llvm.org GIT mirror llvm / release_30 lib / Target / PowerPC / PPCInstrInfo.cpp
release_30

Tree @release_30 (Download .tar.gz)

PPCInstrInfo.cpp @release_30raw · 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
//===- PPCInstrInfo.cpp - PowerPC32 Instruction Information -----*- C++ -*-===//
//
//                     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 PowerPC implementation of the TargetInstrInfo class.
//
//===----------------------------------------------------------------------===//

#include "PPCInstrInfo.h"
#include "PPC.h"
#include "PPCInstrBuilder.h"
#include "PPCMachineFunctionInfo.h"
#include "PPCTargetMachine.h"
#include "PPCHazardRecognizers.h"
#include "MCTargetDesc/PPCPredicates.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/PseudoSourceValue.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/ADT/STLExtras.h"

#define GET_INSTRINFO_CTOR
#include "PPCGenInstrInfo.inc"

namespace llvm {
extern cl::opt<bool> EnablePPC32RS;  // FIXME (64-bit): See PPCRegisterInfo.cpp.
extern cl::opt<bool> EnablePPC64RS;  // FIXME (64-bit): See PPCRegisterInfo.cpp.
}

using namespace llvm;

PPCInstrInfo::PPCInstrInfo(PPCTargetMachine &tm)
  : PPCGenInstrInfo(PPC::ADJCALLSTACKDOWN, PPC::ADJCALLSTACKUP),
    TM(tm), RI(*TM.getSubtargetImpl(), *this) {}

/// CreateTargetHazardRecognizer - Return the hazard recognizer to use for
/// this target when scheduling the DAG.
ScheduleHazardRecognizer *PPCInstrInfo::CreateTargetHazardRecognizer(
  const TargetMachine *TM,
  const ScheduleDAG *DAG) const {
  // Should use subtarget info to pick the right hazard recognizer.  For
  // now, always return a PPC970 recognizer.
  const TargetInstrInfo *TII = TM->getInstrInfo();
  assert(TII && "No InstrInfo?");
  return new PPCHazardRecognizer970(*TII);
}

unsigned PPCInstrInfo::isLoadFromStackSlot(const MachineInstr *MI,
                                           int &FrameIndex) const {
  switch (MI->getOpcode()) {
  default: break;
  case PPC::LD:
  case PPC::LWZ:
  case PPC::LFS:
  case PPC::LFD:
    if (MI->getOperand(1).isImm() && !MI->getOperand(1).getImm() &&
        MI->getOperand(2).isFI()) {
      FrameIndex = MI->getOperand(2).getIndex();
      return MI->getOperand(0).getReg();
    }
    break;
  }
  return 0;
}

unsigned PPCInstrInfo::isStoreToStackSlot(const MachineInstr *MI,
                                          int &FrameIndex) const {
  switch (MI->getOpcode()) {
  default: break;
  case PPC::STD:
  case PPC::STW:
  case PPC::STFS:
  case PPC::STFD:
    if (MI->getOperand(1).isImm() && !MI->getOperand(1).getImm() &&
        MI->getOperand(2).isFI()) {
      FrameIndex = MI->getOperand(2).getIndex();
      return MI->getOperand(0).getReg();
    }
    break;
  }
  return 0;
}

// commuteInstruction - We can commute rlwimi instructions, but only if the
// rotate amt is zero.  We also have to munge the immediates a bit.
MachineInstr *
PPCInstrInfo::commuteInstruction(MachineInstr *MI, bool NewMI) const {
  MachineFunction &MF = *MI->getParent()->getParent();

  // Normal instructions can be commuted the obvious way.
  if (MI->getOpcode() != PPC::RLWIMI)
    return TargetInstrInfoImpl::commuteInstruction(MI, NewMI);

  // Cannot commute if it has a non-zero rotate count.
  if (MI->getOperand(3).getImm() != 0)
    return 0;

