
return SignExtendBits<6>(checked_bits);
}
// Regular instruction register utils
static uint32_t GetRd(uint32_t insn32) { return (insn32 >> 7) & 0x1fu; }
static uint32_t GetRs1(uint32_t insn32) { return (insn32 >> 15) & 0x1fu; }
static uint32_t GetRs2(uint32_t insn32) { return (insn32 >> 20) & 0x1fu; }
static uint32_t GetRs3(uint32_t insn32) { return insn32 >> 27; }
static uint32_t GetRoundingMode(uint32_t insn32) { return (insn32 >> 12) & 7u; }
// Compressed instruction register utils
static uint32_t GetRs1Short16(uint32_t insn16) { return BitFieldExtract(insn16, 7, 3) + 8u; }
static uint32_t GetRs2Short16(uint32_t insn16) { return BitFieldExtract(insn16, 2, 3) + 8u; }
static uint32_t GetRs1_16(uint32_t insn16) { return BitFieldExtract(insn16, 7, 5); }
static uint32_t GetRs2_16(uint32_t insn16) { return BitFieldExtract(insn16, 2, 5); }
void PrintBranchOffset(int32_t offset);
void PrintLoadStoreAddress(uint32_t rs1, int32_t offset);
void Print32Lui(uint32_t insn32);
void Print32Auipc(const uint8_t* insn, uint32_t insn32);
void Print32Jal(const uint8_t* insn, uint32_t insn32);
void Print32Jalr(const uint8_t* insn, uint32_t insn32);
void Print32BCond(const uint8_t* insn, uint32_t insn32);
void Print32Load(uint32_t insn32);
void Print32Store(uint32_t insn32);
void Print32FLoad(uint32_t insn32);
void Print32FStore(uint32_t insn32);
void Print32BinOpImm(uint32_t insn32);
void Print32BinOp(uint32_t insn32);
void Print32Atomic(uint32_t insn32);
void Print32FpOp(uint32_t insn32);
void Print32RVVOp(uint32_t insn32);
void Print32FpFma(uint32_t insn32);
void Print32Zicsr(uint32_t insn32);
void Print32Fence(uint32_t insn32);
void AppendVType(uint32_t zimm);
static const char* DecodeRVVMemMnemonic(const uint32_t insn32,
bool is_load,
/*out*/ const char** rs2);
DisassemblerRiscv64* const disassembler_;
std::ostream& os_;
};
const char* DisassemblerRiscv64::Printer::XRegName(uint32_t regno) {
static const char* const kXRegisterNames[] = {
"zero",
"ra",
"sp",
"gp",
"tp",
"t0",
"t1",
"t2",
"fp", // s0/fp
"tr", // s1/tr - ART thread register
"a0",
"a1",
"a2",
"a3",
"a4",
"a5",
"a6",
"a7",
"s2",
"s3",
"s4",
"s5",
"s6",
"s7",
"s8",
"s9",
"s10",
"s11",
"t3",
"t4",
"t5",
"t6",
};
static_assert(std::size(kXRegisterNames) == 32);
DCHECK_LT(regno, 32u);
return kXRegisterNames[regno];
}
const char* DisassemblerRiscv64::Printer::FRegName(uint32_t regno) {
static const char* const kFRegisterNames[] = {
"ft0",
"ft1",
"ft2",
"ft3",
"ft4",
"ft5",
"ft6",
"ft7",
"fs0",
"fs1",
"fa0",
"fa1",
"fa2",
"fa3",
"fa4",
"fa5",
"fa6",
"fa7",
"fs2",
"fs3",
"fs4",
"fs5",
"fs6",
"fs7",
"fs8",
"fs9",
"fs10",
"fs11",
"ft8",
"ft9",
"ft10",
"ft11",
};
static_assert(std::size(kFRegisterNames) == 32);
DCHECK_LT(regno, 32u);
return kFRegisterNames[regno];
}
const char* DisassemblerRiscv64::Printer::VRegName(uint32_t regno) {
static const char* const kVRegisterNames[] = {
"V0",
"V1",
"V2",
"V3",
"V4",
"V5",
"V6",
"V7",
"V8",
"V9",
"V10",
"V11",
"V12",
"V13",
"V14",
"V15",
"V16",
"V17",
"V18",
"V19",
"V20",
"V21",
"V22",
"V23",
"V24",
"V25",
"V26",
"V27",
"V28",
"V29",
"V30",
"V31",
};
static_assert(std::size(kVRegisterNames) == 32);
DCHECK_LT(regno, 32u);
return kVRegisterNames[regno];
}
const char* DisassemblerRiscv64::Printer::RoundingModeName(uint32_t rm) {
// Note: We do not print the rounding mode for DYN.
