1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
|
//===- ARMRegisterInfo.cpp - ARM Register Information -----------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the "Instituto Nokia de Tecnologia" and
// is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains the ARM implementation of the MRegisterInfo class.
//
//===----------------------------------------------------------------------===//
#include "ARM.h"
#include "ARMAddressingModes.h"
#include "ARMInstrInfo.h"
#include "ARMMachineFunctionInfo.h"
#include "ARMRegisterInfo.h"
#include "ARMSubtarget.h"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineLocation.h"
#include "llvm/CodeGen/RegisterScavenging.h"
#include "llvm/Target/TargetFrameInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/Support/CommandLine.h"
#include <algorithm>
using namespace llvm;
static cl::opt<bool> ThumbRegScavenging("enable-thumb-reg-scavenging",
cl::Hidden,
cl::desc("Enable register scavenging on Thumb"));
unsigned ARMRegisterInfo::getRegisterNumbering(unsigned RegEnum) {
using namespace ARM;
switch (RegEnum) {
case R0: case S0: case D0: return 0;
case R1: case S1: case D1: return 1;
case R2: case S2: case D2: return 2;
case R3: case S3: case D3: return 3;
case R4: case S4: case D4: return 4;
case R5: case S5: case D5: return 5;
case R6: case S6: case D6: return 6;
case R7: case S7: case D7: return 7;
case R8: case S8: case D8: return 8;
case R9: case S9: case D9: return 9;
case R10: case S10: case D10: return 10;
case R11: case S11: case D11: return 11;
case R12: case S12: case D12: return 12;
case SP: case S13: case D13: return 13;
case LR: case S14: case D14: return 14;
case PC: case S15: case D15: return 15;
case S16: return 16;
case S17: return 17;
case S18: return 18;
case S19: return 19;
case S20: return 20;
case S21: return 21;
case S22: return 22;
case S23: return 23;
case S24: return 24;
case S25: return 25;
case S26: return 26;
case S27: return 27;
case S28: return 28;
case S29: return 29;
case S30: return 30;
case S31: return 31;
default:
assert(0 && "Unknown ARM register!");
abort();
}
}
ARMRegisterInfo::ARMRegisterInfo(const TargetInstrInfo &tii,
const ARMSubtarget &sti)
: ARMGenRegisterInfo(ARM::ADJCALLSTACKDOWN, ARM::ADJCALLSTACKUP),
TII(tii), STI(sti),
FramePtr((STI.useThumbBacktraces() || STI.isThumb()) ? ARM::R7 : ARM::R11) {
}
bool ARMRegisterInfo::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI) const {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (!AFI->isThumbFunction() || CSI.empty())
return false;
MachineInstrBuilder MIB = BuildMI(MBB, MI, TII.get(ARM::tPUSH));
for (unsigned i = CSI.size(); i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
// Add the callee-saved register as live-in. It's killed at the spill.
MBB.addLiveIn(Reg);
MIB.addReg(Reg, false/*isDef*/,false/*isImp*/,true/*isKill*/);
}
return true;
}
bool ARMRegisterInfo::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI) const {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (!AFI->isThumbFunction() || CSI.empty())
return false;
bool isVarArg = AFI->getVarArgsRegSaveSize() > 0;
MachineInstr *PopMI = new MachineInstr(TII.get(ARM::tPOP));
MBB.insert(MI, PopMI);
for (unsigned i = CSI.size(); i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
if (Reg == ARM::LR) {
// Special epilogue for vararg functions. See emitEpilogue
if (isVarArg)
continue;
Reg = ARM::PC;
PopMI->setInstrDescriptor(TII.get(ARM::tPOP_RET));
MBB.erase(MI);
}
PopMI->addRegOperand(Reg, true);
}
return true;
}
static inline
const MachineInstrBuilder &AddDefaultPred(const MachineInstrBuilder &MIB) {
return MIB.addImm((int64_t)ARMCC::AL).addReg(0);
}
static inline
const MachineInstrBuilder &AddDefaultCC(const MachineInstrBuilder &MIB) {
return MIB.addReg(0);
}
static const MachineInstrBuilder &ARMInstrAddOperand(MachineInstrBuilder &MIB,
MachineOperand &MO) {
if (MO.isRegister())
MIB = MIB.addReg(MO.getReg(), MO.isDef(), MO.isImplicit());
else if (MO.isImmediate())
MIB = MIB.addImm(MO.getImm());
else if (MO.isFrameIndex())
MIB = MIB.addFrameIndex(MO.getFrameIndex());
else
assert(0 && "Unknown operand for ARMInstrAddOperand!");
return MIB;
}
void ARMRegisterInfo::
storeRegToStackSlot(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
unsigned SrcReg, bool isKill, int FI,
const TargetRegisterClass *RC) const {
if (RC == ARM::GPRRegisterClass) {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (AFI->isThumbFunction())
BuildMI(MBB, I, TII.get(ARM::tSpill)).addReg(SrcReg, false, false, isKill)
.addFrameIndex(FI).addImm(0);
else
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::STR))
.addReg(SrcReg, false, false, isKill)
.addFrameIndex(FI).addReg(0).addImm(0));
} else if (RC == ARM::DPRRegisterClass) {
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FSTD))
.addReg(SrcReg, false, false, isKill)
.addFrameIndex(FI).addImm(0));
} else {
assert(RC == ARM::SPRRegisterClass && "Unknown regclass!");
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FSTS))
.addReg(SrcReg, false, false, isKill)
.addFrameIndex(FI).addImm(0));
}
}
void ARMRegisterInfo::storeRegToAddr(MachineFunction &MF, unsigned SrcReg,
bool isKill,
SmallVectorImpl<MachineOperand> &Addr,
const TargetRegisterClass *RC,
SmallVectorImpl<MachineInstr*> &NewMIs) const {
unsigned Opc = 0;
if (RC == ARM::GPRRegisterClass) {
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (AFI->isThumbFunction()) {
Opc = Addr[0].isFrameIndex() ? ARM::tSpill : ARM::tSTR;
MachineInstrBuilder MIB =
BuildMI(TII.get(Opc)).addReg(SrcReg, false, false, isKill);
for (unsigned i = 0, e = Addr.size(); i != e; ++i)
MIB = ARMInstrAddOperand(MIB, Addr[i]);
NewMIs.push_back(MIB);
return;
}
Opc = ARM::STR;
} else if (RC == ARM::DPRRegisterClass) {
Opc = ARM::FSTD;
} else {
assert(RC == ARM::SPRRegisterClass && "Unknown regclass!");
Opc = ARM::FSTS;
}
MachineInstrBuilder MIB =
BuildMI(TII.get(Opc)).addReg(SrcReg, false, false, isKill);
for (unsigned i = 0, e = Addr.size(); i != e; ++i)
MIB = ARMInstrAddOperand(MIB, Addr[i]);
AddDefaultPred(MIB);
NewMIs.push_back(MIB);
return;
}
void ARMRegisterInfo::
loadRegFromStackSlot(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
unsigned DestReg, int FI,
const TargetRegisterClass *RC) const {
if (RC == ARM::GPRRegisterClass) {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (AFI->isThumbFunction())
BuildMI(MBB, I, TII.get(ARM::tRestore), DestReg)
.addFrameIndex(FI).addImm(0);
else
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::LDR), DestReg)
.addFrameIndex(FI).addReg(0).addImm(0));
} else if (RC == ARM::DPRRegisterClass) {
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FLDD), DestReg)
.addFrameIndex(FI).addImm(0));
} else {
assert(RC == ARM::SPRRegisterClass && "Unknown regclass!");
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FLDS), DestReg)
.addFrameIndex(FI).addImm(0));
}
}
void ARMRegisterInfo::loadRegFromAddr(MachineFunction &MF, unsigned DestReg,
SmallVectorImpl<MachineOperand> &Addr,
const TargetRegisterClass *RC,
SmallVectorImpl<MachineInstr*> &NewMIs) const {
unsigned Opc = 0;
if (RC == ARM::GPRRegisterClass) {
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (AFI->isThumbFunction()) {
Opc = Addr[0].isFrameIndex() ? ARM::tRestore : ARM::tLDR;
MachineInstrBuilder MIB = BuildMI(TII.get(Opc), DestReg);
for (unsigned i = 0, e = Addr.size(); i != e; ++i)
MIB = ARMInstrAddOperand(MIB, Addr[i]);
NewMIs.push_back(MIB);
return;
}
Opc = ARM::LDR;
} else if (RC == ARM::DPRRegisterClass) {
Opc = ARM::FLDD;
} else {
assert(RC == ARM::SPRRegisterClass && "Unknown regclass!");
Opc = ARM::FLDS;
}
MachineInstrBuilder MIB = BuildMI(TII.get(Opc), DestReg);
for (unsigned i = 0, e = Addr.size(); i != e; ++i)
MIB = ARMInstrAddOperand(MIB, Addr[i]);
AddDefaultPred(MIB);
NewMIs.push_back(MIB);
return;
}
void ARMRegisterInfo::copyRegToReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg, unsigned SrcReg,
const TargetRegisterClass *DestRC,
const TargetRegisterClass *SrcRC) const {
if (DestRC != SrcRC) {
cerr << "Not yet supported!";
abort();
}
if (DestRC == ARM::GPRRegisterClass) {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (AFI->isThumbFunction())
BuildMI(MBB, I, TII.get(ARM::tMOVr), DestReg).addReg(SrcReg);
else
AddDefaultCC(AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::MOVr), DestReg)
.addReg(SrcReg)));
} else if (DestRC == ARM::SPRRegisterClass)
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FCPYS), DestReg)
.addReg(SrcReg));
else if (DestRC == ARM::DPRRegisterClass)
AddDefaultPred(BuildMI(MBB, I, TII.get(ARM::FCPYD), DestReg)
.addReg(SrcReg));
else
abort();
}
/// emitLoadConstPool - Emits a load from constpool to materialize the
/// specified immediate.