  // If we have a zero rotate count, we have:
  //   M = mask(MB,ME)
  //   Op0 = (Op1 & ~M) | (Op2 & M)
  // Change this to:
  //   M = mask((ME+1)&31, (MB-1)&31)
  //   Op0 = (Op2 & ~M) | (Op1 & M)

  // Swap op1/op2
  unsigned Reg0 = MI->getOperand(0).getReg();
  unsigned Reg1 = MI->getOperand(1).getReg();
  unsigned Reg2 = MI->getOperand(2).getReg();
  bool Reg1IsKill = MI->getOperand(1).isKill();
  bool Reg2IsKill = MI->getOperand(2).isKill();
  bool ChangeReg0 = false;
  // If machine instrs are no longer in two-address forms, update
  // destination register as well.
  if (Reg0 == Reg1) {
    // Must be two address instruction!
    assert(MI->getDesc().getOperandConstraint(0, MCOI::TIED_TO) &&
           "Expecting a two-address instruction!");
    Reg2IsKill = false;
    ChangeReg0 = true;
  }

  // Masks.
  unsigned MB = MI->getOperand(4).getImm();
  unsigned ME = MI->getOperand(5).getImm();

  if (NewMI) {
    // Create a new instruction.
    unsigned Reg0 = ChangeReg0 ? Reg2 : MI->getOperand(0).getReg();
    bool Reg0IsDead = MI->getOperand(0).isDead();
    return BuildMI(MF, MI->getDebugLoc(), MI->getDesc())
      .addReg(Reg0, RegState::Define | getDeadRegState(Reg0IsDead))
      .addReg(Reg2, getKillRegState(Reg2IsKill))
      .addReg(Reg1, getKillRegState(Reg1IsKill))
      .addImm((ME+1) & 31)
      .addImm((MB-1) & 31);
  }

  if (ChangeReg0)
    MI->getOperand(0).setReg(Reg2);
  MI->getOperand(2).setReg(Reg1);
  MI->getOperand(1).setReg(Reg2);
  MI->getOperand(2).setIsKill(Reg1IsKill);
  MI->getOperand(1).setIsKill(Reg2IsKill);

  // Swap the mask around.
  MI->getOperand(4).setImm((ME+1) & 31);
  MI->getOperand(5).setImm((MB-1) & 31);
  return MI;
}

void PPCInstrInfo::insertNoop(MachineBasicBlock &MBB,
                              MachineBasicBlock::iterator MI) const {
  DebugLoc DL;
  BuildMI(MBB, MI, DL, get(PPC::NOP));
}


// Branch analysis.
bool PPCInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
                                 MachineBasicBlock *&FBB,
                                 SmallVectorImpl<MachineOperand> &Cond,
                                 bool AllowModify) const {
  // If the block has no terminators, it just falls into the block after it.
  MachineBasicBlock::iterator I = MBB.end();
  if (I == MBB.begin())
    return false;
  --I;
  while (I->isDebugValue()) {
    if (I == MBB.begin())
      return false;
    --I;
  }
  if (!isUnpredicatedTerminator(I))
    return false;

  // Get the last instruction in the block.
  MachineInstr *LastInst = I;

  // If there is only one terminator instruction, process it.
  if (I == MBB.begin() || !isUnpredicatedTerminator(--I)) {
    if (LastInst->getOpcode() == PPC::B) {
      if (!LastInst->getOperand(0).isMBB())
        return true;
      TBB = LastInst->getOperand(0).getMBB();
      return false;
    } else if (LastInst->getOpcode() == PPC::BCC) {
      if (!LastInst->getOperand(2).isMBB())
        return true;
      // Block ends with fall-through condbranch.
      TBB = LastInst->getOperand(2).getMBB();
      Cond.push_back(LastInst->getOperand(0));
      Cond.push_back(LastInst->getOperand(1));
      return false;
    }
    // Otherwise, don't know what this is.
    return true;
  }