static const char* const kRoundingModeNames[] = {
".rne", ".rtz", ".rdn", ".rup", ".rmm", ".<reserved-rm>", ".<reserved-rm>", /*DYN*/ ""
};
static_assert(std::size(kRoundingModeNames) == 8);
DCHECK_LT(rm, 8u);
return kRoundingModeNames[rm];
}
void DisassemblerRiscv64::Printer::PrintBranchOffset(int32_t offset) {
os_ << (offset >= 0 ? "+" : "") << offset;
}
void DisassemblerRiscv64::Printer::PrintLoadStoreAddress(uint32_t rs1, int32_t offset) {
if (offset != 0) {
os_ << StringPrintf("%d", offset);
}
os_ << "(" << XRegName(rs1) << ")";
if (rs1 == TR && offset >= 0) {
// Add entrypoint name.
os_ << " ; ";
disassembler_->GetDisassemblerOptions()->thread_offset_name_function_(
os_, dchecked_integral_cast<uint32_t>(offset));
}
}
void DisassemblerRiscv64::Printer::Print32Lui(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x37u);
// TODO(riscv64): Should we also print the actual sign-extend value?
os_ << StringPrintf("lui %s, %u", XRegName(GetRd(insn32)), insn32 >> 12);
}
void DisassemblerRiscv64::Printer::Print32Auipc([[maybe_unused]] const uint8_t* insn,
uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x17u);
// TODO(riscv64): Should we also print the calculated address?
os_ << StringPrintf("auipc %s, %u", XRegName(GetRd(insn32)), insn32 >> 12);
}
void DisassemblerRiscv64::Printer::Print32Jal(const uint8_t* insn, uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x6fu);
// Print an alias if available.
uint32_t rd = GetRd(insn32);
os_ << (rd == Zero ? "j " : "jal ");
if (rd != Zero && rd != RA) {
os_ << XRegName(rd) << ", ";
}
uint32_t bit20 = (insn32 >> 31);
uint32_t bits1_10 = (insn32 >> 21) & 0x3ffu;
uint32_t bit11 = (insn32 >> 20) & 1u;
uint32_t bits12_19 = (insn32 >> 12) & 0xffu;
uint32_t imm = (bits1_10 << 1) + (bit11 << 11) + (bits12_19 << 12) + (bit20 << 20);
int32_t offset = static_cast<int32_t>(imm) - static_cast<int32_t>(bit20 << 21); // Sign-extend.
PrintBranchOffset(offset);
os_ << " ; " << disassembler_->FormatInstructionPointer(insn + offset);
// TODO(riscv64): When we implement shared thunks to reduce AOT slow-path code size,
// check if this JAL lands at an entrypoint load from TR and, if so, print its name.
}
void DisassemblerRiscv64::Printer::Print32Jalr([[maybe_unused]] const uint8_t* insn,
uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x67u);
DCHECK_EQ((insn32 >> 12) & 7u, 0u);
uint32_t rd = GetRd(insn32);
uint32_t rs1 = GetRs1(insn32);
int32_t imm12 = Decode32Imm12(insn32);
// Print shorter macro instruction notation if available.
if (rd == Zero && rs1 == RA && imm12 == 0) {
os_ << "ret";
} else if (rd == Zero && imm12 == 0) {
os_ << "jr " << XRegName(rs1);
} else if (rd == RA && imm12 == 0) {
os_ << "jalr " << XRegName(rs1);
} else {
// TODO(riscv64): Should we also print the calculated address if the preceding
// instruction is AUIPC? (We would need to record the previous instruction.)
os_ << "jalr " << XRegName(rd) << ", ";
// Use the same format as llvm-objdump: "rs1" if `imm12` is zero, otherwise "imm12(rs1)".
if (imm12 == 0) {
os_ << XRegName(rs1);
} else {
os_ << imm12 << "(" << XRegName(rs1) << ")";
}
}
}
void DisassemblerRiscv64::Printer::Print32BCond(const uint8_t* insn, uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x63u);
static const char* const kOpcodes[] = {
"beq", "bne", nullptr, nullptr, "blt", "bge", "bltu", "bgeu"
};
uint32_t funct3 = (insn32 >> 12) & 7u;
const char* opcode = kOpcodes[funct3];
if (opcode == nullptr) {
os_ << "<unknown32>";
return;
}
// Print shorter macro instruction notation if available.
