forked from llvm/llvm-project
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathX86InstructionSelector.cpp
1693 lines (1431 loc) · 59.4 KB
/
X86InstructionSelector.cpp
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
//===- X86InstructionSelector.cpp -----------------------------------------===//
//
// 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 implements the targeting of the InstructionSelector class for
/// X86.
/// \todo This should be generated by TableGen.
//===----------------------------------------------------------------------===//
#include "MCTargetDesc/X86BaseInfo.h"
#include "X86.h"
#include "X86InstrBuilder.h"
#include "X86InstrInfo.h"
#include "X86RegisterBankInfo.h"
#include "X86RegisterInfo.h"
#include "X86Subtarget.h"
#include "X86TargetMachine.h"
#include "llvm/CodeGen/GlobalISel/InstructionSelector.h"
#include "llvm/CodeGen/GlobalISel/InstructionSelectorImpl.h"
#include "llvm/CodeGen/GlobalISel/RegisterBank.h"
#include "llvm/CodeGen/GlobalISel/Utils.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/TargetOpcodes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/IntrinsicsX86.h"
#include "llvm/Support/AtomicOrdering.h"
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/LowLevelTypeImpl.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include <cassert>
#include <cstdint>
#include <tuple>
#define DEBUG_TYPE "X86-isel"
using namespace llvm;
namespace {
#define GET_GLOBALISEL_PREDICATE_BITSET
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_PREDICATE_BITSET
class X86InstructionSelector : public InstructionSelector {
public:
X86InstructionSelector(const X86TargetMachine &TM, const X86Subtarget &STI,
const X86RegisterBankInfo &RBI);
bool select(MachineInstr &I) override;
static const char *getName() { return DEBUG_TYPE; }
private:
/// tblgen-erated 'select' implementation, used as the initial selector for
/// the patterns that don't require complex C++.
bool selectImpl(MachineInstr &I, CodeGenCoverage &CoverageInfo) const;
// TODO: remove after supported by Tablegen-erated instruction selection.
unsigned getLoadStoreOp(const LLT &Ty, const RegisterBank &RB, unsigned Opc,
Align Alignment) const;
bool selectLoadStoreOp(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectFrameIndexOrGep(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectGlobalValue(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectConstant(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectTruncOrPtrToInt(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectZext(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectAnyext(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectCmp(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectFCmp(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectUadde(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectCopy(MachineInstr &I, MachineRegisterInfo &MRI) const;
bool selectUnmergeValues(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF);
bool selectMergeValues(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF);
bool selectInsert(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectExtract(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectCondBranch(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectTurnIntoCOPY(MachineInstr &I, MachineRegisterInfo &MRI,
const unsigned DstReg,
const TargetRegisterClass *DstRC,
const unsigned SrcReg,
const TargetRegisterClass *SrcRC) const;
bool materializeFP(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectImplicitDefOrPHI(MachineInstr &I, MachineRegisterInfo &MRI) const;
bool selectDivRem(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
bool selectIntrinsicWSideEffects(MachineInstr &I, MachineRegisterInfo &MRI,
MachineFunction &MF) const;
// emit insert subreg instruction and insert it before MachineInstr &I
bool emitInsertSubreg(unsigned DstReg, unsigned SrcReg, MachineInstr &I,
MachineRegisterInfo &MRI, MachineFunction &MF) const;
// emit extract subreg instruction and insert it before MachineInstr &I
bool emitExtractSubreg(unsigned DstReg, unsigned SrcReg, MachineInstr &I,
MachineRegisterInfo &MRI, MachineFunction &MF) const;
const TargetRegisterClass *getRegClass(LLT Ty, const RegisterBank &RB) const;
const TargetRegisterClass *getRegClass(LLT Ty, unsigned Reg,
MachineRegisterInfo &MRI) const;
const X86TargetMachine &TM;
const X86Subtarget &STI;
const X86InstrInfo &TII;
const X86RegisterInfo &TRI;
const X86RegisterBankInfo &RBI;
#define GET_GLOBALISEL_PREDICATES_DECL
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_PREDICATES_DECL
#define GET_GLOBALISEL_TEMPORARIES_DECL
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_TEMPORARIES_DECL
};
} // end anonymous namespace
#define GET_GLOBALISEL_IMPL
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_IMPL
X86InstructionSelector::X86InstructionSelector(const X86TargetMachine &TM,
const X86Subtarget &STI,
const X86RegisterBankInfo &RBI)
: TM(TM), STI(STI), TII(*STI.getInstrInfo()), TRI(*STI.getRegisterInfo()),
RBI(RBI),
#define GET_GLOBALISEL_PREDICATES_INIT
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_PREDICATES_INIT
#define GET_GLOBALISEL_TEMPORARIES_INIT
#include "X86GenGlobalISel.inc"
#undef GET_GLOBALISEL_TEMPORARIES_INIT
{
}
// FIXME: This should be target-independent, inferred from the types declared
// for each class in the bank.