  // Get the instruction before it if it's a terminator.
  MachineInstr *SecondLastInst = I;

  // If there are three terminators, we don't know what sort of block this is.
  if (SecondLastInst && I != MBB.begin() &&
      isUnpredicatedTerminator(--I))
    return true;

  // If the block ends with PPC::B and PPC:BCC, handle it.
  if (SecondLastInst->getOpcode() == PPC::BCC &&
      LastInst->getOpcode() == PPC::B) {
    if (!SecondLastInst->getOperand(2).isMBB() ||
        !LastInst->getOperand(0).isMBB())
      return true;
    TBB =  SecondLastInst->getOperand(2).getMBB();
    Cond.push_back(SecondLastInst->getOperand(0));
    Cond.push_back(SecondLastInst->getOperand(1));
    FBB = LastInst->getOperand(0).getMBB();
    return false;
  }

  // If the block ends with two PPC:Bs, handle it.  The second one is not
  // executed, so remove it.
  if (SecondLastInst->getOpcode() == PPC::B &&
      LastInst->getOpcode() == PPC::B) {
    if (!SecondLastInst->getOperand(0).isMBB())
      return true;
    TBB = SecondLastInst->getOperand(0).getMBB();
    I = LastInst;
    if (AllowModify)
      I->eraseFromParent();
    return false;
  }

  // Otherwise, can't handle this.
  return true;
}

unsigned PPCInstrInfo::RemoveBranch(MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator I = MBB.end();
  if (I == MBB.begin()) return 0;
  --I;
  while (I->isDebugValue()) {
    if (I == MBB.begin())
      return 0;
    --I;
  }
  if (I->getOpcode() != PPC::B && I->getOpcode() != PPC::BCC)
    return 0;

  // Remove the branch.
  I->eraseFromParent();

  I = MBB.end();

  if (I == MBB.begin()) return 1;
  --I;
  if (I->getOpcode() != PPC::BCC)
    return 1;

  // Remove the branch.
  I->eraseFromParent();
  return 2;
}

unsigned
PPCInstrInfo::InsertBranch(MachineBasicBlock &MBB, MachineBasicBlock *TBB,
                           MachineBasicBlock *FBB,
                           const SmallVectorImpl<MachineOperand> &Cond,
                           DebugLoc DL) const {
  // Shouldn't be a fall through.
  assert(TBB && "InsertBranch must not be told to insert a fallthrough");
  assert((Cond.size() == 2 || Cond.size() == 0) &&
         "PPC branch conditions have two components!");

  // One-way branch.
  if (FBB == 0) {
    if (Cond.empty())   // Unconditional branch
      BuildMI(&MBB, DL, get(PPC::B)).addMBB(TBB);
    else                // Conditional branch
      BuildMI(&MBB, DL, get(PPC::BCC))
        .addImm(Cond[0].getImm()).addReg(Cond[1].getReg()).addMBB(TBB);
    return 1;
  }

  // Two-way Conditional Branch.
  BuildMI(&MBB, DL, get(PPC::BCC))
    .addImm(Cond[0].getImm()).addReg(Cond[1].getReg()).addMBB(TBB);
  BuildMI(&MBB, DL, get(PPC::B)).addMBB(FBB);
  return 2;
}

void PPCInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
                               MachineBasicBlock::iterator I, DebugLoc DL,
                               unsigned DestReg, unsigned SrcReg,
                               bool KillSrc) const {
  unsigned Opc;
  if (PPC::GPRCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::OR;
  else if (PPC::G8RCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::OR8;
  else if (PPC::F4RCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::FMR;
  else if (PPC::CRRCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::MCRF;
  else if (PPC::VRRCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::VOR;
  else if (PPC::CRBITRCRegClass.contains(DestReg, SrcReg))
    Opc = PPC::CROR;
  else
    llvm_unreachable("Impossible reg-to-reg copy");