uint32_t rs1 = GetRs1(insn32);
uint32_t rs2 = GetRs2(insn32);
if (rs2 == Zero) {
os_ << opcode << "z " << XRegName(rs1);
} else if (rs1 == Zero && (funct3 == 4u || funct3 == 5u)) {
// blt zero, rs2, offset ... bgtz rs2, offset
// bge zero, rs2, offset ... blez rs2, offset
os_ << (funct3 == 4u ? "bgtz " : "blez ") << XRegName(rs2);
} else {
os_ << opcode << " " << XRegName(rs1) << ", " << XRegName(rs2);
}
os_ << ", ";
uint32_t bit12 = insn32 >> 31;
uint32_t bits5_10 = (insn32 >> 25) & 0x3fu;
uint32_t bits1_4 = (insn32 >> 8) & 0xfu;
uint32_t bit11 = (insn32 >> 7) & 1u;
uint32_t imm = (bit12 << 12) + (bit11 << 11) + (bits5_10 << 5) + (bits1_4 << 1);
int32_t offset = static_cast<int32_t>(imm) - static_cast<int32_t>(bit12 << 13); // Sign-extend.
PrintBranchOffset(offset);
os_ << " ; " << disassembler_->FormatInstructionPointer(insn + offset);
}
void DisassemblerRiscv64::Printer::Print32Load(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x03u);
static const char* const kOpcodes[] = {
"lb", "lh", "lw", "ld", "lbu", "lhu", "lwu", nullptr
};
uint32_t funct3 = (insn32 >> 12) & 7u;
const char* opcode = kOpcodes[funct3];
if (opcode == nullptr) {
os_ << "<unknown32>";
return;
}
os_ << opcode << " " << XRegName(GetRd(insn32)) << ", ";
PrintLoadStoreAddress(GetRs1(insn32), Decode32Imm12(insn32));
// TODO(riscv64): If previous instruction is AUIPC for current `rs1` and we load
// from the range specified by assembler options, print the loaded literal.
}
void DisassemblerRiscv64::Printer::Print32Store(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x23u);
static const char* const kOpcodes[] = {
"sb", "sh", "sw", "sd", nullptr, nullptr, nullptr, nullptr
};
uint32_t funct3 = (insn32 >> 12) & 7u;
const char* opcode = kOpcodes[funct3];
if (opcode == nullptr) {
os_ << "<unknown32>";
return;
}
os_ << opcode << " " << XRegName(GetRs2(insn32)) << ", ";
PrintLoadStoreAddress(GetRs1(insn32), Decode32StoreOffset(insn32));
}
const char* DisassemblerRiscv64::Printer::DecodeRVVMemMnemonic(const uint32_t insn32,
bool is_load,
/*out*/ const char** rs2) {
const uint32_t width_index = (insn32 >> 12) & 3u;
DCHECK_EQ(width_index != 0u, (insn32 & 0x4000u) != 0u);
const uint32_t imm7 = Decode32UImm7(insn32);
const enum Nf nf = static_cast<enum Nf>((imm7 >> 4) & 0x7u);
const enum MemAddressMode mop = static_cast<enum MemAddressMode>((imm7 >> 1) & 0x3u);
const uint32_t mew = (insn32 >> 28) & 1u;
if (mew == 1u) {
// 7.3. Vector Load/Store Width Encoding
// The mew bit (inst[28]) when set is expected to be used to encode
// expanded memory sizes of 128 bits and above,
// but these encodings are currently reserved.
return nullptr;
}
switch (mop) {
case MemAddressMode::kUnitStride: {
const uint32_t umop = GetRs2(insn32);
switch (umop) {
case 0b00000: // Vector Unit-Stride Load/Store
static constexpr const char* kVUSMnemonics[8][4] = {
{"e8.v", "e16.v", "e32.v", "e64.v"},
{"seg2e8.v", "seg2e16.v", "seg2e32.v", "seg2e64.v"},
{"seg3e8.v", "seg3e16.v", "seg3e32.v", "seg3e64.v"},
{"seg4e8.v", "seg4e16.v", "seg4e32.v", "seg4e64.v"},
{"seg5e8.v", "seg5e16.v", "seg5e32.v", "seg5e64.v"},
{"seg6e8.v", "seg6e16.v", "seg6e32.v", "seg6e64.v"},