static void emitLoadConstPool(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
unsigned DestReg, int Val,
ARMCC::CondCodes Pred, unsigned PredReg,
const TargetInstrInfo &TII, bool isThumb) {
MachineFunction &MF = *MBB.getParent();
MachineConstantPool *ConstantPool = MF.getConstantPool();
Constant *C = ConstantInt::get(Type::Int32Ty, Val);
unsigned Idx = ConstantPool->getConstantPoolIndex(C, 2);
if (isThumb)
BuildMI(MBB, MBBI, TII.get(ARM::tLDRcp), DestReg).addConstantPoolIndex(Idx);
else
BuildMI(MBB, MBBI, TII.get(ARM::LDRcp), DestReg).addConstantPoolIndex(Idx)
.addReg(0).addImm(0).addImm((unsigned)Pred).addReg(PredReg);
}
void ARMRegisterInfo::reMaterialize(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg,
const MachineInstr *Orig) const {
if (Orig->getOpcode() == ARM::MOVi2pieces) {
emitLoadConstPool(MBB, I, DestReg,
Orig->getOperand(1).getImmedValue(),
(ARMCC::CondCodes)Orig->getOperand(2).getImmedValue(),
Orig->getOperand(3).getReg(),
TII, false);
return;
}
MachineInstr *MI = Orig->clone();
MI->getOperand(0).setReg(DestReg);
MBB.insert(I, MI);
}
/// isLowRegister - Returns true if the register is low register r0-r7.
///
static bool isLowRegister(unsigned Reg) {
using namespace ARM;
switch (Reg) {
case R0: case R1: case R2: case R3:
case R4: case R5: case R6: case R7:
return true;
default:
return false;
}
}
MachineInstr *ARMRegisterInfo::foldMemoryOperand(MachineInstr *MI,
SmallVectorImpl<unsigned> &Ops,
int FI) const {
if (Ops.size() != 1) return NULL;
unsigned OpNum = Ops[0];
unsigned Opc = MI->getOpcode();
MachineInstr *NewMI = NULL;
switch (Opc) {
default: break;
case ARM::MOVr: {
if (MI->getOperand(4).getReg() == ARM::CPSR)
// If it is updating CPSR, then it cannot be foled.
break;
unsigned Pred = MI->getOperand(2).getImmedValue();
unsigned PredReg = MI->getOperand(3).getReg();
if (OpNum == 0) { // move -> store
unsigned SrcReg = MI->getOperand(1).getReg();
NewMI = BuildMI(TII.get(ARM::STR)).addReg(SrcReg).addFrameIndex(FI)
.addReg(0).addImm(0).addImm(Pred).addReg(PredReg);
} else { // move -> load
unsigned DstReg = MI->getOperand(0).getReg();
NewMI = BuildMI(TII.get(ARM::LDR), DstReg).addFrameIndex(FI).addReg(0)
.addImm(0).addImm(Pred).addReg(PredReg);
}
break;
}
case ARM::tMOVr: {
if (OpNum == 0) { // move -> store
unsigned SrcReg = MI->getOperand(1).getReg();
if (isPhysicalRegister(SrcReg) && !isLowRegister(SrcReg))
// tSpill cannot take a high register operand.
break;
NewMI = BuildMI(TII.get(ARM::tSpill)).addReg(SrcReg).addFrameIndex(FI)
.addImm(0);
} else { // move -> load
unsigned DstReg = MI->getOperand(0).getReg();
if (isPhysicalRegister(DstReg) && !isLowRegister(DstReg))
// tRestore cannot target a high register operand.
break;
NewMI = BuildMI(TII.get(ARM::tRestore), DstReg).addFrameIndex(FI)
.addImm(0);
}
break;
}
case ARM::FCPYS: {
unsigned Pred = MI->getOperand(2).getImmedValue();
unsigned PredReg = MI->getOperand(3).getReg();
if (OpNum == 0) { // move -> store
unsigned SrcReg = MI->getOperand(1).getReg();
NewMI = BuildMI(TII.get(ARM::FSTS)).addReg(SrcReg).addFrameIndex(FI)
.addImm(0).addImm(Pred).addReg(PredReg);
} else { // move -> load
unsigned DstReg = MI->getOperand(0).getReg();
NewMI = BuildMI(TII.get(ARM::FLDS), DstReg).addFrameIndex(FI)
.addImm(0).addImm(Pred).addReg(PredReg);
}
break;
}
case ARM::FCPYD: {
unsigned Pred = MI->getOperand(2).getImmedValue();
unsigned PredReg = MI->getOperand(3).getReg();
if (OpNum == 0) { // move -> store
unsigned SrcReg = MI->getOperand(1).getReg();
NewMI = BuildMI(TII.get(ARM::FSTD)).addReg(SrcReg).addFrameIndex(FI)
.addImm(0).addImm(Pred).addReg(PredReg);
} else { // move -> load
unsigned DstReg = MI->getOperand(0).getReg();
NewMI = BuildMI(TII.get(ARM::FLDD), DstReg).addFrameIndex(FI)
.addImm(0).addImm(Pred).addReg(PredReg);
}
break;
}
}
if (NewMI)
NewMI->copyKillDeadInfo(MI);
return NewMI;
}
bool ARMRegisterInfo::canFoldMemoryOperand(MachineInstr *MI,
SmallVectorImpl<unsigned> &Ops) const {
if (Ops.size() != 1) return false;
unsigned OpNum = Ops[0];
unsigned Opc = MI->getOpcode();
switch (Opc) {
default: break;
case ARM::MOVr:
// If it is updating CPSR, then it cannot be foled.
return MI->getOperand(4).getReg() != ARM::CPSR;
case ARM::tMOVr: {
if (OpNum == 0) { // move -> store
unsigned SrcReg = MI->getOperand(1).getReg();
if (isPhysicalRegister(SrcReg) && !isLowRegister(SrcReg))
// tSpill cannot take a high register operand.