  const MCInstrDesc &MCID = get(Opc);
  if (MCID.getNumOperands() == 3)
    BuildMI(MBB, I, DL, MCID, DestReg)
      .addReg(SrcReg).addReg(SrcReg, getKillRegState(KillSrc));
  else
    BuildMI(MBB, I, DL, MCID, DestReg).addReg(SrcReg, getKillRegState(KillSrc));
}

bool
PPCInstrInfo::StoreRegToStackSlot(MachineFunction &MF,
                                  unsigned SrcReg, bool isKill,
                                  int FrameIdx,
                                  const TargetRegisterClass *RC,
                                  SmallVectorImpl<MachineInstr*> &NewMIs) const{
  DebugLoc DL;
  if (PPC::GPRCRegisterClass->hasSubClassEq(RC)) {
    if (SrcReg != PPC::LR) {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW))
                                         .addReg(SrcReg,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
    } else {
      // FIXME: this spills LR immediately to memory in one step.  To do this,
      // we use R11, which we know cannot be used in the prolog/epilog.  This is
      // a hack.
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::MFLR), PPC::R11));
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW))
                                         .addReg(PPC::R11,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
    }
  } else if (PPC::G8RCRegisterClass->hasSubClassEq(RC)) {
    if (SrcReg != PPC::LR8) {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STD))
                                         .addReg(SrcReg,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
    } else {
      // FIXME: this spills LR immediately to memory in one step.  To do this,
      // we use R11, which we know cannot be used in the prolog/epilog.  This is
      // a hack.
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::MFLR8), PPC::X11));
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STD))
                                         .addReg(PPC::X11,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
    }
  } else if (PPC::F8RCRegisterClass->hasSubClassEq(RC)) {
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STFD))
                                       .addReg(SrcReg,
                                               getKillRegState(isKill)),
                                       FrameIdx));
  } else if (PPC::F4RCRegisterClass->hasSubClassEq(RC)) {
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STFS))
                                       .addReg(SrcReg,
                                               getKillRegState(isKill)),
                                       FrameIdx));
  } else if (PPC::CRRCRegisterClass->hasSubClassEq(RC)) {
    if ((EnablePPC32RS && !TM.getSubtargetImpl()->isPPC64()) ||
        (EnablePPC64RS && TM.getSubtargetImpl()->isPPC64())) {
      // FIXME (64-bit): Enable
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPILL_CR))
                                         .addReg(SrcReg,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
      return true;
    } else {
      // FIXME: We need a scatch reg here.  The trouble with using R0 is that
      // it's possible for the stack frame to be so big the save location is
      // out of range of immediate offsets, necessitating another register.
      // We hack this on Darwin by reserving R2.  It's probably broken on Linux
      // at the moment.

      // We need to store the CR in the low 4-bits of the saved value.  First,
      // issue a MFCR to save all of the CRBits.
      unsigned ScratchReg = TM.getSubtargetImpl()->isDarwinABI() ?
                                                           PPC::R2 : PPC::R0;
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::MFCRpseud), ScratchReg)
                               .addReg(SrcReg, getKillRegState(isKill)));

      // If the saved register wasn't CR0, shift the bits left so that they are
      // in CR0's slot.
      if (SrcReg != PPC::CR0) {
        unsigned ShiftBits = getPPCRegisterNumbering(SrcReg)*4;
        // rlwinm scratch, scratch, ShiftBits, 0, 31.
        NewMIs.push_back(BuildMI(MF, DL, get(PPC::RLWINM), ScratchReg)
                       .addReg(ScratchReg).addImm(ShiftBits)
                       .addImm(0).addImm(31));
      }

