llvm.org GIT mirror llvm / 7d7d996 lib / Target / ARM / ARMFrameLowering.cpp
7d7d996

Tree @7d7d996 (Download .tar.gz)

ARMFrameLowering.cpp @7d7d996raw · history · blame

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

#include "ARMFrameLowering.h"
#include "ARMBaseInstrInfo.h"
#include "ARMBaseRegisterInfo.h"
#include "ARMMachineFunctionInfo.h"
#include "MCTargetDesc/ARMAddressingModes.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RegisterScavenging.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Function.h"
#include "llvm/MC/MCContext.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Target/TargetOptions.h"

using namespace llvm;

static cl::opt<bool>
SpillAlignedNEONRegs("align-neon-spills", cl::Hidden, cl::init(true),
                     cl::desc("Align ARM NEON spills in prolog and epilog"));

static MachineBasicBlock::iterator
skipAlignedDPRCS2Spills(MachineBasicBlock::iterator MI,
                        unsigned NumAlignedDPRCS2Regs);

/// 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 ARMFrameLowering::hasFP(const MachineFunction &MF) const {
  const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();

  // iOS requires FP not to be clobbered for backtracing purpose.
  if (STI.isTargetIOS())
    return true;

  const MachineFrameInfo *MFI = MF.getFrameInfo();
  // Always eliminate non-leaf frame pointers.
  return ((MF.getTarget().Options.DisableFramePointerElim(MF) &&
           MFI->hasCalls()) ||
          RegInfo->needsStackRealignment(MF) ||
          MFI->hasVarSizedObjects() ||
          MFI->isFrameAddressTaken());
}

/// 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 ARMFrameLowering::hasReservedCallFrame(const MachineFunction &MF) const {
  const MachineFrameInfo *FFI = MF.getFrameInfo();
  unsigned CFSize = FFI->getMaxCallFrameSize();
  // 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 (CFSize >= ((1 << 12) - 1) / 2)  // Half of imm12
    return false;

  return !MF.getFrameInfo()->hasVarSizedObjects();
}

/// canSimplifyCallFramePseudos - If there is a reserved call frame, the
/// call frame pseudos can be simplified.  Unlike most targets, having a FP
/// is not sufficient here since we still may reference some objects via SP
/// even when FP is available in Thumb2 mode.
bool
ARMFrameLowering::canSimplifyCallFramePseudos(const MachineFunction &MF) const {
  return hasReservedCallFrame(MF) || MF.getFrameInfo()->hasVarSizedObjects();
}

static bool isCSRestore(MachineInstr *MI,
                        const ARMBaseInstrInfo &TII,
                        const uint16_t *CSRegs) {
  // Integer spill area is handled with "pop".
  if (isPopOpcode(MI->getOpcode())) {
    // The first two operands are predicates. The last two are
    // imp-def and imp-use of SP. Check everything in between.
    for (int i = 5, e = MI->getNumOperands(); i != e; ++i)
      if (!isCalleeSavedRegister(MI->getOperand(i).getReg(), CSRegs))
        return false;
    return true;
  }
  if ((MI->getOpcode() == ARM::LDR_POST_IMM ||
       MI->getOpcode() == ARM::LDR_POST_REG ||
       MI->getOpcode() == ARM::t2LDR_POST) &&
      isCalleeSavedRegister(MI->getOperand(0).getReg(), CSRegs) &&
      MI->getOperand(1).getReg() == ARM::SP)
    return true;

  return false;
}

static void emitRegPlusImmediate(bool isARM, MachineBasicBlock &MBB,
                                 MachineBasicBlock::iterator &MBBI, DebugLoc dl,
                                 const ARMBaseInstrInfo &TII, unsigned DestReg,
                                 unsigned SrcReg, int NumBytes,
                                 unsigned MIFlags = MachineInstr::NoFlags,
                                 ARMCC::CondCodes Pred = ARMCC::AL,
                                 unsigned PredReg = 0) {
  if (isARM)
    emitARMRegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
                            Pred, PredReg, TII, MIFlags);
  else
    emitT2RegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
                           Pred, PredReg, TII, MIFlags);
}

static void emitSPUpdate(bool isARM, MachineBasicBlock &MBB,
                         MachineBasicBlock::iterator &MBBI, DebugLoc dl,
                         const ARMBaseInstrInfo &TII, int NumBytes,
                         unsigned MIFlags = MachineInstr::NoFlags,
                         ARMCC::CondCodes Pred = ARMCC::AL,
                         unsigned PredReg = 0) {
  emitRegPlusImmediate(isARM, MBB, MBBI, dl, TII, ARM::SP, ARM::SP, NumBytes,
                       MIFlags, Pred, PredReg);
}

static int sizeOfSPAdjustment(const MachineInstr *MI) {
  assert(MI->getOpcode() == ARM::VSTMDDB_UPD);
  int count = 0;
  // ARM and Thumb2 push/pop insts have explicit "sp, sp" operands (+
  // pred) so the list starts at 4.
  for (int i = MI->getNumOperands() - 1; i >= 4; --i)
    count += 8;
  return count;
}

void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
  MachineBasicBlock &MBB = MF.front();
  MachineBasicBlock::iterator MBBI = MBB.begin();
  MachineFrameInfo  *MFI = MF.getFrameInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  MachineModuleInfo &MMI = MF.getMMI();
  MCContext &Context = MMI.getContext();
  const MCRegisterInfo *MRI = Context.getRegisterInfo();
  const ARMBaseRegisterInfo *RegInfo =
    static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
  const ARMBaseInstrInfo &TII =
    *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
  assert(!AFI->isThumb1OnlyFunction() &&
         "This emitPrologue does not support Thumb1!");
  bool isARM = !AFI->isThumbFunction();
  unsigned Align = MF.getTarget().getFrameLowering()->getStackAlignment();
  unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize(Align);
  unsigned NumBytes = MFI->getStackSize();
  const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
  DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
  unsigned FramePtr = RegInfo->getFrameRegister(MF);
  int CFAOffset = 0;

  // 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;
  int D8SpillFI = 0;

  // All calls are tail calls in GHC calling conv, and functions have no
  // prologue/epilogue.
  if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
    return;

  // Allocate the vararg register save area.
  if (ArgRegsSaveSize) {
    emitSPUpdate(isARM, MBB, MBBI, dl, TII, -ArgRegsSaveSize,
                 MachineInstr::FrameSetup);
    CFAOffset -= ArgRegsSaveSize;
    unsigned CFIIndex = MMI.addFrameInst(
        MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
    BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
        .addCFIIndex(CFIIndex);
  }

  if (!AFI->hasStackFrame()) {
    if (NumBytes - ArgRegsSaveSize != 0) {
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, -(NumBytes - ArgRegsSaveSize),
                   MachineInstr::FrameSetup);
      CFAOffset -= NumBytes - ArgRegsSaveSize;
      unsigned CFIIndex = MMI.addFrameInst(
          MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
      BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex);
    }
    return;
  }