{"seg7e8.v", "seg7e16.v", "seg7e32.v", "seg7e64.v"},
{"seg8e8.v", "seg8e16.v", "seg8e32.v", "seg8e64.v"},
};
return kVUSMnemonics[enum_cast<uint32_t>(nf)][width_index];
case 0b01000: { // Vector Whole Register Load/Store
if (is_load) {
static constexpr const char* kVWRLMnemonics[8][4] = {
{"1re8.v", "1re16.v", "1re32.v", "1re64.v"},
{"2re8.v", "2re16.v", "2re32.v", "2re64.v"},
{nullptr, nullptr, nullptr, nullptr},
{"4re8.v", "4re16.v", "4re32.v", "4re64.v"},
{nullptr, nullptr, nullptr, nullptr},
{nullptr, nullptr, nullptr, nullptr},
{nullptr, nullptr, nullptr, nullptr},
{"8re8.v", "8re16.v", "8re32.v", "8re64.v"},
};
return kVWRLMnemonics[enum_cast<uint32_t>(nf)][width_index];
} else {
if (width_index != 0) {
return nullptr;
}
static constexpr const char* kVWRSMnemonics[8] = {
"1r", "2r", nullptr, "4r", nullptr, nullptr, nullptr, "8r"
};
return kVWRSMnemonics[enum_cast<uint32_t>(nf)];
}
}
case 0b01011: // Vector Unit-Stride Mask Load/Store
if (nf == Nf::k1 && width_index == 0 && (imm7 & 1u) == 1u) {
return "m.v";
} else {
return nullptr;
}
case 0b10000: // Vector Unit-Stride Fault-Only-First Load
static constexpr const char* kVUSFFLMnemonics[8][4] = {
{"e8ff.v", "e16ff.v", "e32ff.v", "e64ff.v"},
{"seg2e8ff.v", "seg2e16ff.v", "seg2e32ff.v", "seg2e64ff.v"},
{"seg3e8ff.v", "seg3e16ff.v", "seg3e32ff.v", "seg3e64ff.v"},
{"seg4e8ff.v", "seg4e16ff.v", "seg4e32ff.v", "seg4e64ff.v"},
{"seg5e8ff.v", "seg5e16ff.v", "seg5e32ff.v", "seg5e64ff.v"},
{"seg6e8ff.v", "seg6e16ff.v", "seg6e32ff.v", "seg6e64ff.v"},
{"seg7e8ff.v", "seg7e16ff.v", "seg7e32ff.v", "seg7e64ff.v"},
{"seg8e8ff.v", "seg8e16ff.v", "seg8e32ff.v", "seg8e64ff.v"},
};
return is_load ? kVUSFFLMnemonics[enum_cast<uint32_t>(nf)][width_index] : nullptr;
default: // Unknown
return nullptr;
}
}
case MemAddressMode::kIndexedUnordered: {
static constexpr const char* kVIUMnemonics[8][4] = {
{"uxei8.v", "uxei16.v", "uxei32.v", "uxei64.v"},
{"uxseg2ei8.v", "uxseg2ei16.v", "uxseg2ei32.v", "uxseg2ei64.v"},
{"uxseg3ei8.v", "uxseg3ei16.v", "uxseg3ei32.v", "uxseg3ei64.v"},
{"uxseg4ei8.v", "uxseg4ei16.v", "uxseg4ei32.v", "uxseg4ei64.v"},
{"uxseg5ei8.v", "uxseg5ei16.v", "uxseg5ei32.v", "uxseg5ei64.v"},
{"uxseg6ei8.v", "uxseg6ei16.v", "uxseg6ei32.v", "uxseg6ei64.v"},
{"uxseg7ei8.v", "uxseg7ei16.v", "uxseg7ei32.v", "uxseg7ei64.v"},
{"uxseg8ei8.v", "uxseg8ei16.v", "uxseg8ei32.v", "uxseg8ei64.v"},
};
*rs2 = VRegName(GetRs2(insn32));
return kVIUMnemonics[enum_cast<uint32_t>(nf)][width_index];
}
case MemAddressMode::kStrided: {
static constexpr const char* kVSMnemonics[8][4] = {
{"se8.v", "se16.v", "se32.v", "se64.v"},
{"sseg2e8.v", "sseg2e16.v", "sseg2e32.v", "sseg2e64.v"},
{"sseg3e8.v", "sseg3e16.v", "sseg3e32.v", "sseg3e64.v"},
{"sseg4e8.v", "sseg4e16.v", "sseg4e32.v", "sseg4e64.v"},
{"sseg5e8.v", "sseg5e16.v", "sseg5e32.v", "sseg5e64.v"},
{"sseg6e8.v", "sseg6e16.v", "sseg6e32.v", "sseg6e64.v"},
{"sseg7e8.v", "sseg7e16.v", "sseg7e32.v", "sseg7e64.v"},
{"sseg8e8.v", "sseg8e16.v", "sseg8e32.v", "sseg8e64.v"},
};
*rs2 = XRegName(GetRs2(insn32));
return kVSMnemonics[enum_cast<uint32_t>(nf)][width_index];
}
case MemAddressMode::kIndexedOrdered: {