      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW))
                                         .addReg(ScratchReg,
                                                 getKillRegState(isKill)),
                                         FrameIdx));
    }
  } else if (PPC::CRBITRCRegisterClass->hasSubClassEq(RC)) {
    // FIXME: We use CRi here because there is no mtcrf on a bit. Since the
    // backend currently only uses CR1EQ as an individual bit, this should
    // not cause any bug. If we need other uses of CR bits, the following
    // code may be invalid.
    unsigned Reg = 0;
    if (SrcReg == PPC::CR0LT || SrcReg == PPC::CR0GT ||
        SrcReg == PPC::CR0EQ || SrcReg == PPC::CR0UN)
      Reg = PPC::CR0;
    else if (SrcReg == PPC::CR1LT || SrcReg == PPC::CR1GT ||
             SrcReg == PPC::CR1EQ || SrcReg == PPC::CR1UN)
      Reg = PPC::CR1;
    else if (SrcReg == PPC::CR2LT || SrcReg == PPC::CR2GT ||
             SrcReg == PPC::CR2EQ || SrcReg == PPC::CR2UN)
      Reg = PPC::CR2;
    else if (SrcReg == PPC::CR3LT || SrcReg == PPC::CR3GT ||
             SrcReg == PPC::CR3EQ || SrcReg == PPC::CR3UN)
      Reg = PPC::CR3;
    else if (SrcReg == PPC::CR4LT || SrcReg == PPC::CR4GT ||
             SrcReg == PPC::CR4EQ || SrcReg == PPC::CR4UN)
      Reg = PPC::CR4;
    else if (SrcReg == PPC::CR5LT || SrcReg == PPC::CR5GT ||
             SrcReg == PPC::CR5EQ || SrcReg == PPC::CR5UN)
      Reg = PPC::CR5;
    else if (SrcReg == PPC::CR6LT || SrcReg == PPC::CR6GT ||
             SrcReg == PPC::CR6EQ || SrcReg == PPC::CR6UN)
      Reg = PPC::CR6;
    else if (SrcReg == PPC::CR7LT || SrcReg == PPC::CR7GT ||
             SrcReg == PPC::CR7EQ || SrcReg == PPC::CR7UN)
      Reg = PPC::CR7;

    return StoreRegToStackSlot(MF, Reg, isKill, FrameIdx,
                               PPC::CRRCRegisterClass, NewMIs);

  } else if (PPC::VRRCRegisterClass->hasSubClassEq(RC)) {
    // We don't have indexed addressing for vector loads.  Emit:
    // R0 = ADDI FI#
    // STVX VAL, 0, R0
    //
    // FIXME: We use R0 here, because it isn't available for RA.
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::ADDI), PPC::R0),
                                       FrameIdx, 0, 0));
    NewMIs.push_back(BuildMI(MF, DL, get(PPC::STVX))
                     .addReg(SrcReg, getKillRegState(isKill))
                     .addReg(PPC::R0)
                     .addReg(PPC::R0));
  } else {
    llvm_unreachable("Unknown regclass!");
  }

  return false;
}

void
PPCInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
                                  MachineBasicBlock::iterator MI,
                                  unsigned SrcReg, bool isKill, int FrameIdx,
                                  const TargetRegisterClass *RC,
                                  const TargetRegisterInfo *TRI) const {
  MachineFunction &MF = *MBB.getParent();
  SmallVector<MachineInstr*, 4> NewMIs;

  if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs)) {
    PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
    FuncInfo->setSpillsCR();
  }

  for (unsigned i = 0, e = NewMIs.size(); i != e; ++i)
    MBB.insert(MI, NewMIs[i]);

  const MachineFrameInfo &MFI = *MF.getFrameInfo();
  MachineMemOperand *MMO =
    MF.getMachineMemOperand(
                MachinePointerInfo(PseudoSourceValue::getFixedStack(FrameIdx)),
                            MachineMemOperand::MOStore,
                            MFI.getObjectSize(FrameIdx),
                            MFI.getObjectAlignment(FrameIdx));
  NewMIs.back()->addMemOperand(MF, MMO);
}