return false;
} else { // move -> load
unsigned DstReg = MI->getOperand(0).getReg();
if (isPhysicalRegister(DstReg) && !isLowRegister(DstReg))
// tRestore cannot target a high register operand.
return false;
}
return true;
}
case ARM::FCPYS:
case ARM::FCPYD:
return true;
}
return false;
}
const unsigned*
ARMRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
static const unsigned CalleeSavedRegs[] = {
ARM::LR, ARM::R11, ARM::R10, ARM::R9, ARM::R8,
ARM::R7, ARM::R6, ARM::R5, ARM::R4,
ARM::D15, ARM::D14, ARM::D13, ARM::D12,
ARM::D11, ARM::D10, ARM::D9, ARM::D8,
0
};
static const unsigned DarwinCalleeSavedRegs[] = {
ARM::LR, ARM::R7, ARM::R6, ARM::R5, ARM::R4,
ARM::R11, ARM::R10, ARM::R9, ARM::R8,
ARM::D15, ARM::D14, ARM::D13, ARM::D12,
ARM::D11, ARM::D10, ARM::D9, ARM::D8,
0
};
return STI.isTargetDarwin() ? DarwinCalleeSavedRegs : CalleeSavedRegs;
}
const TargetRegisterClass* const *
ARMRegisterInfo::getCalleeSavedRegClasses(const MachineFunction *MF) const {
static const TargetRegisterClass * const CalleeSavedRegClasses[] = {
&ARM::GPRRegClass, &ARM::GPRRegClass, &ARM::GPRRegClass,
&ARM::GPRRegClass, &ARM::GPRRegClass, &ARM::GPRRegClass,
&ARM::GPRRegClass, &ARM::GPRRegClass, &ARM::GPRRegClass,
&ARM::DPRRegClass, &ARM::DPRRegClass, &ARM::DPRRegClass, &ARM::DPRRegClass,
&ARM::DPRRegClass, &ARM::DPRRegClass, &ARM::DPRRegClass, &ARM::DPRRegClass,
0
};
return CalleeSavedRegClasses;
}
BitVector ARMRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
// FIXME: avoid re-calculating this everytime.
BitVector Reserved(getNumRegs());
Reserved.set(ARM::SP);
Reserved.set(ARM::PC);
if (STI.isTargetDarwin() || hasFP(MF))
Reserved.set(FramePtr);
// Some targets reserve R9.
if (STI.isR9Reserved())
Reserved.set(ARM::R9);
return Reserved;
}
bool
ARMRegisterInfo::isReservedReg(const MachineFunction &MF, unsigned Reg) const {
switch (Reg) {
default: break;
case ARM::SP:
case ARM::PC:
return true;
case ARM::R7:
case ARM::R11:
if (FramePtr == Reg && (STI.isTargetDarwin() || hasFP(MF)))
return true;
break;
case ARM::R9:
return STI.isR9Reserved();
}
return false;
}
bool
ARMRegisterInfo::requiresRegisterScavenging(const MachineFunction &MF) const {
const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
return ThumbRegScavenging || !AFI->isThumbFunction();
}
/// hasFP - Return true if the specified function should have a dedicated frame
/// pointer register. This is true if the function has variable sized allocas
/// or if frame pointer elimination is disabled.
///
bool ARMRegisterInfo::hasFP(const MachineFunction &MF) const {
return NoFramePointerElim || MF.getFrameInfo()->hasVarSizedObjects();
}
// hasReservedCallFrame - Under normal circumstances, when a frame pointer is
// not required, we reserve argument space for call sites in the function
// immediately on entry to the current function. This eliminates the need for
// add/sub sp brackets around call sites. Returns true if the call frame is
// included as part of the stack frame.
bool ARMRegisterInfo::hasReservedCallFrame(MachineFunction &MF) const {
const MachineFrameInfo *FFI = MF.getFrameInfo();
unsigned CFSize = FFI->getMaxCallFrameSize();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
// It's not always a good idea to include the call frame as part of the
// stack frame. ARM (especially Thumb) has small immediate offset to
// address the stack frame. So a large call frame can cause poor codegen
// and may even makes it impossible to scavenge a register.
if (AFI->isThumbFunction()) {
if (CFSize >= ((1 << 8) - 1) * 4 / 2) // Half of imm8 * 4
return false;
} else {
if (CFSize >= ((1 << 12) - 1) / 2) // Half of imm12
return false;
}
return !MF.getFrameInfo()->hasVarSizedObjects();
}
/// emitARMRegPlusImmediate - Emits a series of instructions to materialize
/// a destreg = basereg + immediate in ARM code.
static
void emitARMRegPlusImmediate(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
unsigned DestReg, unsigned BaseReg, int NumBytes,
ARMCC::CondCodes Pred, unsigned PredReg,
const TargetInstrInfo &TII) {
bool isSub = NumBytes < 0;
if (isSub) NumBytes = -NumBytes;
while (NumBytes) {
unsigned RotAmt = ARM_AM::getSOImmValRotate(NumBytes);
unsigned ThisVal = NumBytes & ARM_AM::rotr32(0xFF, RotAmt);
assert(ThisVal && "Didn't extract field correctly");
// We will handle these bits from offset, clear them.
NumBytes &= ~ThisVal;
// Get the properly encoded SOImmVal field.
int SOImmVal = ARM_AM::getSOImmVal(ThisVal);
assert(SOImmVal != -1 && "Bit extraction didn't work?");
// Build the new ADD / SUB.
BuildMI(MBB, MBBI, TII.get(isSub ? ARM::SUBri : ARM::ADDri), DestReg)
.addReg(BaseReg, false, false, true).addImm(SOImmVal)
.addImm((unsigned)Pred).addReg(PredReg).addReg(0);
BaseReg = DestReg;
}
}
/// calcNumMI - Returns the number of instructions required to materialize
/// the specific add / sub r, c instruction.
static unsigned calcNumMI(int Opc, int ExtraOpc, unsigned Bytes,
unsigned NumBits, unsigned Scale) {
unsigned NumMIs = 0;
unsigned Chunk = ((1 << NumBits) - 1) * Scale;
if (Opc == ARM::tADDrSPi) {
unsigned ThisVal = (Bytes > Chunk) ? Chunk : Bytes;
Bytes -= ThisVal;
NumMIs++;
NumBits = 8;
Scale = 1; // Followed by a number of tADDi8.
Chunk = ((1 << NumBits) - 1) * Scale;
}
NumMIs += Bytes / Chunk;
if ((Bytes % Chunk) != 0)
NumMIs++;
if (ExtraOpc)
NumMIs++;
return NumMIs;
}
/// emitThumbRegPlusImmInReg - Emits a series of instructions to materialize
/// a destreg = basereg + immediate in Thumb code. Materialize the immediate
/// in a register using mov / mvn sequences or load the immediate from a
/// constpool entry.
static
void emitThumbRegPlusImmInReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
unsigned DestReg, unsigned BaseReg,
int NumBytes, bool CanChangeCC,
const TargetInstrInfo &TII) {
bool isHigh = !isLowRegister(DestReg) ||
(BaseReg != 0 && !isLowRegister(BaseReg));
bool isSub = false;
// Subtract doesn't have high register version. Load the negative value
// if either base or dest register is a high register. Also, if do not
// issue sub as part of the sequence if condition register is to be
// preserved.
if (NumBytes < 0 && !isHigh && CanChangeCC) {
isSub = true;
NumBytes = -NumBytes;
}
unsigned LdReg = DestReg;
if (DestReg == ARM::SP) {
assert(BaseReg == ARM::SP && "Unexpected!");
LdReg = ARM::R3;
BuildMI(MBB, MBBI, TII.get(ARM::tMOVr), ARM::R12)
.addReg(ARM::R3, false, false, true);
}
if (NumBytes <= 255 && NumBytes >= 0)
BuildMI(MBB, MBBI, TII.get(ARM::tMOVi8), LdReg).addImm(NumBytes);
else if (NumBytes < 0 && NumBytes >= -255) {
BuildMI(MBB, MBBI, TII.get(ARM::tMOVi8), LdReg).addImm(NumBytes);
BuildMI(MBB, MBBI, TII.get(ARM::tNEG), LdReg)
.addReg(LdReg, false, false, true);
} else
emitLoadConstPool(MBB, MBBI, LdReg, NumBytes, ARMCC::AL, 0, TII, true);
// Emit add / sub.