static constexpr const char* kVIOMnemonics[8][4] = {
{"oxei8.v", "oxei16.v", "oxei32.v", "oxei64.v"},
{"oxseg2ei8.v", "oxseg2ei16.v", "oxseg2ei32.v", "oxseg2ei64.v"},
{"oxseg3ei8.v", "oxseg3ei16.v", "oxseg3ei32.v", "oxseg3ei64.v"},
{"oxseg4ei8.v", "oxseg4ei16.v", "oxseg4ei32.v", "oxseg4ei64.v"},
{"oxseg5ei8.v", "oxseg5ei16.v", "oxseg5ei32.v", "oxseg5ei64.v"},
{"oxseg6ei8.v", "oxseg6ei16.v", "oxseg6ei32.v", "oxseg6ei64.v"},
{"oxseg7ei8.v", "oxseg7ei16.v", "oxseg7ei32.v", "oxseg7ei64.v"},
{"oxseg8ei8.v", "oxseg8ei16.v", "oxseg8ei32.v", "oxseg8ei64.v"},
};
*rs2 = VRegName(GetRs2(insn32));
return kVIOMnemonics[enum_cast<uint32_t>(nf)][width_index];
}
}
}
static constexpr const char* kFpMemMnemonics[] = {
nullptr, "h", "w", "d", "q", nullptr, nullptr, nullptr
};
void DisassemblerRiscv64::Printer::Print32FLoad(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x07u);
int32_t offset = 0;
const char* rd = nullptr;
const char* rs2 = nullptr;
const char* vm = "";
const uint32_t funct3 = (insn32 >> 12) & 7u;
const char* mnemonic = kFpMemMnemonics[funct3];
const char* prefix = "f";
if (mnemonic == nullptr) {
// Vector Loads
prefix = "v";
mnemonic = DecodeRVVMemMnemonic(insn32, /*is_load=*/true, &rs2);
rd = VRegName(GetRd(insn32));
if ((Decode32UImm7(insn32) & 0x1U) == 0) {
vm = ", v0.t";
}
} else {
rd = FRegName(GetRd(insn32));
offset = Decode32Imm12(insn32);
}
if (mnemonic == nullptr) {
os_ << "<unknown32>";
return;
}
os_ << prefix << "l" << mnemonic << " " << rd << ", ";
PrintLoadStoreAddress(GetRs1(insn32), offset);
if (rs2) {
os_ << ", " << rs2;
}
os_ << vm;
// TODO(riscv64): If previous instruction is AUIPC for current `rs1` and we load
// from the range specified by assembler options, print the loaded literal.
}
void DisassemblerRiscv64::Printer::Print32FStore(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x27u);
uint32_t funct3 = (insn32 >> 12) & 3u;
const char* prefix = "f";
const char* mnemonic = kFpMemMnemonics[funct3];
if (mnemonic == nullptr) {
// Vector Stores
const char* rs2 = nullptr;
prefix = "v";
mnemonic = DecodeRVVMemMnemonic(insn32, /*is_load=*/false, &rs2);
if (mnemonic == nullptr) {
os_ << "<unknown32>";
return;
}
os_ << prefix << "s" << mnemonic << " " << VRegName(GetRd(insn32)) << ", ";
PrintLoadStoreAddress(GetRs1(insn32), 0);
if (rs2) {
os_ << ", " << rs2;
}
if ((Decode32UImm7(insn32) & 0x1U) == 0) {
os_ << ", v0.t";
}
} else {
os_ << prefix << "s" << mnemonic << " " << FRegName(GetRs2(insn32)) << ", ";
PrintLoadStoreAddress(GetRs1(insn32), Decode32StoreOffset(insn32));
}
}
void DisassemblerRiscv64::Printer::Print32BinOpImm(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x77u, 0x13u); // Note: Bit 0x8 selects narrow binop.
bool narrow = (insn32 & 0x8u) != 0u;
uint32_t funct3 = (insn32 >> 12) & 7u;
uint32_t rd = GetRd(insn32);
uint32_t rs1 = GetRs1(insn32);
int32_t imm = Decode32Imm12(insn32);
// Print shorter macro instruction notation if available.
if (funct3 == /*ADDI*/ 0u && imm == 0u) {
if (narrow) {
os_ << "sextw " << XRegName(rd) << ", " << XRegName(rs1);
} else if (rd == Zero && rs1 == Zero) {
os_ << "nop"; // Only canonical nop. Non-Zero `rd == rs1` nops are printed as "mv".