void
PPCInstrInfo::LoadRegFromStackSlot(MachineFunction &MF, DebugLoc DL,
                                   unsigned DestReg, int FrameIdx,
                                   const TargetRegisterClass *RC,
                                   SmallVectorImpl<MachineInstr*> &NewMIs)const{
  if (PPC::GPRCRegisterClass->hasSubClassEq(RC)) {
    if (DestReg != PPC::LR) {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ),
                                                 DestReg), FrameIdx));
    } else {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ),
                                                 PPC::R11), FrameIdx));
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::MTLR)).addReg(PPC::R11));
    }
  } else if (PPC::G8RCRegisterClass->hasSubClassEq(RC)) {
    if (DestReg != PPC::LR8) {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD), DestReg),
                                         FrameIdx));
    } else {
      NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD),
                                                 PPC::R11), FrameIdx));
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::MTLR8)).addReg(PPC::R11));
    }
  } else if (PPC::F8RCRegisterClass->hasSubClassEq(RC)) {
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LFD), DestReg),
                                       FrameIdx));
  } else if (PPC::F4RCRegisterClass->hasSubClassEq(RC)) {
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LFS), DestReg),
                                       FrameIdx));
  } else if (PPC::CRRCRegisterClass->hasSubClassEq(RC)) {
    // FIXME: We need a scatch reg here.  The trouble with using R0 is that
    // it's possible for the stack frame to be so big the save location is
    // out of range of immediate offsets, necessitating another register.
    // We hack this on Darwin by reserving R2.  It's probably broken on Linux
    // at the moment.
    unsigned ScratchReg = TM.getSubtargetImpl()->isDarwinABI() ?
                                                          PPC::R2 : PPC::R0;
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ),
                                       ScratchReg), FrameIdx));

    // If the reloaded register isn't CR0, shift the bits right so that they are
    // in the right CR's slot.
    if (DestReg != PPC::CR0) {
      unsigned ShiftBits = getPPCRegisterNumbering(DestReg)*4;
      // rlwinm r11, r11, 32-ShiftBits, 0, 31.
      NewMIs.push_back(BuildMI(MF, DL, get(PPC::RLWINM), ScratchReg)
                    .addReg(ScratchReg).addImm(32-ShiftBits).addImm(0)
                    .addImm(31));
    }

    NewMIs.push_back(BuildMI(MF, DL, get(PPC::MTCRF), DestReg)
                     .addReg(ScratchReg));
  } else if (PPC::CRBITRCRegisterClass->hasSubClassEq(RC)) {

    unsigned Reg = 0;
    if (DestReg == PPC::CR0LT || DestReg == PPC::CR0GT ||
        DestReg == PPC::CR0EQ || DestReg == PPC::CR0UN)
      Reg = PPC::CR0;
    else if (DestReg == PPC::CR1LT || DestReg == PPC::CR1GT ||
             DestReg == PPC::CR1EQ || DestReg == PPC::CR1UN)
      Reg = PPC::CR1;
    else if (DestReg == PPC::CR2LT || DestReg == PPC::CR2GT ||
             DestReg == PPC::CR2EQ || DestReg == PPC::CR2UN)
      Reg = PPC::CR2;
    else if (DestReg == PPC::CR3LT || DestReg == PPC::CR3GT ||
             DestReg == PPC::CR3EQ || DestReg == PPC::CR3UN)
      Reg = PPC::CR3;
    else if (DestReg == PPC::CR4LT || DestReg == PPC::CR4GT ||
             DestReg == PPC::CR4EQ || DestReg == PPC::CR4UN)
      Reg = PPC::CR4;
    else if (DestReg == PPC::CR5LT || DestReg == PPC::CR5GT ||
             DestReg == PPC::CR5EQ || DestReg == PPC::CR5UN)
      Reg = PPC::CR5;
    else if (DestReg == PPC::CR6LT || DestReg == PPC::CR6GT ||
             DestReg == PPC::CR6EQ || DestReg == PPC::CR6UN)
      Reg = PPC::CR6;
    else if (DestReg == PPC::CR7LT || DestReg == PPC::CR7GT ||
             DestReg == PPC::CR7EQ || DestReg == PPC::CR7UN)
      Reg = PPC::CR7;

    return LoadRegFromStackSlot(MF, DL, Reg, FrameIdx,
                                PPC::CRRCRegisterClass, NewMIs);