int Opc = (isSub) ? ARM::tSUBrr : (isHigh ? ARM::tADDhirr : ARM::tADDrr);
const MachineInstrBuilder MIB = BuildMI(MBB, MBBI, TII.get(Opc), DestReg);
if (DestReg == ARM::SP || isSub)
MIB.addReg(BaseReg).addReg(LdReg, false, false, true);
else
MIB.addReg(LdReg).addReg(BaseReg, false, false, true);
if (DestReg == ARM::SP)
BuildMI(MBB, MBBI, TII.get(ARM::tMOVr), ARM::R3)
.addReg(ARM::R12, false, false, true);
}
/// emitThumbRegPlusImmediate - Emits a series of instructions to materialize
/// a destreg = basereg + immediate in Thumb code.
static
void emitThumbRegPlusImmediate(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
unsigned DestReg, unsigned BaseReg,
int NumBytes, const TargetInstrInfo &TII) {
bool isSub = NumBytes < 0;
unsigned Bytes = (unsigned)NumBytes;
if (isSub) Bytes = -NumBytes;
bool isMul4 = (Bytes & 3) == 0;
bool isTwoAddr = false;
bool DstNotEqBase = false;
unsigned NumBits = 1;
unsigned Scale = 1;
int Opc = 0;
int ExtraOpc = 0;
if (DestReg == BaseReg && BaseReg == ARM::SP) {
assert(isMul4 && "Thumb sp inc / dec size must be multiple of 4!");
NumBits = 7;
Scale = 4;
Opc = isSub ? ARM::tSUBspi : ARM::tADDspi;
isTwoAddr = true;
} else if (!isSub && BaseReg == ARM::SP) {
// r1 = add sp, 403
// =>
// r1 = add sp, 100 * 4
// r1 = add r1, 3
if (!isMul4) {
Bytes &= ~3;
ExtraOpc = ARM::tADDi3;
}
NumBits = 8;
Scale = 4;
Opc = ARM::tADDrSPi;
} else {
// sp = sub sp, c
// r1 = sub sp, c
// r8 = sub sp, c
if (DestReg != BaseReg)
DstNotEqBase = true;
NumBits = 8;
Opc = isSub ? ARM::tSUBi8 : ARM::tADDi8;
isTwoAddr = true;
}
unsigned NumMIs = calcNumMI(Opc, ExtraOpc, Bytes, NumBits, Scale);
unsigned Threshold = (DestReg == ARM::SP) ? 3 : 2;
if (NumMIs > Threshold) {
// This will expand into too many instructions. Load the immediate from a
// constpool entry.
emitThumbRegPlusImmInReg(MBB, MBBI, DestReg, BaseReg, NumBytes, true, TII);
return;
}
if (DstNotEqBase) {
if (isLowRegister(DestReg) && isLowRegister(BaseReg)) {
// If both are low registers, emit DestReg = add BaseReg, max(Imm, 7)
unsigned Chunk = (1 << 3) - 1;
unsigned ThisVal = (Bytes > Chunk) ? Chunk : Bytes;
Bytes -= ThisVal;
BuildMI(MBB, MBBI, TII.get(isSub ? ARM::tSUBi3 : ARM::tADDi3), DestReg)
.addReg(BaseReg, false, false, true).addImm(ThisVal);
} else {
BuildMI(MBB, MBBI, TII.get(ARM::tMOVr), DestReg)
.addReg(BaseReg, false, false, true);
}
BaseReg = DestReg;
}
unsigned Chunk = ((1 << NumBits) - 1) * Scale;
while (Bytes) {
unsigned ThisVal = (Bytes > Chunk) ? Chunk : Bytes;
Bytes -= ThisVal;
ThisVal /= Scale;
// Build the new tADD / tSUB.
if (isTwoAddr)
BuildMI(MBB, MBBI, TII.get(Opc), DestReg).addReg(DestReg).addImm(ThisVal);
else {
bool isKill = BaseReg != ARM::SP;
BuildMI(MBB, MBBI, TII.get(Opc), DestReg)
.addReg(BaseReg, false, false, isKill).addImm(ThisVal);
BaseReg = DestReg;
if (Opc == ARM::tADDrSPi) {
// r4 = add sp, imm
// r4 = add r4, imm
// ...
NumBits = 8;
Scale = 1;
Chunk = ((1 << NumBits) - 1) * Scale;
Opc = isSub ? ARM::tSUBi8 : ARM::tADDi8;
isTwoAddr = true;
}
}
}
if (ExtraOpc)
BuildMI(MBB, MBBI, TII.get(ExtraOpc), DestReg)
.addReg(DestReg, false, false, true)
.addImm(((unsigned)NumBytes) & 3);
}
static
void emitSPUpdate(MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI,
int NumBytes, ARMCC::CondCodes Pred, unsigned PredReg,
bool isThumb, const TargetInstrInfo &TII) {
if (isThumb)
emitThumbRegPlusImmediate(MBB, MBBI, ARM::SP, ARM::SP, NumBytes, TII);
else
emitARMRegPlusImmediate(MBB, MBBI, ARM::SP, ARM::SP, NumBytes,
Pred, PredReg, TII);
}
void ARMRegisterInfo::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I) const {
if (!hasReservedCallFrame(MF)) {
// If we have alloca, convert as follows:
// ADJCALLSTACKDOWN -> sub, sp, sp, amount
// ADJCALLSTACKUP -> add, sp, sp, amount
MachineInstr *Old = I;
unsigned Amount = Old->getOperand(0).getImmedValue();
if (Amount != 0) {
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
// alignment boundary.
unsigned Align = MF.getTarget().getFrameInfo()->getStackAlignment();
Amount = (Amount+Align-1)/Align*Align;
// Replace the pseudo instruction with a new instruction...
unsigned Opc = Old->getOpcode();
bool isThumb = AFI->isThumbFunction();
ARMCC::CondCodes Pred = isThumb
? ARMCC::AL : (ARMCC::CondCodes)Old->getOperand(1).getImmedValue();
if (Opc == ARM::ADJCALLSTACKDOWN || Opc == ARM::tADJCALLSTACKDOWN) {
// Note: PredReg is operand 2 for ADJCALLSTACKDOWN.
unsigned PredReg = isThumb ? 0 : Old->getOperand(2).getReg();
emitSPUpdate(MBB, I, -Amount, Pred, PredReg, isThumb, TII);
} else {
// Note: PredReg is operand 3 for ADJCALLSTACKUP.
unsigned PredReg = isThumb ? 0 : Old->getOperand(3).getReg();
assert(Opc == ARM::ADJCALLSTACKUP || Opc == ARM::tADJCALLSTACKUP);
emitSPUpdate(MBB, I, Amount, Pred, PredReg, isThumb, TII);
}
}
}
MBB.erase(I);
}
/// emitThumbConstant - Emit a series of instructions to materialize a
/// constant.
static void emitThumbConstant(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
unsigned DestReg, int Imm,
const TargetInstrInfo &TII) {
bool isSub = Imm < 0;
if (isSub) Imm = -Imm;
int Chunk = (1 << 8) - 1;
int ThisVal = (Imm > Chunk) ? Chunk : Imm;
Imm -= ThisVal;
BuildMI(MBB, MBBI, TII.get(ARM::tMOVi8), DestReg).addImm(ThisVal);
if (Imm > 0)
emitThumbRegPlusImmediate(MBB, MBBI, DestReg, DestReg, Imm, TII);
if (isSub)
BuildMI(MBB, MBBI, TII.get(ARM::tNEG), DestReg)
.addReg(DestReg, false, false, true);
}
/// findScratchRegister - Find a 'free' ARM register. If register scavenger
/// is not being used, R12 is available. Otherwise, try for a call-clobbered
/// register first and then a spilled callee-saved register if that fails.
static
unsigned findScratchRegister(RegScavenger *RS, const TargetRegisterClass *RC,
ARMFunctionInfo *AFI) {
unsigned Reg = RS ? RS->FindUnusedReg(RC, true) : (unsigned) ARM::R12;
if (Reg == 0)
// Try a already spilled CS register.