const TargetRegisterClass *
X86InstructionSelector::getRegClass(LLT Ty, const RegisterBank &RB) const {
if (RB.getID() == X86::GPRRegBankID) {
if (Ty.getSizeInBits() <= 8)
return &X86::GR8RegClass;
if (Ty.getSizeInBits() == 16)
return &X86::GR16RegClass;
if (Ty.getSizeInBits() == 32)
return &X86::GR32RegClass;
if (Ty.getSizeInBits() == 64)
return &X86::GR64RegClass;
}
if (RB.getID() == X86::VECRRegBankID) {
if (Ty.getSizeInBits() == 32)
return STI.hasAVX512() ? &X86::FR32XRegClass : &X86::FR32RegClass;
if (Ty.getSizeInBits() == 64)
return STI.hasAVX512() ? &X86::FR64XRegClass : &X86::FR64RegClass;
if (Ty.getSizeInBits() == 128)
return STI.hasAVX512() ? &X86::VR128XRegClass : &X86::VR128RegClass;
if (Ty.getSizeInBits() == 256)
return STI.hasAVX512() ? &X86::VR256XRegClass : &X86::VR256RegClass;
if (Ty.getSizeInBits() == 512)
return &X86::VR512RegClass;
}
llvm_unreachable("Unknown RegBank!");
}
const TargetRegisterClass *
X86InstructionSelector::getRegClass(LLT Ty, unsigned Reg,
MachineRegisterInfo &MRI) const {
const RegisterBank &RegBank = *RBI.getRegBank(Reg, MRI, TRI);
return getRegClass(Ty, RegBank);
}
static unsigned getSubRegIndex(const TargetRegisterClass *RC) {
unsigned SubIdx = X86::NoSubRegister;
if (RC == &X86::GR32RegClass) {
SubIdx = X86::sub_32bit;
} else if (RC == &X86::GR16RegClass) {
SubIdx = X86::sub_16bit;
} else if (RC == &X86::GR8RegClass) {
SubIdx = X86::sub_8bit;
}
return SubIdx;
}
static const TargetRegisterClass *getRegClassFromGRPhysReg(Register Reg) {
assert(Reg.isPhysical());
if (X86::GR64RegClass.contains(Reg))
return &X86::GR64RegClass;
if (X86::GR32RegClass.contains(Reg))
return &X86::GR32RegClass;
if (X86::GR16RegClass.contains(Reg))
return &X86::GR16RegClass;
if (X86::GR8RegClass.contains(Reg))
return &X86::GR8RegClass;
llvm_unreachable("Unknown RegClass for PhysReg!");
}
// Set X86 Opcode and constrain DestReg.
bool X86InstructionSelector::selectCopy(MachineInstr &I,
MachineRegisterInfo &MRI) const {
Register DstReg = I.getOperand(0).getReg();
const unsigned DstSize = RBI.getSizeInBits(DstReg, MRI, TRI);
const RegisterBank &DstRegBank = *RBI.getRegBank(DstReg, MRI, TRI);
Register SrcReg = I.getOperand(1).getReg();
const unsigned SrcSize = RBI.getSizeInBits(SrcReg, MRI, TRI);
const RegisterBank &SrcRegBank = *RBI.getRegBank(SrcReg, MRI, TRI);
if (DstReg.isPhysical()) {
assert(I.isCopy() && "Generic operators do not allow physical registers");
if (DstSize > SrcSize && SrcRegBank.getID() == X86::GPRRegBankID &&
DstRegBank.getID() == X86::GPRRegBankID) {
const TargetRegisterClass *SrcRC =
getRegClass(MRI.getType(SrcReg), SrcRegBank);
const TargetRegisterClass *DstRC = getRegClassFromGRPhysReg(DstReg);
if (SrcRC != DstRC) {
// This case can be generated by ABI lowering, performe anyext
Register ExtSrc = MRI.createVirtualRegister(DstRC);
BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(TargetOpcode::SUBREG_TO_REG))
.addDef(ExtSrc)
.addImm(0)
.addReg(SrcReg)
.addImm(getSubRegIndex(SrcRC));
I.getOperand(1).setReg(ExtSrc);
}
}
return true;
}
assert((!SrcReg.isPhysical() || I.isCopy()) &&
"No phys reg on generic operators");
assert((DstSize == SrcSize ||
// Copies are a mean to setup initial types, the number of
// bits may not exactly match.