  } else if (PPC::VRRCRegisterClass->hasSubClassEq(RC)) {
    // We don't have indexed addressing for vector loads.  Emit:
    // R0 = ADDI FI#
    // Dest = LVX 0, R0
    //
    // FIXME: We use R0 here, because it isn't available for RA.
    NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::ADDI), PPC::R0),
                                       FrameIdx, 0, 0));
    NewMIs.push_back(BuildMI(MF, DL, get(PPC::LVX),DestReg).addReg(PPC::R0)
                     .addReg(PPC::R0));
  } else {
    llvm_unreachable("Unknown regclass!");
  }
}

void
PPCInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
                                   MachineBasicBlock::iterator MI,
                                   unsigned DestReg, int FrameIdx,
                                   const TargetRegisterClass *RC,
                                   const TargetRegisterInfo *TRI) const {
  MachineFunction &MF = *MBB.getParent();
  SmallVector<MachineInstr*, 4> NewMIs;
  DebugLoc DL;
  if (MI != MBB.end()) DL = MI->getDebugLoc();
  LoadRegFromStackSlot(MF, DL, DestReg, FrameIdx, RC, NewMIs);
  for (unsigned i = 0, e = NewMIs.size(); i != e; ++i)
    MBB.insert(MI, NewMIs[i]);

  const MachineFrameInfo &MFI = *MF.getFrameInfo();
  MachineMemOperand *MMO =
    MF.getMachineMemOperand(
                MachinePointerInfo(PseudoSourceValue::getFixedStack(FrameIdx)),
                            MachineMemOperand::MOLoad,
                            MFI.getObjectSize(FrameIdx),
                            MFI.getObjectAlignment(FrameIdx));
  NewMIs.back()->addMemOperand(MF, MMO);
}

MachineInstr*
PPCInstrInfo::emitFrameIndexDebugValue(MachineFunction &MF,
                                       int FrameIx, uint64_t Offset,
                                       const MDNode *MDPtr,
                                       DebugLoc DL) const {
  MachineInstrBuilder MIB = BuildMI(MF, DL, get(PPC::DBG_VALUE));
  addFrameReference(MIB, FrameIx, 0, false).addImm(Offset).addMetadata(MDPtr);
  return &*MIB;
}

bool PPCInstrInfo::
ReverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) const {
  assert(Cond.size() == 2 && "Invalid PPC branch opcode!");
  // Leave the CR# the same, but invert the condition.
  Cond[0].setImm(PPC::InvertPredicate((PPC::Predicate)Cond[0].getImm()));
  return false;
}

/// GetInstSize - Return the number of bytes of code the specified
/// instruction may be.  This returns the maximum number of bytes.
///
unsigned PPCInstrInfo::GetInstSizeInBytes(const MachineInstr *MI) const {
  switch (MI->getOpcode()) {
  case PPC::INLINEASM: {       // Inline Asm: Variable size.
    const MachineFunction *MF = MI->getParent()->getParent();
    const char *AsmStr = MI->getOperand(0).getSymbolName();
    return getInlineAsmLength(AsmStr, *MF->getTarget().getMCAsmInfo());
  }
  case PPC::PROLOG_LABEL:
  case PPC::EH_LABEL:
  case PPC::GC_LABEL:
  case PPC::DBG_VALUE:
    return 0;
  default:
    return 4; // PowerPC instructions are all 4 bytes
  }
}