  // Determine spill area sizes.
  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
    unsigned Reg = CSI[i].getReg();
    int FI = CSI[i].getFrameIdx();
    switch (Reg) {
    case ARM::R8:
    case ARM::R9:
    case ARM::R10:
    case ARM::R11:
    case ARM::R12:
      if (STI.isTargetMachO()) {
        GPRCS2Size += 4;
        break;
      }
      // fallthrough
    case ARM::R0:
    case ARM::R1:
    case ARM::R2:
    case ARM::R3:
    case ARM::R4:
    case ARM::R5:
    case ARM::R6:
    case ARM::R7:
    case ARM::LR:
      if (Reg == FramePtr)
        FramePtrSpillFI = FI;
      GPRCS1Size += 4;
      break;
    default:
      // This is a DPR. Exclude the aligned DPRCS2 spills.
      if (Reg == ARM::D8)
        D8SpillFI = FI;
      if (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())
        DPRCSSize += 8;
    }
  }

  // Move past area 1.
  MachineBasicBlock::iterator LastPush = MBB.end(), GPRCS1Push, GPRCS2Push,
      DPRCSPush;
  if (GPRCS1Size > 0)
    GPRCS1Push = LastPush = MBBI++;

  // Determine starting offsets of spill areas.
  bool HasFP = hasFP(MF);
  unsigned DPRCSOffset  = NumBytes - (ArgRegsSaveSize + GPRCS1Size
                                      + GPRCS2Size + DPRCSSize);
  unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
  unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
  int FramePtrOffsetInPush = 0;
  if (HasFP) {
    FramePtrOffsetInPush = MFI->getObjectOffset(FramePtrSpillFI)
                           + GPRCS1Size + ArgRegsSaveSize;
    AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) +
                                NumBytes);
  }
  AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
  AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
  AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);

  // Move past area 2.
  if (GPRCS2Size > 0)
    GPRCS2Push = LastPush = MBBI++;

  // Move past area 3.
  if (DPRCSSize > 0) {
    DPRCSPush = MBBI;
    // Since vpush register list cannot have gaps, there may be multiple vpush
    // instructions in the prologue.
    while (MBBI->getOpcode() == ARM::VSTMDDB_UPD)
      LastPush = MBBI++;
  }

  // Move past the aligned DPRCS2 area.
  if (AFI->getNumAlignedDPRCS2Regs() > 0) {
    MBBI = skipAlignedDPRCS2Spills(MBBI, AFI->getNumAlignedDPRCS2Regs());
    // The code inserted by emitAlignedDPRCS2Spills realigns the stack, and
    // leaves the stack pointer pointing to the DPRCS2 area.
    //
    // Adjust NumBytes to represent the stack slots below the DPRCS2 area.
    NumBytes += MFI->getObjectOffset(D8SpillFI);
  } else
    NumBytes = DPRCSOffset;

  unsigned adjustedGPRCS1Size = GPRCS1Size;
  if (NumBytes) {
    // Adjust SP after all the callee-save spills.
    if (tryFoldSPUpdateIntoPushPop(STI, MF, LastPush, NumBytes)) {
      if (LastPush == GPRCS1Push) {
        FramePtrOffsetInPush += NumBytes;
        adjustedGPRCS1Size += NumBytes;
        NumBytes = 0;
      }
    } else
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
                   MachineInstr::FrameSetup);

    if (HasFP && isARM)
      // Restore from fp only in ARM mode: e.g. sub sp, r7, #24
      // Note it's not safe to do this in Thumb2 mode because it would have
      // taken two instructions:
      // mov sp, r7
      // sub sp, #24
      // If an interrupt is taken between the two instructions, then sp is in
      // an inconsistent state (pointing to the middle of callee-saved area).
      // The interrupt handler can end up clobbering the registers.
      AFI->setShouldRestoreSPFromFP(true);
  }

  if (adjustedGPRCS1Size > 0) {
    CFAOffset -= adjustedGPRCS1Size;
    unsigned CFIIndex = MMI.addFrameInst(
        MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
    MachineBasicBlock::iterator Pos = ++GPRCS1Push;
    BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
        .addCFIIndex(CFIIndex);
    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
           E = CSI.end(); I != E; ++I) {
      unsigned Reg = I->getReg();
      int FI = I->getFrameIdx();
      switch (Reg) {
      case ARM::R8:
      case ARM::R9:
      case ARM::R10:
      case ARM::R11:
      case ARM::R12:
        if (STI.isTargetMachO())
          break;
        // fallthrough
      case ARM::R0:
      case ARM::R1:
      case ARM::R2:
      case ARM::R3:
      case ARM::R4:
      case ARM::R5:
      case ARM::R6:
      case ARM::R7:
      case ARM::LR:
        CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
            nullptr, MRI->getDwarfRegNum(Reg, true), MFI->getObjectOffset(FI)));
        BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
            .addCFIIndex(CFIIndex);
        break;
      }
    }
  }

  // Set FP to point to the stack slot that contains the previous FP.
  // For iOS, FP is R7, which has now been stored in spill area 1.
  // Otherwise, if this is not iOS, all the callee-saved registers go
  // into spill area 1, including the FP in R11.  In either case, it
  // is in area one and the adjustment needs to take place just after
  // that push.
  if (HasFP) {
    emitRegPlusImmediate(!AFI->isThumbFunction(), MBB, GPRCS1Push, dl, TII,
                         FramePtr, ARM::SP, FramePtrOffsetInPush,
                         MachineInstr::FrameSetup);
    if (FramePtrOffsetInPush) {
      CFAOffset += FramePtrOffsetInPush;
      unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfa(
          nullptr, MRI->getDwarfRegNum(FramePtr, true), CFAOffset));
      BuildMI(MBB, GPRCS1Push, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex);

    } else {
      unsigned CFIIndex =
          MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(
              nullptr, MRI->getDwarfRegNum(FramePtr, true)));
      BuildMI(MBB, GPRCS1Push, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex);
    }
  }

  if (GPRCS2Size > 0) {
    MachineBasicBlock::iterator Pos = ++GPRCS2Push;
    if (!HasFP) {
      CFAOffset -= GPRCS2Size;
      unsigned CFIIndex = MMI.addFrameInst(
          MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
      BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex);
    }
    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
           E = CSI.end(); I != E; ++I) {
      unsigned Reg = I->getReg();
      int FI = I->getFrameIdx();
      switch (Reg) {
      case ARM::R8:
      case ARM::R9:
      case ARM::R10:
      case ARM::R11:
      case ARM::R12:
        if (STI.isTargetMachO()) {
          unsigned DwarfReg =  MRI->getDwarfRegNum(Reg, true);
          unsigned Offset = MFI->getObjectOffset(FI);
          unsigned CFIIndex = MMI.addFrameInst(
              MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
          BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
              .addCFIIndex(CFIIndex);
        }
        break;
      }
    }
  }

  if (DPRCSSize > 0) {
    // Since vpush register list cannot have gaps, there may be multiple vpush
    // instructions in the prologue.
    do {
      MachineBasicBlock::iterator Push = DPRCSPush++;
      if (!HasFP) {
        CFAOffset -= sizeOfSPAdjustment(Push);;
        unsigned CFIIndex = MMI.addFrameInst(
            MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
        BuildMI(MBB, DPRCSPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
            .addCFIIndex(CFIIndex);
      }
    } while (DPRCSPush->getOpcode() == ARM::VSTMDDB_UPD);