(SrcReg.isPhysical() &&
DstSize <= RBI.getSizeInBits(SrcReg, MRI, TRI))) &&
"Copy with different width?!");
const TargetRegisterClass *DstRC =
getRegClass(MRI.getType(DstReg), DstRegBank);
if (SrcRegBank.getID() == X86::GPRRegBankID &&
DstRegBank.getID() == X86::GPRRegBankID && SrcSize > DstSize &&
SrcReg.isPhysical()) {
// Change the physical register to performe truncate.
const TargetRegisterClass *SrcRC = getRegClassFromGRPhysReg(SrcReg);
if (DstRC != SrcRC) {
I.getOperand(1).setSubReg(getSubRegIndex(DstRC));
I.getOperand(1).substPhysReg(SrcReg, TRI);
}
}
// No need to constrain SrcReg. It will get constrained when
// we hit another of its use or its defs.
// Copies do not have constraints.
const TargetRegisterClass *OldRC = MRI.getRegClassOrNull(DstReg);
if (!OldRC || !DstRC->hasSubClassEq(OldRC)) {
if (!RBI.constrainGenericRegister(DstReg, *DstRC, MRI)) {
LLVM_DEBUG(dbgs() << "Failed to constrain " << TII.getName(I.getOpcode())
<< " operand\n");
return false;
}
}
I.setDesc(TII.get(X86::COPY));
return true;
}
bool X86InstructionSelector::select(MachineInstr &I) {
assert(I.getParent() && "Instruction should be in a basic block!");
assert(I.getParent()->getParent() && "Instruction should be in a function!");
MachineBasicBlock &MBB = *I.getParent();
MachineFunction &MF = *MBB.getParent();
MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned Opcode = I.getOpcode();
if (!isPreISelGenericOpcode(Opcode)) {
// Certain non-generic instructions also need some special handling.
if (Opcode == TargetOpcode::LOAD_STACK_GUARD)
return false;
if (I.isCopy())
return selectCopy(I, MRI);
return true;
}
assert(I.getNumOperands() == I.getNumExplicitOperands() &&
"Generic instruction has unexpected implicit operands\n");
if (selectImpl(I, *CoverageInfo))
return true;
LLVM_DEBUG(dbgs() << " C++ instruction selection: "; I.print(dbgs()));
// TODO: This should be implemented by tblgen.
switch (I.getOpcode()) {
default:
return false;
case TargetOpcode::G_STORE:
case TargetOpcode::G_LOAD:
return selectLoadStoreOp(I, MRI, MF);
case TargetOpcode::G_PTR_ADD:
case TargetOpcode::G_FRAME_INDEX:
return selectFrameIndexOrGep(I, MRI, MF);
case TargetOpcode::G_GLOBAL_VALUE:
return selectGlobalValue(I, MRI, MF);
case TargetOpcode::G_CONSTANT:
return selectConstant(I, MRI, MF);
case TargetOpcode::G_FCONSTANT:
return materializeFP(I, MRI, MF);
case TargetOpcode::G_PTRTOINT:
case TargetOpcode::G_TRUNC:
return selectTruncOrPtrToInt(I, MRI, MF);
case TargetOpcode::G_INTTOPTR:
return selectCopy(I, MRI);
case TargetOpcode::G_ZEXT:
return selectZext(I, MRI, MF);
case TargetOpcode::G_ANYEXT:
return selectAnyext(I, MRI, MF);
case TargetOpcode::G_ICMP:
return selectCmp(I, MRI, MF);
case TargetOpcode::G_FCMP:
return selectFCmp(I, MRI, MF);
case TargetOpcode::G_UADDE:
return selectUadde(I, MRI, MF);
case TargetOpcode::G_UNMERGE_VALUES:
return selectUnmergeValues(I, MRI, MF);
case TargetOpcode::G_MERGE_VALUES:
case TargetOpcode::G_CONCAT_VECTORS:
return selectMergeValues(I, MRI, MF);
case TargetOpcode::G_EXTRACT:
return selectExtract(I, MRI, MF);
case TargetOpcode::G_INSERT:
return selectInsert(I, MRI, MF);
case TargetOpcode::G_BRCOND:
return selectCondBranch(I, MRI, MF);
case TargetOpcode::G_IMPLICIT_DEF:
case TargetOpcode::G_PHI:
return selectImplicitDefOrPHI(I, MRI);
case TargetOpcode::G_SDIV:
case TargetOpcode::G_UDIV:
case TargetOpcode::G_SREM:
case TargetOpcode::G_UREM:
return selectDivRem(I, MRI, MF);
case TargetOpcode::G_INTRINSIC_W_SIDE_EFFECTS:
return selectIntrinsicWSideEffects(I, MRI, MF);
}
return false;
}
unsigned X86InstructionSelector::getLoadStoreOp(const LLT &Ty,
const RegisterBank &RB,
unsigned Opc,
Align Alignment) const {
bool Isload = (Opc == TargetOpcode::G_LOAD);
bool HasAVX = STI.hasAVX();
bool HasAVX512 = STI.hasAVX512();
bool HasVLX = STI.hasVLX();
if (Ty == LLT::scalar(8)) {
if (X86::GPRRegBankID == RB.getID())
return Isload ? X86::MOV8rm : X86::MOV8mr;
} else if (Ty == LLT::scalar(16)) {
if (X86::GPRRegBankID == RB.getID())
return Isload ? X86::MOV16rm : X86::MOV16mr;
} else if (Ty == LLT::scalar(32) || Ty == LLT::pointer(0, 32)) {
if (X86::GPRRegBankID == RB.getID())
return Isload ? X86::MOV32rm : X86::MOV32mr;
if (X86::VECRRegBankID == RB.getID())
return Isload ? (HasAVX512 ? X86::VMOVSSZrm_alt :
HasAVX ? X86::VMOVSSrm_alt :
X86::MOVSSrm_alt)
: (HasAVX512 ? X86::VMOVSSZmr :
HasAVX ? X86::VMOVSSmr :
X86::MOVSSmr);
} else if (Ty == LLT::scalar(64) || Ty == LLT::pointer(0, 64)) {
if (X86::GPRRegBankID == RB.getID())
return Isload ? X86::MOV64rm : X86::MOV64mr;
if (X86::VECRRegBankID == RB.getID())
return Isload ? (HasAVX512 ? X86::VMOVSDZrm_alt :
HasAVX ? X86::VMOVSDrm_alt :
X86::MOVSDrm_alt)
: (HasAVX512 ? X86::VMOVSDZmr :
HasAVX ? X86::VMOVSDmr :
X86::MOVSDmr);
} else if (Ty.isVector() && Ty.getSizeInBits() == 128) {
if (Alignment >= Align(16))
return Isload ? (HasVLX ? X86::VMOVAPSZ128rm
: HasAVX512
? X86::VMOVAPSZ128rm_NOVLX
: HasAVX ? X86::VMOVAPSrm : X86::MOVAPSrm)
: (HasVLX ? X86::VMOVAPSZ128mr
: HasAVX512
? X86::VMOVAPSZ128mr_NOVLX
: HasAVX ? X86::VMOVAPSmr : X86::MOVAPSmr);
else
return Isload ? (HasVLX ? X86::VMOVUPSZ128rm
: HasAVX512
? X86::VMOVUPSZ128rm_NOVLX
: HasAVX ? X86::VMOVUPSrm : X86::MOVUPSrm)
: (HasVLX ? X86::VMOVUPSZ128mr
: HasAVX512
? X86::VMOVUPSZ128mr_NOVLX
: HasAVX ? X86::VMOVUPSmr : X86::MOVUPSmr);
} else if (Ty.isVector() && Ty.getSizeInBits() == 256) {
if (Alignment >= Align(32))
return Isload ? (HasVLX ? X86::VMOVAPSZ256rm
: HasAVX512 ? X86::VMOVAPSZ256rm_NOVLX
: X86::VMOVAPSYrm)
: (HasVLX ? X86::VMOVAPSZ256mr
: HasAVX512 ? X86::VMOVAPSZ256mr_NOVLX
: X86::VMOVAPSYmr);
else
return Isload ? (HasVLX ? X86::VMOVUPSZ256rm
: HasAVX512 ? X86::VMOVUPSZ256rm_NOVLX
: X86::VMOVUPSYrm)
: (HasVLX ? X86::VMOVUPSZ256mr
: HasAVX512 ? X86::VMOVUPSZ256mr_NOVLX
: X86::VMOVUPSYmr);
} else if (Ty.isVector() && Ty.getSizeInBits() == 512) {
if (Alignment >= Align(64))
return Isload ? X86::VMOVAPSZrm : X86::VMOVAPSZmr;
else
return Isload ? X86::VMOVUPSZrm : X86::VMOVUPSZmr;
}
return Opc;
}
// Fill in an address from the given instruction.
static void X86SelectAddress(const MachineInstr &I,
const MachineRegisterInfo &MRI,
X86AddressMode &AM) {
assert(I.getOperand(0).isReg() && "unsupported opperand.");
assert(MRI.getType(I.getOperand(0).getReg()).isPointer() &&
"unsupported type.");
if (I.getOpcode() == TargetOpcode::G_PTR_ADD) {
if (auto COff = getIConstantVRegSExtVal(I.getOperand(2).getReg(), MRI)) {
int64_t Imm = *COff;
if (isInt<32>(Imm)) { // Check for displacement overflow.