Reg = RS->FindUnusedReg(RC, AFI->getSpilledCSRegisters());
return Reg;
}
void ARMRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
int SPAdj, RegScavenger *RS) const{
unsigned i = 0;
MachineInstr &MI = *II;
MachineBasicBlock &MBB = *MI.getParent();
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
bool isThumb = AFI->isThumbFunction();
while (!MI.getOperand(i).isFrameIndex()) {
++i;
assert(i < MI.getNumOperands() && "Instr doesn't have FrameIndex operand!");
}
unsigned FrameReg = ARM::SP;
int FrameIndex = MI.getOperand(i).getFrameIndex();
int Offset = MF.getFrameInfo()->getObjectOffset(FrameIndex) +
MF.getFrameInfo()->getStackSize() + SPAdj;
if (AFI->isGPRCalleeSavedArea1Frame(FrameIndex))
Offset -= AFI->getGPRCalleeSavedArea1Offset();
else if (AFI->isGPRCalleeSavedArea2Frame(FrameIndex))
Offset -= AFI->getGPRCalleeSavedArea2Offset();
else if (AFI->isDPRCalleeSavedAreaFrame(FrameIndex))
Offset -= AFI->getDPRCalleeSavedAreaOffset();
else if (hasFP(MF)) {
assert(SPAdj == 0 && "Unexpected");
// There is alloca()'s in this function, must reference off the frame
// pointer instead.
FrameReg = getFrameRegister(MF);
Offset -= AFI->getFramePtrSpillOffset();
}
unsigned Opcode = MI.getOpcode();
const TargetInstrDescriptor &Desc = TII.get(Opcode);
unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask);
bool isSub = false;
if (Opcode == ARM::ADDri) {
Offset += MI.getOperand(i+1).getImm();
if (Offset == 0) {
// Turn it into a move.
MI.setInstrDescriptor(TII.get(ARM::MOVr));
MI.getOperand(i).ChangeToRegister(FrameReg, false);
MI.RemoveOperand(i+1);
return;
} else if (Offset < 0) {
Offset = -Offset;
isSub = true;
MI.setInstrDescriptor(TII.get(ARM::SUBri));
}
// Common case: small offset, fits into instruction.
int ImmedOffset = ARM_AM::getSOImmVal(Offset);
if (ImmedOffset != -1) {
// Replace the FrameIndex with sp / fp
MI.getOperand(i).ChangeToRegister(FrameReg, false);
MI.getOperand(i+1).ChangeToImmediate(ImmedOffset);
return;
}
// Otherwise, we fallback to common code below to form the imm offset with
// a sequence of ADDri instructions. First though, pull as much of the imm
// into this ADDri as possible.
unsigned RotAmt = ARM_AM::getSOImmValRotate(Offset);
unsigned ThisImmVal = Offset & ARM_AM::rotr32(0xFF, RotAmt);
// We will handle these bits from offset, clear them.
Offset &= ~ThisImmVal;
// Get the properly encoded SOImmVal field.
int ThisSOImmVal = ARM_AM::getSOImmVal(ThisImmVal);
assert(ThisSOImmVal != -1 && "Bit extraction didn't work?");
MI.getOperand(i+1).ChangeToImmediate(ThisSOImmVal);
} else if (Opcode == ARM::tADDrSPi) {
Offset += MI.getOperand(i+1).getImm();
// Can't use tADDrSPi if it's based off the frame pointer.
unsigned NumBits = 0;
unsigned Scale = 1;
if (FrameReg != ARM::SP) {
Opcode = ARM::tADDi3;
MI.setInstrDescriptor(TII.get(ARM::tADDi3));
NumBits = 3;
} else {
NumBits = 8;
Scale = 4;
assert((Offset & 3) == 0 &&
"Thumb add/sub sp, #imm immediate must be multiple of 4!");
}
if (Offset == 0) {
// Turn it into a move.
MI.setInstrDescriptor(TII.get(ARM::tMOVr));
MI.getOperand(i).ChangeToRegister(FrameReg, false);
MI.RemoveOperand(i+1);
return;
}
// Common case: small offset, fits into instruction.
unsigned Mask = (1 << NumBits) - 1;
if (((Offset / Scale) & ~Mask) == 0) {
// Replace the FrameIndex with sp / fp
MI.getOperand(i).ChangeToRegister(FrameReg, false);
MI.getOperand(i+1).ChangeToImmediate(Offset / Scale);
return;
}
unsigned DestReg = MI.getOperand(0).getReg();
unsigned Bytes = (Offset > 0) ? Offset : -Offset;
unsigned NumMIs = calcNumMI(Opcode, 0, Bytes, NumBits, Scale);
// MI would expand into a large number of instructions. Don't try to
// simplify the immediate.
if (NumMIs > 2) {
emitThumbRegPlusImmediate(MBB, II, DestReg, FrameReg, Offset, TII);
MBB.erase(II);
return;
}
if (Offset > 0) {
// Translate r0 = add sp, imm to
// r0 = add sp, 255*4
// r0 = add r0, (imm - 255*4)
MI.getOperand(i).ChangeToRegister(FrameReg, false);
MI.getOperand(i+1).ChangeToImmediate(Mask);
Offset = (Offset - Mask * Scale);
MachineBasicBlock::iterator NII = next(II);
emitThumbRegPlusImmediate(MBB, NII, DestReg, DestReg, Offset, TII);
} else {
// Translate r0 = add sp, -imm to
// r0 = -imm (this is then translated into a series of instructons)
// r0 = add r0, sp
emitThumbConstant(MBB, II, DestReg, Offset, TII);
MI.setInstrDescriptor(TII.get(ARM::tADDhirr));
MI.getOperand(i).ChangeToRegister(DestReg, false, false, true);
MI.getOperand(i+1).ChangeToRegister(FrameReg, false);
}
return;
} else {
unsigned ImmIdx = 0;
int InstrOffs = 0;
unsigned NumBits = 0;
unsigned Scale = 1;
switch (AddrMode) {
case ARMII::AddrMode2: {
ImmIdx = i+2;
InstrOffs = ARM_AM::getAM2Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM2Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 12;
break;
}
case ARMII::AddrMode3: {
ImmIdx = i+2;
InstrOffs = ARM_AM::getAM3Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM3Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 8;
break;
}
case ARMII::AddrMode5: {
ImmIdx = i+1;
InstrOffs = ARM_AM::getAM5Offset(MI.getOperand(ImmIdx).getImm());
if (ARM_AM::getAM5Op(MI.getOperand(ImmIdx).getImm()) == ARM_AM::sub)
InstrOffs *= -1;
NumBits = 8;
Scale = 4;
break;
}
case ARMII::AddrModeTs: {
ImmIdx = i+1;
InstrOffs = MI.getOperand(ImmIdx).getImm();
NumBits = (FrameReg == ARM::SP) ? 8 : 5;
Scale = 4;
break;
}
default:
assert(0 && "Unsupported addressing mode!");
abort();
break;
}
Offset += InstrOffs * Scale;
assert((Offset & (Scale-1)) == 0 && "Can't encode this offset!");
if (Offset < 0 && !isThumb) {
Offset = -Offset;
isSub = true;
}
// Common case: small offset, fits into instruction.