    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
           E = CSI.end(); I != E; ++I) {
      unsigned Reg = I->getReg();
      int FI = I->getFrameIdx();
      if ((Reg >= ARM::D0 && Reg <= ARM::D31) &&
          (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())) {
        unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true);
        unsigned Offset = MFI->getObjectOffset(FI);
        unsigned CFIIndex = MMI.addFrameInst(
            MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
        BuildMI(MBB, DPRCSPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
            .addCFIIndex(CFIIndex);
      }
    }
  }

  if (NumBytes) {
    if (!HasFP) {
      CFAOffset -= NumBytes;
      unsigned CFIIndex = MMI.addFrameInst(
          MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
      BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex);
    }
  }

  if (STI.isTargetELF() && hasFP(MF))
    MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
                             AFI->getFramePtrSpillOffset());

  AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
  AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
  AFI->setDPRCalleeSavedAreaSize(DPRCSSize);

  // If we need dynamic stack realignment, do it here. Be paranoid and make
  // sure if we also have VLAs, we have a base pointer for frame access.
  // If aligned NEON registers were spilled, the stack has already been
  // realigned.
  if (!AFI->getNumAlignedDPRCS2Regs() && RegInfo->needsStackRealignment(MF)) {
    unsigned MaxAlign = MFI->getMaxAlignment();
    assert (!AFI->isThumb1OnlyFunction());
    if (!AFI->isThumbFunction()) {
      // Emit bic sp, sp, MaxAlign
      AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
                                          TII.get(ARM::BICri), ARM::SP)
                                  .addReg(ARM::SP, RegState::Kill)
                                  .addImm(MaxAlign-1)));
    } else {
      // We cannot use sp as source/dest register here, thus we're emitting the
      // following sequence:
      // mov r4, sp
      // bic r4, r4, MaxAlign
      // mov sp, r4
      // FIXME: It will be better just to find spare register here.
      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::R4)
        .addReg(ARM::SP, RegState::Kill));
      AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
                                          TII.get(ARM::t2BICri), ARM::R4)
                                  .addReg(ARM::R4, RegState::Kill)
                                  .addImm(MaxAlign-1)));
      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
        .addReg(ARM::R4, RegState::Kill));
    }

    AFI->setShouldRestoreSPFromFP(true);
  }

  // If we need a base pointer, set it up here. It's whatever the value
  // of the stack pointer is at this point. Any variable size objects
  // will be allocated after this, so we can still use the base pointer
  // to reference locals.
  // FIXME: Clarify FrameSetup flags here.
  if (RegInfo->hasBasePointer(MF)) {
    if (isARM)
      BuildMI(MBB, MBBI, dl,
              TII.get(ARM::MOVr), RegInfo->getBaseRegister())
        .addReg(ARM::SP)
        .addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
    else
      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
                             RegInfo->getBaseRegister())
        .addReg(ARM::SP));
  }

  // If the frame has variable sized objects then the epilogue must restore
  // the sp from fp. We can assume there's an FP here since hasFP already
  // checks for hasVarSizedObjects.
  if (MFI->hasVarSizedObjects())
    AFI->setShouldRestoreSPFromFP(true);
}

void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
                                    MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
  assert(MBBI->isReturn() && "Can only insert epilog into returning blocks");
  unsigned RetOpcode = MBBI->getOpcode();
  DebugLoc dl = MBBI->getDebugLoc();
  MachineFrameInfo *MFI = MF.getFrameInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();
  const ARMBaseInstrInfo &TII =
    *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
  assert(!AFI->isThumb1OnlyFunction() &&
         "This emitEpilogue does not support Thumb1!");
  bool isARM = !AFI->isThumbFunction();

  unsigned Align = MF.getTarget().getFrameLowering()->getStackAlignment();
  unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize(Align);
  int NumBytes = (int)MFI->getStackSize();
  unsigned FramePtr = RegInfo->getFrameRegister(MF);

  // All calls are tail calls in GHC calling conv, and functions have no
  // prologue/epilogue.
  if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
    return;

  if (!AFI->hasStackFrame()) {
    if (NumBytes - ArgRegsSaveSize != 0)
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes - ArgRegsSaveSize);
  } else {
    // Unwind MBBI to point to first LDR / VLDRD.
    const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
    if (MBBI != MBB.begin()) {
      do {
        --MBBI;
      } while (MBBI != MBB.begin() && isCSRestore(MBBI, TII, CSRegs));
      if (!isCSRestore(MBBI, TII, CSRegs))
        ++MBBI;
    }

    // Move SP to start of FP callee save spill area.
    NumBytes -= (ArgRegsSaveSize +
                 AFI->getGPRCalleeSavedArea1Size() +
                 AFI->getGPRCalleeSavedArea2Size() +
                 AFI->getDPRCalleeSavedAreaSize());

    // 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->shouldRestoreSPFromFP()) {
      NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
      if (NumBytes) {
        if (isARM)
          emitARMRegPlusImmediate(MBB, MBBI, dl, ARM::SP, FramePtr, -NumBytes,
                                  ARMCC::AL, 0, TII);
        else {
          // It's not possible to restore SP from FP in a single instruction.
          // For iOS, this looks like:
          // mov sp, r7
          // sub sp, #24
          // This is bad, if an interrupt is taken after the mov, sp is in an
          // inconsistent state.
          // Use the first callee-saved register as a scratch register.
          assert(MF.getRegInfo().isPhysRegUsed(ARM::R4) &&
                 "No scratch register to restore SP from FP!");
          emitT2RegPlusImmediate(MBB, MBBI, dl, ARM::R4, FramePtr, -NumBytes,
                                 ARMCC::AL, 0, TII);
          AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
                                 ARM::SP)
            .addReg(ARM::R4));
        }
      } else {
        // Thumb2 or ARM.
        if (isARM)
          BuildMI(MBB, MBBI, dl, TII.get(ARM::MOVr), ARM::SP)
            .addReg(FramePtr).addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
        else
          AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
                                 ARM::SP)
            .addReg(FramePtr));
      }
    } else if (NumBytes &&
               !tryFoldSPUpdateIntoPushPop(STI, MF, MBBI, NumBytes))
        emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);

    // Increment past our save areas.
    if (AFI->getDPRCalleeSavedAreaSize()) {
      MBBI++;
      // Since vpop register list cannot have gaps, there may be multiple vpop
      // instructions in the epilogue.
      while (MBBI->getOpcode() == ARM::VLDMDIA_UPD)
        MBBI++;
    }
    if (AFI->getGPRCalleeSavedArea2Size()) MBBI++;
    if (AFI->getGPRCalleeSavedArea1Size()) MBBI++;
  }

  if (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNri) {
    // Tail call return: adjust the stack pointer and jump to callee.
    MBBI = MBB.getLastNonDebugInstr();
    MachineOperand &JumpTarget = MBBI->getOperand(0);

    // Jump to label or value in register.
    if (RetOpcode == ARM::TCRETURNdi) {
      unsigned TCOpcode = STI.isThumb() ?
               (STI.isTargetMachO() ? ARM::tTAILJMPd : ARM::tTAILJMPdND) :
               ARM::TAILJMPd;
      MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII.get(TCOpcode));
      if (JumpTarget.isGlobal())
        MIB.addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset(),
                             JumpTarget.getTargetFlags());
      else {
        assert(JumpTarget.isSymbol());
        MIB.addExternalSymbol(JumpTarget.getSymbolName(),
                              JumpTarget.getTargetFlags());
      }