AM.Disp = static_cast<int32_t>(Imm);
AM.Base.Reg = I.getOperand(1).getReg();
return;
}
}
} else if (I.getOpcode() == TargetOpcode::G_FRAME_INDEX) {
AM.Base.FrameIndex = I.getOperand(1).getIndex();
AM.BaseType = X86AddressMode::FrameIndexBase;
return;
}
// Default behavior.
AM.Base.Reg = I.getOperand(0).getReg();
}
bool X86InstructionSelector::selectLoadStoreOp(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
unsigned Opc = I.getOpcode();
assert((Opc == TargetOpcode::G_STORE || Opc == TargetOpcode::G_LOAD) &&
"unexpected instruction");
const Register DefReg = I.getOperand(0).getReg();
LLT Ty = MRI.getType(DefReg);
const RegisterBank &RB = *RBI.getRegBank(DefReg, MRI, TRI);
assert(I.hasOneMemOperand());
auto &MemOp = **I.memoperands_begin();
if (MemOp.isAtomic()) {
// Note: for unordered operations, we rely on the fact the appropriate MMO
// is already on the instruction we're mutating, and thus we don't need to
// make any changes. So long as we select an opcode which is capable of
// loading or storing the appropriate size atomically, the rest of the
// backend is required to respect the MMO state.
if (!MemOp.isUnordered()) {
LLVM_DEBUG(dbgs() << "Atomic ordering not supported yet\n");
return false;
}
if (MemOp.getAlign() < Ty.getSizeInBits() / 8) {
LLVM_DEBUG(dbgs() << "Unaligned atomics not supported yet\n");
return false;
}
}
unsigned NewOpc = getLoadStoreOp(Ty, RB, Opc, MemOp.getAlign());
if (NewOpc == Opc)
return false;
X86AddressMode AM;
X86SelectAddress(*MRI.getVRegDef(I.getOperand(1).getReg()), MRI, AM);
I.setDesc(TII.get(NewOpc));
MachineInstrBuilder MIB(MF, I);
if (Opc == TargetOpcode::G_LOAD) {
I.RemoveOperand(1);
addFullAddress(MIB, AM);
} else {
// G_STORE (VAL, Addr), X86Store instruction (Addr, VAL)
I.RemoveOperand(1);
I.RemoveOperand(0);
addFullAddress(MIB, AM).addUse(DefReg);
}
return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
}
static unsigned getLeaOP(LLT Ty, const X86Subtarget &STI) {
if (Ty == LLT::pointer(0, 64))
return X86::LEA64r;
else if (Ty == LLT::pointer(0, 32))
return STI.isTarget64BitILP32() ? X86::LEA64_32r : X86::LEA32r;
else
llvm_unreachable("Can't get LEA opcode. Unsupported type.");
}
bool X86InstructionSelector::selectFrameIndexOrGep(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
unsigned Opc = I.getOpcode();
assert((Opc == TargetOpcode::G_FRAME_INDEX || Opc == TargetOpcode::G_PTR_ADD) &&
"unexpected instruction");
const Register DefReg = I.getOperand(0).getReg();
LLT Ty = MRI.getType(DefReg);
// Use LEA to calculate frame index and GEP
unsigned NewOpc = getLeaOP(Ty, STI);
I.setDesc(TII.get(NewOpc));
MachineInstrBuilder MIB(MF, I);
if (Opc == TargetOpcode::G_FRAME_INDEX) {
addOffset(MIB, 0);
} else {
MachineOperand &InxOp = I.getOperand(2);
I.addOperand(InxOp); // set IndexReg
InxOp.ChangeToImmediate(1); // set Scale
MIB.addImm(0).addReg(0);
}
return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
}
bool X86InstructionSelector::selectGlobalValue(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_GLOBAL_VALUE) &&
"unexpected instruction");
auto GV = I.getOperand(1).getGlobal();
if (GV->isThreadLocal()) {
return false; // TODO: we don't support TLS yet.