} else {
os_ << "mv " << XRegName(rd) << ", " << XRegName(rs1);
}
} else if (!narrow && funct3 == /*XORI*/ 4u && imm == -1) {
os_ << "not " << XRegName(rd) << ", " << XRegName(rs1);
} else if (!narrow && funct3 == /*ANDI*/ 7u && imm == 0xff) {
os_ << "zextb " << XRegName(rd) << ", " << XRegName(rs1);
} else if (!narrow && funct3 == /*SLTIU*/ 3u && imm == 1) {
os_ << "seqz " << XRegName(rd) << ", " << XRegName(rs1);
} else if ((insn32 & 0xfc00707fu) == 0x0800101bu) {
os_ << "slli.uw " << XRegName(rd) << ", " << XRegName(rs1) << ", " << (imm & 0x3fu);
} else if ((imm ^ 0x600u) < 3u && funct3 == 1u) {
static const char* const kBitOpcodes[] = { "clz", "ctz", "cpop" };
os_ << kBitOpcodes[imm ^ 0x600u] << (narrow ? "w " : " ")
<< XRegName(rd) << ", " << XRegName(rs1);
} else if ((imm ^ 0x600u) < (narrow ? 32 : 64) && funct3 == 5u) {
os_ << "rori" << (narrow ? "w " : " ")
<< XRegName(rd) << ", " << XRegName(rs1) << ", " << (imm ^ 0x600u);
} else if (imm == 0x287u && !narrow && funct3 == 5u) {
os_ << "orc.b " << XRegName(rd) << ", " << XRegName(rs1);
} else if (imm == 0x6b8u && !narrow && funct3 == 5u) {
os_ << "rev8 " << XRegName(rd) << ", " << XRegName(rs1);
} else {
bool bad_high_bits = false;
if (funct3 == /*SLLI*/ 1u || funct3 == /*SRLI/SRAI*/ 5u) {
imm &= (narrow ? 0x1fu : 0x3fu);
uint32_t high_bits = insn32 & (narrow ? 0xfe000000u : 0xfc000000u);
if (high_bits == 0x40000000u && funct3 == /*SRAI*/ 5u) {
os_ << "srai";
} else {
os_ << ((funct3 == /*SRLI*/ 5u) ? "srli" : "slli");
bad_high_bits = (high_bits != 0u);
}
} else if (!narrow || funct3 == /*ADDI*/ 0u) {
static const char* const kOpcodes[] = {
"addi", nullptr, "slti", "sltiu", "xori", nullptr, "ori", "andi"
};
DCHECK(kOpcodes[funct3] != nullptr);
os_ << kOpcodes[funct3];
} else {
os_ << "<unknown32>"; // There is no SLTIW/SLTIUW/XORIW/ORIW/ANDIW.
return;
}
os_ << (narrow ? "w " : " ") << XRegName(rd) << ", " << XRegName(rs1) << ", " << imm;
if (bad_high_bits) {
os_ << " (invalid high bits)";
}
}
}
void DisassemblerRiscv64::Printer::Print32BinOp(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x77u, 0x33u); // Note: Bit 0x8 selects narrow binop.
bool narrow = (insn32 & 0x8u) != 0u;
uint32_t funct3 = (insn32 >> 12) & 7u;
uint32_t rd = GetRd(insn32);
uint32_t rs1 = GetRs1(insn32);
uint32_t rs2 = GetRs2(insn32);
uint32_t high_bits = insn32 & 0xfe000000u;
// Print shorter macro instruction notation if available.
if (high_bits == 0x40000000u && funct3 == /*SUB*/ 0u && rs1 == Zero) {
os_ << (narrow ? "negw " : "neg ") << XRegName(rd) << ", " << XRegName(rs2);
} else if (!narrow && funct3 == /*SLT*/ 2u && rs2 == Zero) {
os_ << "sltz " << XRegName(rd) << ", " << XRegName(rs1);
} else if (!narrow && funct3 == /*SLT*/ 2u && rs1 == Zero) {
os_ << "sgtz " << XRegName(rd) << ", " << XRegName(rs2);
} else if (!narrow && funct3 == /*SLTU*/ 3u && rs1 == Zero) {
os_ << "snez " << XRegName(rd) << ", " << XRegName(rs2);
} else if (narrow && high_bits == 0x08000000u && funct3 == /*ADD.UW*/ 0u && rs2 == Zero) {
os_ << "zext.w " << XRegName(rd) << ", " << XRegName(rs1);
} else {
bool bad_high_bits = false;
if (high_bits == 0x40000000u && (funct3 == /*SUB*/ 0u || funct3 == /*SRA*/ 5u)) {
os_ << ((funct3 == /*SUB*/ 0u) ? "sub" : "sra");
} else if (high_bits == 0x02000000u &&
(!narrow || (funct3 == /*MUL*/ 0u || funct3 >= /*DIV/DIVU/REM/REMU*/ 4u))) {
static const char* const kOpcodes[] = {
"mul", "mulh", "mulhsu", "mulhu", "div", "divu", "rem", "remu"
};
os_ << kOpcodes[funct3];
} else if (high_bits == 0x08000000u && narrow && funct3 == /*ADD.UW*/ 0u) {
os_ << "add.u"; // "w" is added below.