      // Add the default predicate in Thumb mode.
      if (STI.isThumb()) MIB.addImm(ARMCC::AL).addReg(0);
    } else if (RetOpcode == ARM::TCRETURNri) {
      BuildMI(MBB, MBBI, dl,
              TII.get(STI.isThumb() ? ARM::tTAILJMPr : ARM::TAILJMPr)).
        addReg(JumpTarget.getReg(), RegState::Kill);
    }

    MachineInstr *NewMI = std::prev(MBBI);
    for (unsigned i = 1, e = MBBI->getNumOperands(); i != e; ++i)
      NewMI->addOperand(MBBI->getOperand(i));

    // Delete the pseudo instruction TCRETURN.
    MBB.erase(MBBI);
    MBBI = NewMI;
  }

  if (ArgRegsSaveSize)
    emitSPUpdate(isARM, MBB, MBBI, dl, TII, ArgRegsSaveSize);
}

/// getFrameIndexReference - Provide a base+offset reference to an FI slot for
/// debug info.  It's the same as what we use for resolving the code-gen
/// references for now.  FIXME: This can go wrong when references are
/// SP-relative and simple call frames aren't used.
int
ARMFrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI,
                                         unsigned &FrameReg) const {
  return ResolveFrameIndexReference(MF, FI, FrameReg, 0);
}

int
ARMFrameLowering::ResolveFrameIndexReference(const MachineFunction &MF,
                                             int FI, unsigned &FrameReg,
                                             int SPAdj) const {
  const MachineFrameInfo *MFI = MF.getFrameInfo();
  const ARMBaseRegisterInfo *RegInfo =
    static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
  const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  int Offset = MFI->getObjectOffset(FI) + MFI->getStackSize();
  int FPOffset = Offset - AFI->getFramePtrSpillOffset();
  bool isFixed = MFI->isFixedObjectIndex(FI);

  FrameReg = ARM::SP;
  Offset += SPAdj;

  // SP can move around if there are allocas.  We may also lose track of SP
  // when emergency spilling inside a non-reserved call frame setup.
  bool hasMovingSP = !hasReservedCallFrame(MF);

  // When dynamically realigning the stack, use the frame pointer for
  // parameters, and the stack/base pointer for locals.
  if (RegInfo->needsStackRealignment(MF)) {
    assert (hasFP(MF) && "dynamic stack realignment without a FP!");
    if (isFixed) {
      FrameReg = RegInfo->getFrameRegister(MF);
      Offset = FPOffset;
    } else if (hasMovingSP) {
      assert(RegInfo->hasBasePointer(MF) &&
             "VLAs and dynamic stack alignment, but missing base pointer!");
      FrameReg = RegInfo->getBaseRegister();
    }
    return Offset;
  }

  // If there is a frame pointer, use it when we can.
  if (hasFP(MF) && AFI->hasStackFrame()) {
    // Use frame pointer to reference fixed objects. Use it for locals if
    // there are VLAs (and thus the SP isn't reliable as a base).
    if (isFixed || (hasMovingSP && !RegInfo->hasBasePointer(MF))) {
      FrameReg = RegInfo->getFrameRegister(MF);
      return FPOffset;
    } else if (hasMovingSP) {
      assert(RegInfo->hasBasePointer(MF) && "missing base pointer!");
      if (AFI->isThumb2Function()) {
        // Try to use the frame pointer if we can, else use the base pointer
        // since it's available. This is handy for the emergency spill slot, in
        // particular.
        if (FPOffset >= -255 && FPOffset < 0) {
          FrameReg = RegInfo->getFrameRegister(MF);
          return FPOffset;
        }
      }
    } else if (AFI->isThumb2Function()) {
      // Use  add <rd>, sp, #<imm8>
      //      ldr <rd>, [sp, #<imm8>]
      // if at all possible to save space.
      if (Offset >= 0 && (Offset & 3) == 0 && Offset <= 1020)
        return Offset;
      // In Thumb2 mode, the negative offset is very limited. Try to avoid
      // out of range references. ldr <rt>,[<rn>, #-<imm8>]
      if (FPOffset >= -255 && FPOffset < 0) {
        FrameReg = RegInfo->getFrameRegister(MF);
        return FPOffset;
      }
    } else if (Offset > (FPOffset < 0 ? -FPOffset : FPOffset)) {
      // Otherwise, use SP or FP, whichever is closer to the stack slot.
      FrameReg = RegInfo->getFrameRegister(MF);
      return FPOffset;
    }
  }
  // Use the base pointer if we have one.
  if (RegInfo->hasBasePointer(MF))
    FrameReg = RegInfo->getBaseRegister();
  return Offset;
}

int ARMFrameLowering::getFrameIndexOffset(const MachineFunction &MF,
                                          int FI) const {
  unsigned FrameReg;
  return getFrameIndexReference(MF, FI, FrameReg);
}

void ARMFrameLowering::emitPushInst(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator MI,
                                    const std::vector<CalleeSavedInfo> &CSI,
                                    unsigned StmOpc, unsigned StrOpc,
                                    bool NoGap,
                                    bool(*Func)(unsigned, bool),
                                    unsigned NumAlignedDPRCS2Regs,
                                    unsigned MIFlags) const {
  MachineFunction &MF = *MBB.getParent();
  const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();

  DebugLoc DL;
  if (MI != MBB.end()) DL = MI->getDebugLoc();

  SmallVector<std::pair<unsigned,bool>, 4> Regs;
  unsigned i = CSI.size();
  while (i != 0) {
    unsigned LastReg = 0;
    for (; i != 0; --i) {
      unsigned Reg = CSI[i-1].getReg();
      if (!(Func)(Reg, STI.isTargetMachO())) continue;

      // D-registers in the aligned area DPRCS2 are NOT spilled here.
      if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
        continue;

      // Add the callee-saved register as live-in unless it's LR and
      // @llvm.returnaddress is called. If LR is returned for
      // @llvm.returnaddress then it's already added to the function and
      // entry block live-in sets.
      bool isKill = true;
      if (Reg == ARM::LR) {
        if (MF.getFrameInfo()->isReturnAddressTaken() &&
            MF.getRegInfo().isLiveIn(Reg))
          isKill = false;
      }

      if (isKill)
        MBB.addLiveIn(Reg);