}
// Can't handle alternate code models yet.
if (TM.getCodeModel() != CodeModel::Small)
return false;
X86AddressMode AM;
AM.GV = GV;
AM.GVOpFlags = STI.classifyGlobalReference(GV);
// TODO: The ABI requires an extra load. not supported yet.
if (isGlobalStubReference(AM.GVOpFlags))
return false;
// TODO: This reference is relative to the pic base. not supported yet.
if (isGlobalRelativeToPICBase(AM.GVOpFlags))
return false;
if (STI.isPICStyleRIPRel()) {
// Use rip-relative addressing.
assert(AM.Base.Reg == 0 && AM.IndexReg == 0);
AM.Base.Reg = X86::RIP;
}
const Register DefReg = I.getOperand(0).getReg();
LLT Ty = MRI.getType(DefReg);
unsigned NewOpc = getLeaOP(Ty, STI);
I.setDesc(TII.get(NewOpc));
MachineInstrBuilder MIB(MF, I);
I.RemoveOperand(1);
addFullAddress(MIB, AM);
return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
}
bool X86InstructionSelector::selectConstant(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_CONSTANT) &&
"unexpected instruction");
const Register DefReg = I.getOperand(0).getReg();
LLT Ty = MRI.getType(DefReg);
if (RBI.getRegBank(DefReg, MRI, TRI)->getID() != X86::GPRRegBankID)
return false;
uint64_t Val = 0;
if (I.getOperand(1).isCImm()) {
Val = I.getOperand(1).getCImm()->getZExtValue();
I.getOperand(1).ChangeToImmediate(Val);
} else if (I.getOperand(1).isImm()) {
Val = I.getOperand(1).getImm();
} else
llvm_unreachable("Unsupported operand type.");
unsigned NewOpc;
switch (Ty.getSizeInBits()) {
case 8:
NewOpc = X86::MOV8ri;
break;
case 16:
NewOpc = X86::MOV16ri;
break;
case 32:
NewOpc = X86::MOV32ri;
break;
case 64:
// TODO: in case isUInt<32>(Val), X86::MOV32ri can be used
if (isInt<32>(Val))
NewOpc = X86::MOV64ri32;
else
NewOpc = X86::MOV64ri;
break;
default:
llvm_unreachable("Can't select G_CONSTANT, unsupported type.");
}
I.setDesc(TII.get(NewOpc));
return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
}
// Helper function for selectTruncOrPtrToInt and selectAnyext.
// Returns true if DstRC lives on a floating register class and
// SrcRC lives on a 128-bit vector class.
static bool canTurnIntoCOPY(const TargetRegisterClass *DstRC,
const TargetRegisterClass *SrcRC) {
return (DstRC == &X86::FR32RegClass || DstRC == &X86::FR32XRegClass ||
DstRC == &X86::FR64RegClass || DstRC == &X86::FR64XRegClass) &&
(SrcRC == &X86::VR128RegClass || SrcRC == &X86::VR128XRegClass);
}
bool X86InstructionSelector::selectTurnIntoCOPY(
MachineInstr &I, MachineRegisterInfo &MRI, const unsigned DstReg,
const TargetRegisterClass *DstRC, const unsigned SrcReg,
const TargetRegisterClass *SrcRC) const {
if (!RBI.constrainGenericRegister(SrcReg, *SrcRC, MRI) ||
!RBI.constrainGenericRegister(DstReg, *DstRC, MRI)) {
LLVM_DEBUG(dbgs() << "Failed to constrain " << TII.getName(I.getOpcode())
<< " operand\n");
return false;
}
I.setDesc(TII.get(X86::COPY));
return true;
}
bool X86InstructionSelector::selectTruncOrPtrToInt(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_TRUNC ||
I.getOpcode() == TargetOpcode::G_PTRTOINT) &&
"unexpected instruction");
const Register DstReg = I.getOperand(0).getReg();
const Register SrcReg = I.getOperand(1).getReg();
const LLT DstTy = MRI.getType(DstReg);
const LLT SrcTy = MRI.getType(SrcReg);
const RegisterBank &DstRB = *RBI.getRegBank(DstReg, MRI, TRI);
const RegisterBank &SrcRB = *RBI.getRegBank(SrcReg, MRI, TRI);
if (DstRB.getID() != SrcRB.getID()) {
LLVM_DEBUG(dbgs() << TII.getName(I.getOpcode())
<< " input/output on different banks\n");
return false;
}
const TargetRegisterClass *DstRC = getRegClass(DstTy, DstRB);
const TargetRegisterClass *SrcRC = getRegClass(SrcTy, SrcRB);
if (!DstRC || !SrcRC)
return false;
// If that's truncation of the value that lives on the vector class and goes
// into the floating class, just replace it with copy, as we are able to
// select it as a regular move.