} else if (high_bits == 0x20000000u && (funct3 & 1u) == 0u && funct3 != 0u) {
static const char* const kZbaOpcodes[] = { nullptr, "sh1add", "sh2add", "sh3add" };
DCHECK(kZbaOpcodes[funct3 >> 1] != nullptr);
os_ << kZbaOpcodes[funct3 >> 1] << (narrow ? ".u" /* "w" is added below. */ : "");
} else if (high_bits == 0x40000000u && !narrow && funct3 >= 4u && funct3 != 5u) {
static const char* const kZbbNegOpcodes[] = { "xnor", nullptr, "orn", "andn" };
DCHECK(kZbbNegOpcodes[funct3 - 4u] != nullptr);
os_ << kZbbNegOpcodes[funct3 - 4u];
} else if (high_bits == 0x0a000000u && !narrow && funct3 >= 4u) {
static const char* const kZbbMinMaxOpcodes[] = { "min", "minu", "max", "maxu" };
DCHECK(kZbbMinMaxOpcodes[funct3 - 4u] != nullptr);
os_ << kZbbMinMaxOpcodes[funct3 - 4u];
} else if (high_bits == 0x60000000u && (funct3 == /*ROL*/ 1u || funct3 == /*ROL*/ 5u)) {
os_ << (funct3 == /*ROL*/ 1u ? "rol" : "ror");
} else if (!narrow || (funct3 == /*ADD*/ 0u || funct3 == /*SLL*/ 1u || funct3 == /*SRL*/ 5u)) {
static const char* const kOpcodes[] = {
"add", "sll", "slt", "sltu", "xor", "srl", "or", "and"
};
os_ << kOpcodes[funct3];
bad_high_bits = (high_bits != 0u);
} else {
DCHECK(narrow);
os_ << "<unknown32>"; // Some of the above instructions do not have a narrow version.
return;
}
os_ << (narrow ? "w " : " ") << XRegName(rd) << ", " << XRegName(rs1) << ", " << XRegName(rs2);
if (bad_high_bits) {
os_ << " (invalid high bits)";
}
}
}
void DisassemblerRiscv64::Printer::Print32Atomic(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x2fu);
uint32_t funct3 = (insn32 >> 12) & 7u;
uint32_t funct5 = (insn32 >> 27);
if ((funct3 != 2u && funct3 != 3u) || // There are only 32-bit and 64-bit LR/SC/AMO*.
(((funct5 & 3u) != 0u) && funct5 >= 4u)) { // Only multiples of 4, or 1-3.
os_ << "<unknown32>";
return;
}
static const char* const kMul4Opcodes[] = {
"amoadd", "amoxor", "amoor", "amoand", "amomin", "amomax", "amominu", "amomaxu"
};
static const char* const kOtherOpcodes[] = {
nullptr, "amoswap", "lr", "sc"
};
const char* opcode = ((funct5 & 3u) == 0u) ? kMul4Opcodes[funct5 >> 2] : kOtherOpcodes[funct5];
DCHECK(opcode != nullptr);
uint32_t rd = GetRd(insn32);
uint32_t rs1 = GetRs1(insn32);
uint32_t rs2 = GetRs2(insn32);
const char* type = (funct3 == 2u) ? ".w" : ".d";
const char* aq = (((insn32 >> 26) & 1u) != 0u) ? ".aq" : "";
const char* rl = (((insn32 >> 25) & 1u) != 0u) ? ".rl" : "";
os_ << opcode << type << aq << rl << " " << XRegName(rd) << ", " << XRegName(rs1);
if (funct5 == /*LR*/ 2u) {
if (rs2 != 0u) {
os_ << " (bad rs2)";
}
} else {
os_ << ", " << XRegName(rs2);
}
}
void DisassemblerRiscv64::Printer::Print32FpOp(uint32_t insn32) {
DCHECK_EQ(insn32 & 0x7fu, 0x53u);
uint32_t rd = GetRd(insn32);
uint32_t rs1 = GetRs1(insn32);
uint32_t rs2 = GetRs2(insn32); // Sometimes used to to differentiate opcodes.
uint32_t rm = GetRoundingMode(insn32); // Sometimes used to to differentiate opcodes.
uint32_t funct7 = insn32 >> 25;
const char* type = ((funct7 & 1u) != 0u) ? ".d" : ".s";
if ((funct7 & 2u) != 0u) {
os_ << "<unknown32>"; // Note: This includes the "H" and "Q" extensions.