      // If NoGap is true, push consecutive registers and then leave the rest
      // for other instructions. e.g.
      // vpush {d8, d10, d11} -> vpush {d8}, vpush {d10, d11}
      if (NoGap && LastReg && LastReg != Reg-1)
        break;
      LastReg = Reg;
      Regs.push_back(std::make_pair(Reg, isKill));
    }

    if (Regs.empty())
      continue;
    if (Regs.size() > 1 || StrOpc== 0) {
      MachineInstrBuilder MIB =
        AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(StmOpc), ARM::SP)
                       .addReg(ARM::SP).setMIFlags(MIFlags));
      for (unsigned i = 0, e = Regs.size(); i < e; ++i)
        MIB.addReg(Regs[i].first, getKillRegState(Regs[i].second));
    } else if (Regs.size() == 1) {
      MachineInstrBuilder MIB = BuildMI(MBB, MI, DL, TII.get(StrOpc),
                                        ARM::SP)
        .addReg(Regs[0].first, getKillRegState(Regs[0].second))
        .addReg(ARM::SP).setMIFlags(MIFlags)
        .addImm(-4);
      AddDefaultPred(MIB);
    }
    Regs.clear();
  }
}

void ARMFrameLowering::emitPopInst(MachineBasicBlock &MBB,
                                   MachineBasicBlock::iterator MI,
                                   const std::vector<CalleeSavedInfo> &CSI,
                                   unsigned LdmOpc, unsigned LdrOpc,
                                   bool isVarArg, bool NoGap,
                                   bool(*Func)(unsigned, bool),
                                   unsigned NumAlignedDPRCS2Regs) const {
  MachineFunction &MF = *MBB.getParent();
  const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL = MI->getDebugLoc();
  unsigned RetOpcode = MI->getOpcode();
  bool isTailCall = (RetOpcode == ARM::TCRETURNdi ||
                     RetOpcode == ARM::TCRETURNri);
  bool isInterrupt =
      RetOpcode == ARM::SUBS_PC_LR || RetOpcode == ARM::t2SUBS_PC_LR;

  SmallVector<unsigned, 4> Regs;
  unsigned i = CSI.size();
  while (i != 0) {
    unsigned LastReg = 0;
    bool DeleteRet = false;
    for (; i != 0; --i) {
      unsigned Reg = CSI[i-1].getReg();
      if (!(Func)(Reg, STI.isTargetMachO())) continue;

      // The aligned reloads from area DPRCS2 are not inserted here.
      if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
        continue;

      if (Reg == ARM::LR && !isTailCall && !isVarArg && !isInterrupt &&
          STI.hasV5TOps()) {
        Reg = ARM::PC;
        LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
        // Fold the return instruction into the LDM.
        DeleteRet = true;
      }

      // If NoGap is true, pop consecutive registers and then leave the rest
      // for other instructions. e.g.
      // vpop {d8, d10, d11} -> vpop {d8}, vpop {d10, d11}
      if (NoGap && LastReg && LastReg != Reg-1)
        break;

      LastReg = Reg;
      Regs.push_back(Reg);
    }

    if (Regs.empty())
      continue;
    if (Regs.size() > 1 || LdrOpc == 0) {
      MachineInstrBuilder MIB =
        AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(LdmOpc), ARM::SP)
                       .addReg(ARM::SP));
      for (unsigned i = 0, e = Regs.size(); i < e; ++i)
        MIB.addReg(Regs[i], getDefRegState(true));
      if (DeleteRet) {
        MIB.copyImplicitOps(&*MI);
        MI->eraseFromParent();
      }
      MI = MIB;
    } else if (Regs.size() == 1) {
      // If we adjusted the reg to PC from LR above, switch it back here. We
      // only do that for LDM.
      if (Regs[0] == ARM::PC)
        Regs[0] = ARM::LR;
      MachineInstrBuilder MIB =
        BuildMI(MBB, MI, DL, TII.get(LdrOpc), Regs[0])
          .addReg(ARM::SP, RegState::Define)
          .addReg(ARM::SP);
      // ARM mode needs an extra reg0 here due to addrmode2. Will go away once
      // that refactoring is complete (eventually).
      if (LdrOpc == ARM::LDR_POST_REG || LdrOpc == ARM::LDR_POST_IMM) {
        MIB.addReg(0);
        MIB.addImm(ARM_AM::getAM2Opc(ARM_AM::add, 4, ARM_AM::no_shift));
      } else
        MIB.addImm(4);
      AddDefaultPred(MIB);
    }
    Regs.clear();
  }
}

/// Emit aligned spill instructions for NumAlignedDPRCS2Regs D-registers
/// starting from d8.  Also insert stack realignment code and leave the stack
/// pointer pointing to the d8 spill slot.
static void emitAlignedDPRCS2Spills(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator MI,
                                    unsigned NumAlignedDPRCS2Regs,
                                    const std::vector<CalleeSavedInfo> &CSI,
                                    const TargetRegisterInfo *TRI) {
  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL = MI->getDebugLoc();
  const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
  MachineFrameInfo &MFI = *MF.getFrameInfo();

  // Mark the D-register spill slots as properly aligned.  Since MFI computes
  // stack slot layout backwards, this can actually mean that the d-reg stack
  // slot offsets can be wrong. The offset for d8 will always be correct.
  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
    unsigned DNum = CSI[i].getReg() - ARM::D8;
    if (DNum >= 8)
      continue;
    int FI = CSI[i].getFrameIdx();
    // The even-numbered registers will be 16-byte aligned, the odd-numbered
    // registers will be 8-byte aligned.
    MFI.setObjectAlignment(FI, DNum % 2 ? 8 : 16);

    // The stack slot for D8 needs to be maximally aligned because this is
    // actually the point where we align the stack pointer.  MachineFrameInfo
    // computes all offsets relative to the incoming stack pointer which is a
    // bit weird when realigning the stack.  Any extra padding for this
    // over-alignment is not realized because the code inserted below adjusts
    // the stack pointer by numregs * 8 before aligning the stack pointer.
    if (DNum == 0)
      MFI.setObjectAlignment(FI, MFI.getMaxAlignment());
  }

  // Move the stack pointer to the d8 spill slot, and align it at the same
  // time. Leave the stack slot address in the scratch register r4.
  //
  //   sub r4, sp, #numregs * 8
  //   bic r4, r4, #align - 1
  //   mov sp, r4
  //
  bool isThumb = AFI->isThumbFunction();
  assert(!AFI->isThumb1OnlyFunction() && "Can't realign stack for thumb1");
  AFI->setShouldRestoreSPFromFP(true);

  // sub r4, sp, #numregs * 8
  // The immediate is <= 64, so it doesn't need any special encoding.
  unsigned Opc = isThumb ? ARM::t2SUBri : ARM::SUBri;
  AddDefaultCC(AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
                              .addReg(ARM::SP)
                              .addImm(8 * NumAlignedDPRCS2Regs)));

  // bic r4, r4, #align-1
  Opc = isThumb ? ARM::t2BICri : ARM::BICri;
  unsigned MaxAlign = MF.getFrameInfo()->getMaxAlignment();
  AddDefaultCC(AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
                              .addReg(ARM::R4, RegState::Kill)
                              .addImm(MaxAlign - 1)));

  // mov sp, r4
  // The stack pointer must be adjusted before spilling anything, otherwise
  // the stack slots could be clobbered by an interrupt handler.
  // Leave r4 live, it is used below.
  Opc = isThumb ? ARM::tMOVr : ARM::MOVr;
  MachineInstrBuilder MIB = BuildMI(MBB, MI, DL, TII.get(Opc), ARM::SP)
                            .addReg(ARM::R4);
  MIB = AddDefaultPred(MIB);
  if (!isThumb)
    AddDefaultCC(MIB);

  // Now spill NumAlignedDPRCS2Regs registers starting from d8.
  // r4 holds the stack slot address.
  unsigned NextReg = ARM::D8;