if (canTurnIntoCOPY(DstRC, SrcRC))
return selectTurnIntoCOPY(I, MRI, DstReg, DstRC, SrcReg, SrcRC);
if (DstRB.getID() != X86::GPRRegBankID)
return false;
unsigned SubIdx;
if (DstRC == SrcRC) {
// Nothing to be done
SubIdx = X86::NoSubRegister;
} else if (DstRC == &X86::GR32RegClass) {
SubIdx = X86::sub_32bit;
} else if (DstRC == &X86::GR16RegClass) {
SubIdx = X86::sub_16bit;
} else if (DstRC == &X86::GR8RegClass) {
SubIdx = X86::sub_8bit;
} else {
return false;
}
SrcRC = TRI.getSubClassWithSubReg(SrcRC, SubIdx);
if (!RBI.constrainGenericRegister(SrcReg, *SrcRC, MRI) ||
!RBI.constrainGenericRegister(DstReg, *DstRC, MRI)) {
LLVM_DEBUG(dbgs() << "Failed to constrain " << TII.getName(I.getOpcode())
<< "\n");
return false;
}
I.getOperand(1).setSubReg(SubIdx);
I.setDesc(TII.get(X86::COPY));
return true;
}
bool X86InstructionSelector::selectZext(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_ZEXT) && "unexpected instruction");
const Register DstReg = I.getOperand(0).getReg();
const Register SrcReg = I.getOperand(1).getReg();
const LLT DstTy = MRI.getType(DstReg);
const LLT SrcTy = MRI.getType(SrcReg);
assert(!(SrcTy == LLT::scalar(8) && DstTy == LLT::scalar(16)) &&
"8=>16 Zext is handled by tablegen");
assert(!(SrcTy == LLT::scalar(8) && DstTy == LLT::scalar(32)) &&
"8=>32 Zext is handled by tablegen");
assert(!(SrcTy == LLT::scalar(16) && DstTy == LLT::scalar(32)) &&
"16=>32 Zext is handled by tablegen");
assert(!(SrcTy == LLT::scalar(8) && DstTy == LLT::scalar(64)) &&
"8=>64 Zext is handled by tablegen");
assert(!(SrcTy == LLT::scalar(16) && DstTy == LLT::scalar(64)) &&
"16=>64 Zext is handled by tablegen");
assert(!(SrcTy == LLT::scalar(32) && DstTy == LLT::scalar(64)) &&
"32=>64 Zext is handled by tablegen");
if (SrcTy != LLT::scalar(1))
return false;
unsigned AndOpc;
if (DstTy == LLT::scalar(8))
AndOpc = X86::AND8ri;
else if (DstTy == LLT::scalar(16))
AndOpc = X86::AND16ri8;
else if (DstTy == LLT::scalar(32))
AndOpc = X86::AND32ri8;
else if (DstTy == LLT::scalar(64))
AndOpc = X86::AND64ri8;
else
return false;
Register DefReg = SrcReg;
if (DstTy != LLT::scalar(8)) {
Register ImpDefReg =
MRI.createVirtualRegister(getRegClass(DstTy, DstReg, MRI));
BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(TargetOpcode::IMPLICIT_DEF), ImpDefReg);
DefReg = MRI.createVirtualRegister(getRegClass(DstTy, DstReg, MRI));
BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(TargetOpcode::INSERT_SUBREG), DefReg)
.addReg(ImpDefReg)
.addReg(SrcReg)
.addImm(X86::sub_8bit);
}
MachineInstr &AndInst =
*BuildMI(*I.getParent(), I, I.getDebugLoc(), TII.get(AndOpc), DstReg)
.addReg(DefReg)
.addImm(1);
constrainSelectedInstRegOperands(AndInst, TII, TRI, RBI);
I.eraseFromParent();
return true;
}
bool X86InstructionSelector::selectAnyext(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_ANYEXT) && "unexpected instruction");
const Register DstReg = I.getOperand(0).getReg();
const Register SrcReg = I.getOperand(1).getReg();
const LLT DstTy = MRI.getType(DstReg);
const LLT SrcTy = MRI.getType(SrcReg);
const RegisterBank &DstRB = *RBI.getRegBank(DstReg, MRI, TRI);
const RegisterBank &SrcRB = *RBI.getRegBank(SrcReg, MRI, TRI);
assert(DstRB.getID() == SrcRB.getID() &&
"G_ANYEXT input/output on different banks\n");
assert(DstTy.getSizeInBits() > SrcTy.getSizeInBits() &&
"G_ANYEXT incorrect operand size");
const TargetRegisterClass *DstRC = getRegClass(DstTy, DstRB);
const TargetRegisterClass *SrcRC = getRegClass(SrcTy, SrcRB);
// If that's ANY_EXT of the value that lives on the floating class and goes
// into the vector class, just replace it with copy, as we are able to select
// it as a regular move.