MachineOperand &ImmOp = MI.getOperand(ImmIdx);
int ImmedOffset = Offset / Scale;
unsigned Mask = (1 << NumBits) - 1;
if ((unsigned)Offset <= Mask * Scale) {
// Replace the FrameIndex with sp
MI.getOperand(i).ChangeToRegister(FrameReg, false);
if (isSub)
ImmedOffset |= 1 << NumBits;
ImmOp.ChangeToImmediate(ImmedOffset);
return;
}
bool isThumSpillRestore = Opcode == ARM::tRestore || Opcode == ARM::tSpill;
if (AddrMode == ARMII::AddrModeTs) {
// Thumb tLDRspi, tSTRspi. These will change to instructions that use
// a different base register.
NumBits = 5;
Mask = (1 << NumBits) - 1;
}
// If this is a thumb spill / restore, we will be using a constpool load to
// materialize the offset.
if (AddrMode == ARMII::AddrModeTs && isThumSpillRestore)
ImmOp.ChangeToImmediate(0);
else {
// Otherwise, it didn't fit. Pull in what we can to simplify the immed.
ImmedOffset = ImmedOffset & Mask;
if (isSub)
ImmedOffset |= 1 << NumBits;
ImmOp.ChangeToImmediate(ImmedOffset);
Offset &= ~(Mask*Scale);
}
}
// If we get here, the immediate doesn't fit into the instruction. We folded
// as much as possible above, handle the rest, providing a register that is
// SP+LargeImm.
assert(Offset && "This code isn't needed if offset already handled!");
if (isThumb) {
if (TII.isLoad(Opcode)) {
// Use the destination register to materialize sp + offset.
unsigned TmpReg = MI.getOperand(0).getReg();
bool UseRR = false;
if (Opcode == ARM::tRestore) {
if (FrameReg == ARM::SP)
emitThumbRegPlusImmInReg(MBB, II, TmpReg, FrameReg,Offset,false,TII);
else {
emitLoadConstPool(MBB, II, TmpReg, Offset, ARMCC::AL, 0, TII, true);
UseRR = true;
}
} else
emitThumbRegPlusImmediate(MBB, II, TmpReg, FrameReg, Offset, TII);
MI.setInstrDescriptor(TII.get(ARM::tLDR));
MI.getOperand(i).ChangeToRegister(TmpReg, false, false, true);
if (UseRR)
MI.addRegOperand(FrameReg, false); // Use [reg, reg] addrmode.
else
MI.addRegOperand(0, false); // tLDR has an extra register operand.
} else if (TII.isStore(Opcode)) {
// FIXME! This is horrific!!! We need register scavenging.
// Our temporary workaround has marked r3 unavailable. Of course, r3 is
// also a ABI register so it's possible that is is the register that is
// being storing here. If that's the case, we do the following:
// r12 = r2
// Use r2 to materialize sp + offset
// str r3, r2
// r2 = r12
unsigned ValReg = MI.getOperand(0).getReg();
unsigned TmpReg = ARM::R3;
bool UseRR = false;
if (ValReg == ARM::R3) {
BuildMI(MBB, II, TII.get(ARM::tMOVr), ARM::R12)
.addReg(ARM::R2, false, false, true);
TmpReg = ARM::R2;
}
if (TmpReg == ARM::R3 && AFI->isR3LiveIn())
BuildMI(MBB, II, TII.get(ARM::tMOVr), ARM::R12)
.addReg(ARM::R3, false, false, true);
if (Opcode == ARM::tSpill) {
if (FrameReg == ARM::SP)
emitThumbRegPlusImmInReg(MBB, II, TmpReg, FrameReg,Offset,false,TII);
else {
emitLoadConstPool(MBB, II, TmpReg, Offset, ARMCC::AL, 0, TII, true);
UseRR = true;
}
} else
emitThumbRegPlusImmediate(MBB, II, TmpReg, FrameReg, Offset, TII);
MI.setInstrDescriptor(TII.get(ARM::tSTR));
MI.getOperand(i).ChangeToRegister(TmpReg, false, false, true);
if (UseRR)
MI.addRegOperand(FrameReg, false); // Use [reg, reg] addrmode.
else
MI.addRegOperand(0, false); // tSTR has an extra register operand.
MachineBasicBlock::iterator NII = next(II);
if (ValReg == ARM::R3)
BuildMI(MBB, NII, TII.get(ARM::tMOVr), ARM::R2)
.addReg(ARM::R12, false, false, true);
if (TmpReg == ARM::R3 && AFI->isR3LiveIn())
BuildMI(MBB, NII, TII.get(ARM::tMOVr), ARM::R3)
.addReg(ARM::R12, false, false, true);
} else
assert(false && "Unexpected opcode!");
} else {
// Insert a set of r12 with the full address: r12 = sp + offset
// If the offset we have is too large to fit into the instruction, we need
// to form it with a series of ADDri's. Do this by taking 8-bit chunks
// out of 'Offset'.
unsigned ScratchReg = findScratchRegister(RS, &ARM::GPRRegClass, AFI);
if (ScratchReg == 0)
// No register is "free". Scavenge a register.
ScratchReg = RS->scavengeRegister(&ARM::GPRRegClass, II, SPAdj);
int PIdx = MI.findFirstPredOperandIdx();
ARMCC::CondCodes Pred = (PIdx == -1)
? ARMCC::AL : (ARMCC::CondCodes)MI.getOperand(PIdx).getImmedValue();
unsigned PredReg = (PIdx == -1) ? 0 : MI.getOperand(PIdx+1).getReg();
emitARMRegPlusImmediate(MBB, II, ScratchReg, FrameReg,
isSub ? -Offset : Offset, Pred, PredReg, TII);
MI.getOperand(i).ChangeToRegister(ScratchReg, false, false, true);
}
}
static unsigned estimateStackSize(MachineFunction &MF, MachineFrameInfo *MFI) {
const MachineFrameInfo *FFI = MF.getFrameInfo();
int Offset = 0;
for (int i = FFI->getObjectIndexBegin(); i != 0; ++i) {
int FixedOff = -FFI->getObjectOffset(i);
if (FixedOff > Offset) Offset = FixedOff;
}
for (unsigned i = 0, e = FFI->getObjectIndexEnd(); i != e; ++i) {
Offset += FFI->getObjectSize(i);
unsigned Align = FFI->getObjectAlignment(i);
// Adjust to alignment boundary
Offset = (Offset+Align-1)/Align*Align;
}
return (unsigned)Offset;
}
void
ARMRegisterInfo::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
RegScavenger *RS) const {
// This tells PEI to spill the FP as if it is any other callee-save register
// to take advantage the eliminateFrameIndex machinery. This also ensures it
// is spilled in the order specified by getCalleeSavedRegs() to make it easier
// to combine multiple loads / stores.
bool CanEliminateFrame = true;
bool CS1Spilled = false;
bool LRSpilled = false;
unsigned NumGPRSpills = 0;
SmallVector<unsigned, 4> UnspilledCS1GPRs;
SmallVector<unsigned, 4> UnspilledCS2GPRs;
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
// Don't spill FP if the frame can be eliminated. This is determined
// by scanning the callee-save registers to see if any is used.
const unsigned *CSRegs = getCalleeSavedRegs();
const TargetRegisterClass* const *CSRegClasses = getCalleeSavedRegClasses();
for (unsigned i = 0; CSRegs[i]; ++i) {
unsigned Reg = CSRegs[i];
bool Spilled = false;
if (MF.isPhysRegUsed(Reg)) {
AFI->setCSRegisterIsSpilled(Reg);
Spilled = true;
CanEliminateFrame = false;
} else {
// Check alias registers too.
for (const unsigned *Aliases = getAliasSet(Reg); *Aliases; ++Aliases) {
if (MF.isPhysRegUsed(*Aliases)) {
Spilled = true;
CanEliminateFrame = false;
}
}
}
if (CSRegClasses[i] == &ARM::GPRRegClass) {
if (Spilled) {
NumGPRSpills++;
if (!STI.isTargetDarwin()) {
if (Reg == ARM::LR)
LRSpilled = true;
CS1Spilled = true;
continue;
}
// Keep track if LR and any of R4, R5, R6, and R7 is spilled.
switch (Reg) {
case ARM::LR:
LRSpilled = true;
// Fallthrough
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
CS1Spilled = true;
break;
default:
break;
}
} else {
if (!STI.isTargetDarwin()) {
UnspilledCS1GPRs.push_back(Reg);
continue;
}
switch (Reg) {
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
UnspilledCS1GPRs.push_back(Reg);
break;
default:
UnspilledCS2GPRs.push_back(Reg);
break;
}
}
}
}
bool ForceLRSpill = false;
if (!LRSpilled && AFI->isThumbFunction()) {
unsigned FnSize = ARM::GetFunctionSize(MF);
// Force LR to be spilled if the Thumb function size is > 2048. This enables
// use of BL to implement far jump. If it turns out that it's not needed
// then the branch fix up path will undo it.
if (FnSize >= (1 << 11)) {
CanEliminateFrame = false;
ForceLRSpill = true;
}
}
bool ExtraCSSpill = false;
if (!CanEliminateFrame || hasFP(MF)) {
AFI->setHasStackFrame(true);
// If LR is not spilled, but at least one of R4, R5, R6, and R7 is spilled.