return;
}
switch (funct7 >> 2) {
case 0u:
case 1u:
case 2u:
case 3u: {
static const char* const kOpcodes[] = { "fadd", "fsub", "fmul", "fdiv" };
os_ << kOpcodes[funct7 >> 2] << type << RoundingModeName(rm) << " "
<< FRegName(rd) << ", " << FRegName(rs1) << ", " << FRegName(rs2);
return;
}
case 4u: { // FSGN*
// Print shorter macro instruction notation if available.
static const char* const kOpcodes[] = { "fsgnj", "fsgnjn", "fsgnjx" };
if (rm < std::size(kOpcodes)) {
if (rs1 == rs2) {
static const char* const kAltOpcodes[] = { "fmv", "fneg", "fabs" };
static_assert(std::size(kOpcodes) == std::size(kAltOpcodes));
os_ << kAltOpcodes[rm] << type << " " << FRegName(rd) << ", " << FRegName(rs1);
} else {
os_ << kOpcodes[rm] << type << " "
<< FRegName(rd) << ", " << FRegName(rs1) << ", " << FRegName(rs2);
}
return;
}
break;
}
case 5u: { // FMIN/FMAX
static const char* const kOpcodes[] = { "fmin", "fmax" };
if (rm < std::size(kOpcodes)) {
os_ << kOpcodes[rm] << type << " "
<< FRegName(rd) << ", " << FRegName(rs1) << ", " << FRegName(rs2);
return;
}
break;
}
case 0x8u: // FCVT between FP numbers.
if ((rs2 ^ 1u) == (funct7 & 1u)) {
os_ << ((rs2 != 0u) ? "fcvt.s.d" : "fcvt.d.s") << RoundingModeName(rm) << " "
<< FRegName(rd) << ", " << FRegName(rs1);
}
break;
case 0xbu:
if (rs2 == 0u) {
os_ << "fsqrt" << type << RoundingModeName(rm) << " "
<< FRegName(rd) << ", " << FRegName(rs1);
return;
}
break;
case 0x14u: { // FLE/FLT/FEQ
static const char* const kOpcodes[] = { "fle", "flt", "feq" };
if (rm < std::size(kOpcodes)) {
os_ << kOpcodes[rm] << type << " "
<< XRegName(rd) << ", " << FRegName(rs1) << ", " << FRegName(rs2);
return;
}
break;
}
case 0x18u: { // FCVT from floating point numbers to integers
static const char* const kIntTypes[] = { "w", "wu", "l", "lu" };
if (rs2 < std::size(kIntTypes)) {
os_ << "fcvt." << kIntTypes[rs2] << type << RoundingModeName(rm) << " "
<< XRegName(rd) << ", " << FRegName(rs1);
return;
}
break;
}
case 0x1au: { // FCVT from integers to floating point numbers
static const char* const kIntTypes[] = { "w", "wu", "l", "lu" };
if (rs2 < std::size(kIntTypes)) {
os_ << "fcvt" << type << "." << kIntTypes[rs2] << RoundingModeName(rm) << " "
<< FRegName(rd) << ", " << XRegName(rs1);
return;
}
break;
}
case 0x1cu: // FMV from FPR to GPR, or FCLASS
if (rs2 == 0u && rm == 0u) {
os_ << (((funct7 & 1u) != 0u) ? "fmv.x.d " : "fmv.x.w ")
<< XRegName(rd) << ", " << FRegName(rs1);
return;
} else if (rs2 == 0u && rm == 1u) {
os_ << "fclass" << type << " " << XRegName(rd) << ", " << FRegName(rs1);
return;
}
break;
case 0x1eu: // FMV from GPR to FPR
if (rs2 == 0u && rm == 0u) {
os_ << (((funct7 & 1u) != 0u) ? "fmv.d.x " : "fmv.w.x ")
<< FRegName(rd) << ", " << XRegName(rs1);
return;
}
break;
default:
break;
}
os_ << "<unknown32>";
}
void DisassemblerRiscv64::Printer::AppendVType(uint32_t vtype) {
const uint32_t lmul_v = vtype & 0x7U;
const uint32_t vsew_v = (vtype >> 3) & 0x7U;
const uint32_t vta_v = (vtype >> 6) & 0x1U;
const uint32_t vma_v = (vtype >> 7) & 0x1U;
if ((vsew_v & 0x4U) == 0u) {
if (lmul_v != 0b100) {
static const char* const vsews[] = {"e8", "e16", "e32", "e64"};
static const char* const lmuls[] = {
"m1", "m2", "m4", "m8", nullptr, "mf8", "mf4", "mf2"
};
const char* vma = vma_v ? "ma" : "mu";
const char* vta = vta_v ? "ta" : "tu";
const char* vsew = vsews[vsew_v & 0x3u];
const char* lmul = lmuls[lmul_v];
os_ << vsew << ", " << lmul << ", " << vta << ", " << vma;
return;
}
}