if (canTurnIntoCOPY(SrcRC, DstRC))
return selectTurnIntoCOPY(I, MRI, SrcReg, SrcRC, DstReg, DstRC);
if (DstRB.getID() != X86::GPRRegBankID)
return false;
if (!RBI.constrainGenericRegister(SrcReg, *SrcRC, MRI) ||
!RBI.constrainGenericRegister(DstReg, *DstRC, MRI)) {
LLVM_DEBUG(dbgs() << "Failed to constrain " << TII.getName(I.getOpcode())
<< " operand\n");
return false;
}
if (SrcRC == DstRC) {
I.setDesc(TII.get(X86::COPY));
return true;
}
BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(TargetOpcode::SUBREG_TO_REG))
.addDef(DstReg)
.addImm(0)
.addReg(SrcReg)
.addImm(getSubRegIndex(SrcRC));
I.eraseFromParent();
return true;
}
bool X86InstructionSelector::selectCmp(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_ICMP) && "unexpected instruction");
X86::CondCode CC;
bool SwapArgs;
std::tie(CC, SwapArgs) = X86::getX86ConditionCode(
(CmpInst::Predicate)I.getOperand(1).getPredicate());
Register LHS = I.getOperand(2).getReg();
Register RHS = I.getOperand(3).getReg();
if (SwapArgs)
std::swap(LHS, RHS);
unsigned OpCmp;
LLT Ty = MRI.getType(LHS);
switch (Ty.getSizeInBits()) {
default:
return false;
case 8:
OpCmp = X86::CMP8rr;
break;
case 16:
OpCmp = X86::CMP16rr;
break;
case 32:
OpCmp = X86::CMP32rr;
break;
case 64:
OpCmp = X86::CMP64rr;
break;
}
MachineInstr &CmpInst =
*BuildMI(*I.getParent(), I, I.getDebugLoc(), TII.get(OpCmp))
.addReg(LHS)
.addReg(RHS);
MachineInstr &SetInst = *BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(X86::SETCCr), I.getOperand(0).getReg()).addImm(CC);
constrainSelectedInstRegOperands(CmpInst, TII, TRI, RBI);
constrainSelectedInstRegOperands(SetInst, TII, TRI, RBI);
I.eraseFromParent();
return true;
}
bool X86InstructionSelector::selectFCmp(MachineInstr &I,
MachineRegisterInfo &MRI,
MachineFunction &MF) const {
assert((I.getOpcode() == TargetOpcode::G_FCMP) && "unexpected instruction");
Register LhsReg = I.getOperand(2).getReg();
Register RhsReg = I.getOperand(3).getReg();
CmpInst::Predicate Predicate =
(CmpInst::Predicate)I.getOperand(1).getPredicate();
// FCMP_OEQ and FCMP_UNE cannot be checked with a single instruction.
static const uint16_t SETFOpcTable[2][3] = {
{X86::COND_E, X86::COND_NP, X86::AND8rr},
{X86::COND_NE, X86::COND_P, X86::OR8rr}};
const uint16_t *SETFOpc = nullptr;
switch (Predicate) {
default:
break;
case CmpInst::FCMP_OEQ:
SETFOpc = &SETFOpcTable[0][0];
break;
case CmpInst::FCMP_UNE:
SETFOpc = &SETFOpcTable[1][0];
break;
}
// Compute the opcode for the CMP instruction.
unsigned OpCmp;
LLT Ty = MRI.getType(LhsReg);
switch (Ty.getSizeInBits()) {
default:
return false;
case 32:
OpCmp = X86::UCOMISSrr;
break;
case 64:
OpCmp = X86::UCOMISDrr;
break;
}
Register ResultReg = I.getOperand(0).getReg();
RBI.constrainGenericRegister(
ResultReg,
*getRegClass(LLT::scalar(8), *RBI.getRegBank(ResultReg, MRI, TRI)), MRI);
if (SETFOpc) {
MachineInstr &CmpInst =
*BuildMI(*I.getParent(), I, I.getDebugLoc(), TII.get(OpCmp))
.addReg(LhsReg)
.addReg(RhsReg);
Register FlagReg1 = MRI.createVirtualRegister(&X86::GR8RegClass);
Register FlagReg2 = MRI.createVirtualRegister(&X86::GR8RegClass);
MachineInstr &Set1 = *BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(X86::SETCCr), FlagReg1).addImm(SETFOpc[0]);
MachineInstr &Set2 = *BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(X86::SETCCr), FlagReg2).addImm(SETFOpc[1]);
MachineInstr &Set3 = *BuildMI(*I.getParent(), I, I.getDebugLoc(),
TII.get(SETFOpc[2]), ResultReg)