// Spill LR as well so we can fold BX_RET to the registers restore (LDM).
if (!LRSpilled && CS1Spilled) {
MF.setPhysRegUsed(ARM::LR);
AFI->setCSRegisterIsSpilled(ARM::LR);
NumGPRSpills++;
UnspilledCS1GPRs.erase(std::find(UnspilledCS1GPRs.begin(),
UnspilledCS1GPRs.end(), (unsigned)ARM::LR));
ForceLRSpill = false;
ExtraCSSpill = true;
}
// Darwin ABI requires FP to point to the stack slot that contains the
// previous FP.
if (STI.isTargetDarwin() || hasFP(MF)) {
MF.setPhysRegUsed(FramePtr);
NumGPRSpills++;
}
// If stack and double are 8-byte aligned and we are spilling an odd number
// of GPRs. Spill one extra callee save GPR so we won't have to pad between
// the integer and double callee save areas.
unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
if (TargetAlign == 8 && (NumGPRSpills & 1)) {
if (CS1Spilled && !UnspilledCS1GPRs.empty()) {
for (unsigned i = 0, e = UnspilledCS1GPRs.size(); i != e; ++i) {
unsigned Reg = UnspilledCS1GPRs[i];
// Don't spiil high register if the function is thumb
if (!AFI->isThumbFunction() || isLowRegister(Reg) || Reg == ARM::LR) {
MF.setPhysRegUsed(Reg);
AFI->setCSRegisterIsSpilled(Reg);
if (!isReservedReg(MF, Reg))
ExtraCSSpill = true;
break;
}
}
} else if (!UnspilledCS2GPRs.empty() &&
!AFI->isThumbFunction()) {
unsigned Reg = UnspilledCS2GPRs.front();
MF.setPhysRegUsed(Reg);
AFI->setCSRegisterIsSpilled(Reg);
if (!isReservedReg(MF, Reg))
ExtraCSSpill = true;
}
}
// Estimate if we might need to scavenge a register at some point in order
// to materialize a stack offset. If so, either spill one additiona
// callee-saved register or reserve a special spill slot to facilitate
// register scavenging.
if (RS && !ExtraCSSpill && !AFI->isThumbFunction()) {
MachineFrameInfo *MFI = MF.getFrameInfo();
unsigned Size = estimateStackSize(MF, MFI);
unsigned Limit = (1 << 12) - 1;
for (MachineFunction::iterator BB = MF.begin(),E = MF.end();BB != E; ++BB)
for (MachineBasicBlock::iterator I= BB->begin(); I != BB->end(); ++I) {
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (I->getOperand(i).isFrameIndex()) {
unsigned Opcode = I->getOpcode();
const TargetInstrDescriptor &Desc = TII.get(Opcode);
unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask);
if (AddrMode == ARMII::AddrMode3) {
Limit = (1 << 8) - 1;
goto DoneEstimating;
} else if (AddrMode == ARMII::AddrMode5) {
unsigned ThisLimit = ((1 << 8) - 1) * 4;
if (ThisLimit < Limit)
Limit = ThisLimit;
}
}
}
DoneEstimating:
if (Size >= Limit) {
// If any non-reserved CS register isn't spilled, just spill one or two
// extra. That should take care of it!
unsigned NumExtras = TargetAlign / 4;
SmallVector<unsigned, 2> Extras;
while (NumExtras && !UnspilledCS1GPRs.empty()) {
unsigned Reg = UnspilledCS1GPRs.back();
UnspilledCS1GPRs.pop_back();
if (!isReservedReg(MF, Reg)) {
Extras.push_back(Reg);
NumExtras--;
}
}
while (NumExtras && !UnspilledCS2GPRs.empty()) {
unsigned Reg = UnspilledCS2GPRs.back();
UnspilledCS2GPRs.pop_back();
if (!isReservedReg(MF, Reg)) {
Extras.push_back(Reg);
NumExtras--;
}
}
if (Extras.size() && NumExtras == 0) {
for (unsigned i = 0, e = Extras.size(); i != e; ++i) {
MF.setPhysRegUsed(Extras[i]);
AFI->setCSRegisterIsSpilled(Extras[i]);
}
} else {
// Reserve a slot closest to SP or frame pointer.
const TargetRegisterClass *RC = &ARM::GPRRegClass;
RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
RC->getAlignment()));
}
}
}
}
if (ForceLRSpill) {
MF.setPhysRegUsed(ARM::LR);
AFI->setCSRegisterIsSpilled(ARM::LR);
AFI->setLRIsSpilledForFarJump(true);
}
}
/// Move iterator pass the next bunch of callee save load / store ops for
/// the particular spill area (1: integer area 1, 2: integer area 2,
/// 3: fp area, 0: don't care).
static void movePastCSLoadStoreOps(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI,
int Opc, unsigned Area,
const ARMSubtarget &STI) {
while (MBBI != MBB.end() &&
MBBI->getOpcode() == Opc && MBBI->getOperand(1).isFrameIndex()) {
if (Area != 0) {
bool Done = false;
unsigned Category = 0;
switch (MBBI->getOperand(0).getReg()) {
case ARM::R4: case ARM::R5: case ARM::R6: case ARM::R7:
case ARM::LR:
Category = 1;
break;
case ARM::R8: case ARM::R9: case ARM::R10: case ARM::R11:
Category = STI.isTargetDarwin() ? 2 : 1;
break;
case ARM::D8: case ARM::D9: case ARM::D10: case ARM::D11:
case ARM::D12: case ARM::D13: case ARM::D14: case ARM::D15:
Category = 3;
break;
default:
Done = true;
break;
}
if (Done || Category != Area)
break;
}
++MBBI;
}
}
void ARMRegisterInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front();
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
bool isThumb = AFI->isThumbFunction();
unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
unsigned NumBytes = MFI->getStackSize();
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
if (isThumb) {
// Check if R3 is live in. It might have to be used as a scratch register.
for (MachineFunction::livein_iterator I=MF.livein_begin(),E=MF.livein_end();
I != E; ++I) {
if ((*I).first == ARM::R3) {
AFI->setR3IsLiveIn(true);
break;
}
}
// Thumb add/sub sp, imm8 instructions implicitly multiply the offset by 4.