  // 16-byte aligned vst1.64 with 4 d-regs and address writeback.
  // The writeback is only needed when emitting two vst1.64 instructions.
  if (NumAlignedDPRCS2Regs >= 6) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    MBB.addLiveIn(SupReg);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Qwb_fixed),
                           ARM::R4)
                   .addReg(ARM::R4, RegState::Kill).addImm(16)
                   .addReg(NextReg)
                   .addReg(SupReg, RegState::ImplicitKill));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // We won't modify r4 beyond this point.  It currently points to the next
  // register to be spilled.
  unsigned R4BaseReg = NextReg;

  // 16-byte aligned vst1.64 with 4 d-regs, no writeback.
  if (NumAlignedDPRCS2Regs >= 4) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    MBB.addLiveIn(SupReg);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Q))
                   .addReg(ARM::R4).addImm(16).addReg(NextReg)
                   .addReg(SupReg, RegState::ImplicitKill));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // 16-byte aligned vst1.64 with 2 d-regs.
  if (NumAlignedDPRCS2Regs >= 2) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QPRRegClass);
    MBB.addLiveIn(SupReg);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1q64))
                   .addReg(ARM::R4).addImm(16).addReg(SupReg));
    NextReg += 2;
    NumAlignedDPRCS2Regs -= 2;
  }

  // Finally, use a vanilla vstr.64 for the odd last register.
  if (NumAlignedDPRCS2Regs) {
    MBB.addLiveIn(NextReg);
    // vstr.64 uses addrmode5 which has an offset scale of 4.
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VSTRD))
                   .addReg(NextReg)
                   .addReg(ARM::R4).addImm((NextReg-R4BaseReg)*2));
  }

  // The last spill instruction inserted should kill the scratch register r4.
  std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}

/// Skip past the code inserted by emitAlignedDPRCS2Spills, and return an
/// iterator to the following instruction.
static MachineBasicBlock::iterator
skipAlignedDPRCS2Spills(MachineBasicBlock::iterator MI,
                        unsigned NumAlignedDPRCS2Regs) {
  //   sub r4, sp, #numregs * 8
  //   bic r4, r4, #align - 1
  //   mov sp, r4
  ++MI; ++MI; ++MI;
  assert(MI->mayStore() && "Expecting spill instruction");

  // These switches all fall through.
  switch(NumAlignedDPRCS2Regs) {
  case 7:
    ++MI;
    assert(MI->mayStore() && "Expecting spill instruction");
  default:
    ++MI;
    assert(MI->mayStore() && "Expecting spill instruction");
  case 1:
  case 2:
  case 4:
    assert(MI->killsRegister(ARM::R4) && "Missed kill flag");
    ++MI;
  }
  return MI;
}

/// Emit aligned reload instructions for NumAlignedDPRCS2Regs D-registers
/// starting from d8.  These instructions are assumed to execute while the
/// stack is still aligned, unlike the code inserted by emitPopInst.
static void emitAlignedDPRCS2Restores(MachineBasicBlock &MBB,
                                      MachineBasicBlock::iterator MI,
                                      unsigned NumAlignedDPRCS2Regs,
                                      const std::vector<CalleeSavedInfo> &CSI,
                                      const TargetRegisterInfo *TRI) {
  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL = MI->getDebugLoc();
  const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();

  // Find the frame index assigned to d8.
  int D8SpillFI = 0;
  for (unsigned i = 0, e = CSI.size(); i != e; ++i)
    if (CSI[i].getReg() == ARM::D8) {
      D8SpillFI = CSI[i].getFrameIdx();
      break;
    }

  // Materialize the address of the d8 spill slot into the scratch register r4.
  // This can be fairly complicated if the stack frame is large, so just use
  // the normal frame index elimination mechanism to do it.  This code runs as
  // the initial part of the epilog where the stack and base pointers haven't
  // been changed yet.
  bool isThumb = AFI->isThumbFunction();
  assert(!AFI->isThumb1OnlyFunction() && "Can't realign stack for thumb1");

  unsigned Opc = isThumb ? ARM::t2ADDri : ARM::ADDri;
  AddDefaultCC(AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
                              .addFrameIndex(D8SpillFI).addImm(0)));

  // Now restore NumAlignedDPRCS2Regs registers starting from d8.
  unsigned NextReg = ARM::D8;

  // 16-byte aligned vld1.64 with 4 d-regs and writeback.
  if (NumAlignedDPRCS2Regs >= 6) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Qwb_fixed), NextReg)
                   .addReg(ARM::R4, RegState::Define)
                   .addReg(ARM::R4, RegState::Kill).addImm(16)
                   .addReg(SupReg, RegState::ImplicitDefine));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // We won't modify r4 beyond this point.  It currently points to the next
  // register to be spilled.
  unsigned R4BaseReg = NextReg;

  // 16-byte aligned vld1.64 with 4 d-regs, no writeback.
  if (NumAlignedDPRCS2Regs >= 4) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Q), NextReg)
                   .addReg(ARM::R4).addImm(16)
                   .addReg(SupReg, RegState::ImplicitDefine));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // 16-byte aligned vld1.64 with 2 d-regs.
  if (NumAlignedDPRCS2Regs >= 2) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QPRRegClass);
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1q64), SupReg)
                   .addReg(ARM::R4).addImm(16));
    NextReg += 2;
    NumAlignedDPRCS2Regs -= 2;
  }

  // Finally, use a vanilla vldr.64 for the remaining odd register.
  if (NumAlignedDPRCS2Regs)
    AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLDRD), NextReg)
                   .addReg(ARM::R4).addImm(2*(NextReg-R4BaseReg)));

  // Last store kills r4.
  std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}

bool ARMFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
                                        MachineBasicBlock::iterator MI,
                                        const std::vector<CalleeSavedInfo> &CSI,
                                        const TargetRegisterInfo *TRI) const {
  if (CSI.empty())
    return false;

  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();

  unsigned PushOpc = AFI->isThumbFunction() ? ARM::t2STMDB_UPD : ARM::STMDB_UPD;
  unsigned PushOneOpc = AFI->isThumbFunction() ?
    ARM::t2STR_PRE : ARM::STR_PRE_IMM;
  unsigned FltOpc = ARM::VSTMDDB_UPD;
  unsigned NumAlignedDPRCS2Regs = AFI->getNumAlignedDPRCS2Regs();
  emitPushInst(MBB, MI, CSI, PushOpc, PushOneOpc, false, &isARMArea1Register, 0,
               MachineInstr::FrameSetup);
  emitPushInst(MBB, MI, CSI, PushOpc, PushOneOpc, false, &isARMArea2Register, 0,
               MachineInstr::FrameSetup);
  emitPushInst(MBB, MI, CSI, FltOpc, 0, true, &isARMArea3Register,
               NumAlignedDPRCS2Regs, MachineInstr::FrameSetup);

  // The code above does not insert spill code for the aligned DPRCS2 registers.
  // The stack realignment code will be inserted between the push instructions
  // and these spills.
  if (NumAlignedDPRCS2Regs)
    emitAlignedDPRCS2Spills(MBB, MI, NumAlignedDPRCS2Regs, CSI, TRI);

  return true;
}

bool ARMFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
                                        MachineBasicBlock::iterator MI,
                                        const std::vector<CalleeSavedInfo> &CSI,
                                        const TargetRegisterInfo *TRI) const {
  if (CSI.empty())
    return false;

  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  bool isVarArg = AFI->getArgRegsSaveSize() > 0;
  unsigned NumAlignedDPRCS2Regs = AFI->getNumAlignedDPRCS2Regs();