NumBytes = (NumBytes + 3) & ~3;
MFI->setStackSize(NumBytes);
}
// Determine the sizes of each callee-save spill areas and record which frame
// belongs to which callee-save spill areas.
unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
int FramePtrSpillFI = 0;
if (VARegSaveSize)
emitSPUpdate(MBB, MBBI, -VARegSaveSize, ARMCC::AL, 0, isThumb, TII);
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(MBB, MBBI, -NumBytes, ARMCC::AL, 0, isThumb, TII);
return;
}
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
break;
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
if (STI.isTargetDarwin()) {
AFI->addGPRCalleeSavedArea2Frame(FI);
GPRCS2Size += 4;
} else {
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
}
break;
default:
AFI->addDPRCalleeSavedAreaFrame(FI);
DPRCSSize += 8;
}
}
if (!isThumb) {
// Build the new SUBri to adjust SP for integer callee-save spill area 1.
emitSPUpdate(MBB, MBBI, -GPRCS1Size, ARMCC::AL, 0, isThumb, TII);
movePastCSLoadStoreOps(MBB, MBBI, ARM::STR, 1, STI);
} else if (MBBI != MBB.end() && MBBI->getOpcode() == ARM::tPUSH)
++MBBI;
// Darwin ABI requires FP to point to the stack slot that contains the
// previous FP.
if (STI.isTargetDarwin() || hasFP(MF)) {
MachineInstrBuilder MIB =
BuildMI(MBB, MBBI, TII.get(isThumb ? ARM::tADDrSPi : ARM::ADDri),FramePtr)
.addFrameIndex(FramePtrSpillFI).addImm(0);
if (!isThumb) AddDefaultCC(AddDefaultPred(MIB));
}
if (!isThumb) {
// Build the new SUBri to adjust SP for integer callee-save spill area 2.
emitSPUpdate(MBB, MBBI, -GPRCS2Size, ARMCC::AL, 0, false, TII);
// Build the new SUBri to adjust SP for FP callee-save spill area.
movePastCSLoadStoreOps(MBB, MBBI, ARM::STR, 2, STI);
emitSPUpdate(MBB, MBBI, -DPRCSSize, ARMCC::AL, 0, false, TII);
}
// Determine starting offsets of spill areas.
unsigned DPRCSOffset = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) + NumBytes);
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
NumBytes = DPRCSOffset;
if (NumBytes) {
// Insert it after all the callee-save spills.
if (!isThumb)
movePastCSLoadStoreOps(MBB, MBBI, ARM::FSTD, 3, STI);
emitSPUpdate(MBB, MBBI, -NumBytes, ARMCC::AL, 0, isThumb, TII);
}
if(STI.isTargetELF() && hasFP(MF)) {
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
}
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
}
static bool isCalleeSavedRegister(unsigned Reg, const unsigned *CSRegs) {
for (unsigned i = 0; CSRegs[i]; ++i)
if (Reg == CSRegs[i])
return true;
return false;
}
static bool isCSRestore(MachineInstr *MI, const unsigned *CSRegs) {
return ((MI->getOpcode() == ARM::FLDD ||
MI->getOpcode() == ARM::LDR ||
MI->getOpcode() == ARM::tRestore) &&
MI->getOperand(1).isFrameIndex() &&
isCalleeSavedRegister(MI->getOperand(0).getReg(), CSRegs));
}
void ARMRegisterInfo::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = prior(MBB.end());
assert((MBBI->getOpcode() == ARM::BX_RET ||
MBBI->getOpcode() == ARM::tBX_RET ||
MBBI->getOpcode() == ARM::tPOP_RET) &&
"Can only insert epilog into returning blocks");
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
bool isThumb = AFI->isThumbFunction();
unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
int NumBytes = (int)MFI->getStackSize();
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(MBB, MBBI, NumBytes, ARMCC::AL, 0, isThumb, TII);
} else {
// Unwind MBBI to point to first LDR / FLDD.
const unsigned *CSRegs = getCalleeSavedRegs();
if (MBBI != MBB.begin()) {
do
--MBBI;
while (MBBI != MBB.begin() && isCSRestore(MBBI, CSRegs));
if (!isCSRestore(MBBI, CSRegs))
++MBBI;
}
// Move SP to start of FP callee save spill area.
NumBytes -= (AFI->getGPRCalleeSavedArea1Size() +
AFI->getGPRCalleeSavedArea2Size() +
AFI->getDPRCalleeSavedAreaSize());
if (isThumb) {
if (hasFP(MF)) {
NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
// Reset SP based on frame pointer only if the stack frame extends beyond
// frame pointer stack slot or target is ELF and the function has FP.
if (NumBytes)
emitThumbRegPlusImmediate(MBB, MBBI, ARM::SP, FramePtr, -NumBytes, TII);
else
BuildMI(MBB, MBBI, TII.get(ARM::tMOVr), ARM::SP).addReg(FramePtr);
} else {
if (MBBI->getOpcode() == ARM::tBX_RET &&
&MBB.front() != MBBI &&
prior(MBBI)->getOpcode() == ARM::tPOP) {
MachineBasicBlock::iterator PMBBI = prior(MBBI);
emitSPUpdate(MBB, PMBBI, NumBytes, ARMCC::AL, 0, isThumb, TII);
} else
emitSPUpdate(MBB, MBBI, NumBytes, ARMCC::AL, 0, isThumb, TII);
}
} else {
// Darwin ABI requires FP to point to the stack slot that contains the
// previous FP.
if ((STI.isTargetDarwin() && NumBytes) || hasFP(MF)) {
NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
// Reset SP based on frame pointer only if the stack frame extends beyond
// frame pointer stack slot or target is ELF and the function has FP.
if (AFI->getGPRCalleeSavedArea2Size() ||
AFI->getDPRCalleeSavedAreaSize() ||
AFI->getDPRCalleeSavedAreaOffset()||
hasFP(MF))
if (NumBytes)
BuildMI(MBB, MBBI, TII.get(ARM::SUBri), ARM::SP).addReg(FramePtr)
.addImm(NumBytes)
.addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
else
BuildMI(MBB, MBBI, TII.get(ARM::MOVr), ARM::SP).addReg(FramePtr)
.addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
} else if (NumBytes) {
emitSPUpdate(MBB, MBBI, NumBytes, ARMCC::AL, 0, false, TII);
}
// Move SP to start of integer callee save spill area 2.
movePastCSLoadStoreOps(MBB, MBBI, ARM::FLDD, 3, STI);
emitSPUpdate(MBB, MBBI, AFI->getDPRCalleeSavedAreaSize(), ARMCC::AL, 0,
false, TII);
// Move SP to start of integer callee save spill area 1.
movePastCSLoadStoreOps(MBB, MBBI, ARM::LDR, 2, STI);
emitSPUpdate(MBB, MBBI, AFI->getGPRCalleeSavedArea2Size(), ARMCC::AL, 0,
false, TII);
// Move SP to SP upon entry to the function.
movePastCSLoadStoreOps(MBB, MBBI, ARM::LDR, 1, STI);
emitSPUpdate(MBB, MBBI, AFI->getGPRCalleeSavedArea1Size(), ARMCC::AL, 0,
false, TII);
}
}
if (VARegSaveSize) {
if (isThumb)
// Epilogue for vararg functions: pop LR to R3 and branch off it.
// FIXME: Verify this is still ok when R3 is no longer being reserved.
BuildMI(MBB, MBBI, TII.get(ARM::tPOP)).addReg(ARM::R3);
emitSPUpdate(MBB, MBBI, VARegSaveSize, ARMCC::AL, 0, isThumb, TII);
if (isThumb) {
BuildMI(MBB, MBBI, TII.get(ARM::tBX_RET_vararg)).addReg(ARM::R3);
MBB.erase(MBBI);
}
}
}
unsigned ARMRegisterInfo::getRARegister() const {
return ARM::LR;
}
unsigned ARMRegisterInfo::getFrameRegister(MachineFunction &MF) const {
if (STI.isTargetDarwin() || hasFP(MF))
return (STI.useThumbBacktraces() || STI.isThumb()) ? ARM::R7 : ARM::R11;
else
return ARM::SP;
}
unsigned ARMRegisterInfo::getEHExceptionRegister() const {
assert(0 && "What is the exception register");
return 0;
}
unsigned ARMRegisterInfo::getEHHandlerRegister() const {
assert(0 && "What is the exception handler register");
return 0;
}
int ARMRegisterInfo::getDwarfRegNum(unsigned RegNum, bool isEH) const {
assert(0 && "What is the dwarf register number");
return -1;
}
#include "ARMGenRegisterInfo.inc"
|