  // The emitPopInst calls below do not insert reloads for the aligned DPRCS2
  // registers. Do that here instead.
  if (NumAlignedDPRCS2Regs)
    emitAlignedDPRCS2Restores(MBB, MI, NumAlignedDPRCS2Regs, CSI, TRI);

  unsigned PopOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_UPD : ARM::LDMIA_UPD;
  unsigned LdrOpc = AFI->isThumbFunction() ? ARM::t2LDR_POST :ARM::LDR_POST_IMM;
  unsigned FltOpc = ARM::VLDMDIA_UPD;
  emitPopInst(MBB, MI, CSI, FltOpc, 0, isVarArg, true, &isARMArea3Register,
              NumAlignedDPRCS2Regs);
  emitPopInst(MBB, MI, CSI, PopOpc, LdrOpc, isVarArg, false,
              &isARMArea2Register, 0);
  emitPopInst(MBB, MI, CSI, PopOpc, LdrOpc, isVarArg, false,
              &isARMArea1Register, 0);

  return true;
}

// FIXME: Make generic?
static unsigned GetFunctionSizeInBytes(const MachineFunction &MF,
                                       const ARMBaseInstrInfo &TII) {
  unsigned FnSize = 0;
  for (MachineFunction::const_iterator MBBI = MF.begin(), E = MF.end();
       MBBI != E; ++MBBI) {
    const MachineBasicBlock &MBB = *MBBI;
    for (MachineBasicBlock::const_iterator I = MBB.begin(),E = MBB.end();
         I != E; ++I)
      FnSize += TII.GetInstSizeInBytes(I);
  }
  return FnSize;
}

/// estimateRSStackSizeLimit - Look at each instruction that references stack
/// frames and return the stack size limit beyond which some of these
/// instructions will require a scratch register during their expansion later.
// FIXME: Move to TII?
static unsigned estimateRSStackSizeLimit(MachineFunction &MF,
                                         const TargetFrameLowering *TFI) {
  const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  unsigned Limit = (1 << 12) - 1;
  for (MachineFunction::iterator BB = MF.begin(),E = MF.end(); BB != E; ++BB) {
    for (MachineBasicBlock::iterator I = BB->begin(), E = BB->end();
         I != E; ++I) {
      for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
        if (!I->getOperand(i).isFI()) continue;

        // When using ADDri to get the address of a stack object, 255 is the
        // largest offset guaranteed to fit in the immediate offset.
        if (I->getOpcode() == ARM::ADDri) {
          Limit = std::min(Limit, (1U << 8) - 1);
          break;
        }

        // Otherwise check the addressing mode.
        switch (I->getDesc().TSFlags & ARMII::AddrModeMask) {
        case ARMII::AddrMode3:
        case ARMII::AddrModeT2_i8:
          Limit = std::min(Limit, (1U << 8) - 1);
          break;
        case ARMII::AddrMode5:
        case ARMII::AddrModeT2_i8s4:
          Limit = std::min(Limit, ((1U << 8) - 1) * 4);
          break;
        case ARMII::AddrModeT2_i12:
          // i12 supports only positive offset so these will be converted to
          // i8 opcodes. See llvm::rewriteT2FrameIndex.
          if (TFI->hasFP(MF) && AFI->hasStackFrame())
            Limit = std::min(Limit, (1U << 8) - 1);
          break;
        case ARMII::AddrMode4:
        case ARMII::AddrMode6:
          // Addressing modes 4 & 6 (load/store) instructions can't encode an
          // immediate offset for stack references.
          return 0;
        default:
          break;
        }
        break; // At most one FI per instruction
      }
    }
  }

  return Limit;
}

// In functions that realign the stack, it can be an advantage to spill the
// callee-saved vector registers after realigning the stack. The vst1 and vld1
// instructions take alignment hints that can improve performance.
//
static void checkNumAlignedDPRCS2Regs(MachineFunction &MF) {
  MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(0);
  if (!SpillAlignedNEONRegs)
    return;

  // Naked functions don't spill callee-saved registers.
  if (MF.getFunction()->getAttributes().hasAttribute(AttributeSet::FunctionIndex,
                                                     Attribute::Naked))
    return;

  // We are planning to use NEON instructions vst1 / vld1.
  if (!MF.getTarget().getSubtarget<ARMSubtarget>().hasNEON())
    return;

  // Don't bother if the default stack alignment is sufficiently high.
  if (MF.getTarget().getFrameLowering()->getStackAlignment() >= 8)
    return;

  // Aligned spills require stack realignment.
  const ARMBaseRegisterInfo *RegInfo =
    static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
  if (!RegInfo->canRealignStack(MF))
    return;

  // We always spill contiguous d-registers starting from d8. Count how many
  // needs spilling.  The register allocator will almost always use the
  // callee-saved registers in order, but it can happen that there are holes in
  // the range.  Registers above the hole will be spilled to the standard DPRCS
  // area.
  MachineRegisterInfo &MRI = MF.getRegInfo();
  unsigned NumSpills = 0;
  for (; NumSpills < 8; ++NumSpills)
    if (!MRI.isPhysRegUsed(ARM::D8 + NumSpills))
      break;

  // Don't do this for just one d-register. It's not worth it.
  if (NumSpills < 2)
    return;

  // Spill the first NumSpills D-registers after realigning the stack.
  MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(NumSpills);

  // A scratch register is required for the vst1 / vld1 instructions.
  MF.getRegInfo().setPhysRegUsed(ARM::R4);
}

void
ARMFrameLowering::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;
  const ARMBaseRegisterInfo *RegInfo =
    static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
  const ARMBaseInstrInfo &TII =
    *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  MachineFrameInfo *MFI = MF.getFrameInfo();
  MachineRegisterInfo &MRI = MF.getRegInfo();
  unsigned FramePtr = RegInfo->getFrameRegister(MF);

  // Spill R4 if Thumb2 function requires stack realignment - it will be used as
  // scratch register. Also spill R4 if Thumb2 function has varsized objects,
  // since it's not always possible to restore sp from fp in a single
  // instruction.
  // FIXME: It will be better just to find spare register here.
  if (AFI->isThumb2Function() &&
      (MFI->hasVarSizedObjects() || RegInfo->needsStackRealignment(MF)))
    MRI.setPhysRegUsed(ARM::R4);

  if (AFI->isThumb1OnlyFunction()) {
    // Spill LR if Thumb1 function uses variable length argument lists.
    if (AFI->getArgRegsSaveSize() > 0)
      MRI.setPhysRegUsed(ARM::LR);

    // Spill R4 if Thumb1 epilogue has to restore SP from FP. We don't know
    // for sure what the stack size will be, but for this, an estimate is good
    // enough. If there anything changes it, it'll be a spill, which implies
    // we've used all the registers and so R4 is already used, so not marking
    // it here will be OK.
    // FIXME: It will be better just to find spare register here.
    unsigned StackSize = MFI->estimateStackSize(MF);
    if (MFI->hasVarSizedObjects() || StackSize > 508)
      MRI.setPhysRegUsed(ARM::R4);
  }

  // See if we can spill vector registers to aligned stack.
  checkNumAlignedDPRCS2Regs(MF);

  // Spill the BasePtr if it's used.
  if (RegInfo->hasBasePointer(MF))
    MRI.setPhysRegUsed(RegInfo->getBaseRegister());

  // 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 uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
  for (unsigned i = 0; CSRegs[i]; ++i) {
    unsigned Reg = CSRegs[i];
    bool Spilled = false;
    if (MRI.isPhysRegUsed(Reg)) {
      Spilled = true;
      CanEliminateFrame = false;
    }

    if (!ARM::GPRRegClass.contains(Reg))
      continue;

    if (Spilled) {
      NumGPRSpills++;

      if (!STI.isTargetMachO()) {
        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::R0: case ARM::R1:
      case ARM::R2: case ARM::R3:
      case ARM::R4: case ARM::R5:
      case ARM::R6: case ARM::R7:
        CS1Spilled = true;
        break;
      default:
        break;
      }
    } else {
      if (!STI.isTargetMachO()) {
        UnspilledCS1GPRs.push_back(Reg);
        continue;
      }

      switch (Reg) {
      case ARM::R0: case ARM::R1:
      case ARM::R2: case ARM::R3:
      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->isThumb1OnlyFunction()) {
    unsigned FnSize = GetFunctionSizeInBytes(MF, TII);
    // 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;
    }
  }

  // If any of the stack slot references may be out of range of an immediate
  // offset, make sure a register (or a spill slot) is available for the
  // register scavenger. Note that if we're indexing off the frame pointer, the
  // effective stack size is 4 bytes larger since the FP points to the stack
  // slot of the previous FP. Also, if we have variable sized objects in the
  // function, stack slot references will often be negative, and some of
  // our instructions are positive-offset only, so conservatively consider
  // that case to want a spill slot (or register) as well. Similarly, if
  // the function adjusts the stack pointer during execution and the
  // adjustments aren't already part of our stack size estimate, our offset
  // calculations may be off, so be conservative.
  // FIXME: We could add logic to be more precise about negative offsets
  //        and which instructions will need a scratch register for them. Is it
  //        worth the effort and added fragility?
  bool BigStack =
    (RS &&
     (MFI->estimateStackSize(MF) +
      ((hasFP(MF) && AFI->hasStackFrame()) ? 4:0) >=
      estimateRSStackSizeLimit(MF, this)))
    || MFI->hasVarSizedObjects()
    || (MFI->adjustsStack() && !canSimplifyCallFramePseudos(MF));

  bool ExtraCSSpill = false;
  if (BigStack || !CanEliminateFrame || RegInfo->cannotEliminateFrame(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) {
      MRI.setPhysRegUsed(ARM::LR);
      NumGPRSpills++;
      SmallVectorImpl<unsigned>::iterator LRPos;
      LRPos = std::find(UnspilledCS1GPRs.begin(), UnspilledCS1GPRs.end(),
                        (unsigned)ARM::LR);
      if (LRPos != UnspilledCS1GPRs.end())
        UnspilledCS1GPRs.erase(LRPos);

      ForceLRSpill = false;
      ExtraCSSpill = true;
    }

    if (hasFP(MF)) {
      MRI.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 = 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 spill high register if the function is thumb1
          if (!AFI->isThumb1OnlyFunction() ||
              isARMLowRegister(Reg) || Reg == ARM::LR) {
            MRI.setPhysRegUsed(Reg);
            if (!MRI.isReserved(Reg))
              ExtraCSSpill = true;
            break;
          }
        }
      } else if (!UnspilledCS2GPRs.empty() && !AFI->isThumb1OnlyFunction()) {
        unsigned Reg = UnspilledCS2GPRs.front();
        MRI.setPhysRegUsed(Reg);
        if (!MRI.isReserved(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 additional
    // callee-saved register or reserve a special spill slot to facilitate
    // register scavenging. Thumb1 needs a spill slot for stack pointer
    // adjustments also, even when the frame itself is small.
    if (BigStack && !ExtraCSSpill) {
      // 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 (!MRI.isReserved(Reg) &&
            (!AFI->isThumb1OnlyFunction() || isARMLowRegister(Reg) ||
             Reg == ARM::LR)) {
          Extras.push_back(Reg);
          NumExtras--;
        }
      }
      // For non-Thumb1 functions, also check for hi-reg CS registers
      if (!AFI->isThumb1OnlyFunction()) {
        while (NumExtras && !UnspilledCS2GPRs.empty()) {
          unsigned Reg = UnspilledCS2GPRs.back();
          UnspilledCS2GPRs.pop_back();
          if (!MRI.isReserved(Reg)) {
            Extras.push_back(Reg);
            NumExtras--;
          }
        }
      }
      if (Extras.size() && NumExtras == 0) {
        for (unsigned i = 0, e = Extras.size(); i != e; ++i) {
          MRI.setPhysRegUsed(Extras[i]);
        }
      } else if (!AFI->isThumb1OnlyFunction()) {
        // note: Thumb1 functions spill to R12, not the stack.  Reserve a slot
        // closest to SP or frame pointer.
        const TargetRegisterClass *RC = &ARM::GPRRegClass;
        RS->addScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
                                                           RC->getAlignment(),
                                                           false));
      }
    }
  }

  if (ForceLRSpill) {
    MRI.setPhysRegUsed(ARM::LR);
    AFI->setLRIsSpilledForFarJump(true);
  }
}


void ARMFrameLowering::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
                              MachineBasicBlock::iterator I) const {
  const ARMBaseInstrInfo &TII =
    *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
  if (!hasReservedCallFrame(MF)) {
    // If we have alloca, convert as follows:
    // ADJCALLSTACKDOWN -> sub, sp, sp, amount
    // ADJCALLSTACKUP   -> add, sp, sp, amount
    MachineInstr *Old = I;
    DebugLoc dl = Old->getDebugLoc();
    unsigned Amount = Old->getOperand(0).getImm();
    if (Amount != 0) {
      // 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 = getStackAlignment();
      Amount = (Amount+Align-1)/Align*Align;

      ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
      assert(!AFI->isThumb1OnlyFunction() &&
             "This eliminateCallFramePseudoInstr does not support Thumb1!");
      bool isARM = !AFI->isThumbFunction();

      // Replace the pseudo instruction with a new instruction...
      unsigned Opc = Old->getOpcode();
      int PIdx = Old->findFirstPredOperandIdx();
      ARMCC::CondCodes Pred = (PIdx == -1)
        ? ARMCC::AL : (ARMCC::CondCodes)Old->getOperand(PIdx).getImm();
      if (Opc == ARM::ADJCALLSTACKDOWN || Opc == ARM::tADJCALLSTACKDOWN) {
        // Note: PredReg is operand 2 for ADJCALLSTACKDOWN.
        unsigned PredReg = Old->getOperand(2).getReg();
        emitSPUpdate(isARM, MBB, I, dl, TII, -Amount, MachineInstr::NoFlags,
                     Pred, PredReg);
      } else {
        // Note: PredReg is operand 3 for ADJCALLSTACKUP.
        unsigned PredReg = Old->getOperand(3).getReg();
        assert(Opc == ARM::ADJCALLSTACKUP || Opc == ARM::tADJCALLSTACKUP);
        emitSPUpdate(isARM, MBB, I, dl, TII, Amount, MachineInstr::NoFlags,
                     Pred, PredReg);
      }
    }
  }
  MBB.erase(I);
}