llvm.org GIT mirror llvm / 0392724 lib / Target / AMDGPU / SILowerControlFlow.cpp
0392724

Tree @0392724 (Download .tar.gz)

SILowerControlFlow.cpp @0392724raw · 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
//===-- SILowerControlFlow.cpp - Use predicates for control flow ----------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
/// \file
/// This pass lowers the pseudo control flow instructions to real
/// machine instructions.
///
/// All control flow is handled using predicated instructions and
/// a predicate stack.  Each Scalar ALU controls the operations of 64 Vector
/// ALUs.  The Scalar ALU can update the predicate for any of the Vector ALUs
/// by writting to the 64-bit EXEC register (each bit corresponds to a
/// single vector ALU).  Typically, for predicates, a vector ALU will write
/// to its bit of the VCC register (like EXEC VCC is 64-bits, one for each
/// Vector ALU) and then the ScalarALU will AND the VCC register with the
/// EXEC to update the predicates.
///
/// For example:
/// %vcc = V_CMP_GT_F32 %vgpr1, %vgpr2
/// %sgpr0 = SI_IF %vcc
///   %vgpr0 = V_ADD_F32 %vgpr0, %vgpr0
/// %sgpr0 = SI_ELSE %sgpr0
///   %vgpr0 = V_SUB_F32 %vgpr0, %vgpr0
/// SI_END_CF %sgpr0
///
/// becomes:
///
/// %sgpr0 = S_AND_SAVEEXEC_B64 %vcc  // Save and update the exec mask
/// %sgpr0 = S_XOR_B64 %sgpr0, %exec  // Clear live bits from saved exec mask
/// S_CBRANCH_EXECZ label0            // This instruction is an optional
///                                   // optimization which allows us to
///                                   // branch if all the bits of
///                                   // EXEC are zero.
/// %vgpr0 = V_ADD_F32 %vgpr0, %vgpr0 // Do the IF block of the branch
///
/// label0:
/// %sgpr0 = S_OR_SAVEEXEC_B64 %exec   // Restore the exec mask for the Then block
/// %exec = S_XOR_B64 %sgpr0, %exec    // Clear live bits from saved exec mask
/// S_BRANCH_EXECZ label1              // Use our branch optimization
///                                    // instruction again.
/// %vgpr0 = V_SUB_F32 %vgpr0, %vgpr   // Do the THEN block
/// label1:
/// %exec = S_OR_B64 %exec, %sgpr0     // Re-enable saved exec mask bits
//===----------------------------------------------------------------------===//

#include "AMDGPU.h"
#include "AMDGPUSubtarget.h"
#include "SIInstrInfo.h"
#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/CodeGen/LiveIntervals.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/SlotIndexes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/Pass.h"
#include <cassert>
#include <iterator>

using namespace llvm;

#define DEBUG_TYPE "si-lower-control-flow"

namespace {

class SILowerControlFlow : public MachineFunctionPass {
private:
  const SIRegisterInfo *TRI = nullptr;
  const SIInstrInfo *TII = nullptr;
  LiveIntervals *LIS = nullptr;
  MachineRegisterInfo *MRI = nullptr;

  const TargetRegisterClass *BoolRC = nullptr;
  unsigned AndOpc;
  unsigned OrOpc;
  unsigned XorOpc;
  unsigned MovTermOpc;
  unsigned Andn2TermOpc;
  unsigned XorTermrOpc;
  unsigned OrSaveExecOpc;
  unsigned Exec;

  void emitIf(MachineInstr &MI);
  void emitElse(MachineInstr &MI);
  void emitIfBreak(MachineInstr &MI);
  void emitLoop(MachineInstr &MI);
  void emitEndCf(MachineInstr &MI);

  void findMaskOperands(MachineInstr &MI, unsigned OpNo,
                        SmallVectorImpl<MachineOperand> &Src) const;

  void combineMasks(MachineInstr &MI);

public:
  static char ID;

  SILowerControlFlow() : MachineFunctionPass(ID) {}

  bool runOnMachineFunction(MachineFunction &MF) override;

  StringRef getPassName() const override {
    return "SI Lower control flow pseudo instructions";
  }

  void getAnalysisUsage(AnalysisUsage &AU) const override {
    // Should preserve the same set that TwoAddressInstructions does.
    AU.addPreserved<SlotIndexes>();
    AU.addPreserved<LiveIntervals>();
    AU.addPreservedID(LiveVariablesID);
    AU.addPreservedID(MachineLoopInfoID);
    AU.addPreservedID(MachineDominatorsID);
    AU.setPreservesCFG();
    MachineFunctionPass::getAnalysisUsage(AU);
  }
};

} // end anonymous namespace

char SILowerControlFlow::ID = 0;

INITIALIZE_PASS(SILowerControlFlow, DEBUG_TYPE,
               "SI lower control flow", false, false)

static void setImpSCCDefDead(MachineInstr &MI, bool IsDead) {
  MachineOperand &ImpDefSCC = MI.getOperand(3);
  assert(ImpDefSCC.getReg() == AMDGPU::SCC && ImpDefSCC.isDef());

  ImpDefSCC.setIsDead(IsDead);
}

char &llvm::SILowerControlFlowID = SILowerControlFlow::ID;

static bool isSimpleIf(const MachineInstr &MI, const MachineRegisterInfo *MRI,
                       const SIInstrInfo *TII) {
  unsigned SaveExecReg = MI.getOperand(0).getReg();
  auto U = MRI->use_instr_nodbg_begin(SaveExecReg);

  if (U == MRI->use_instr_nodbg_end() ||
      std::next(U) != MRI->use_instr_nodbg_end() ||
      U->getOpcode() != AMDGPU::SI_END_CF)
    return false;

  // Check for SI_KILL_*_TERMINATOR on path from if to endif.
  // if there is any such terminator simplififcations are not safe.
  auto SMBB = MI.getParent();
  auto EMBB = U->getParent();
  DenseSet<const MachineBasicBlock*> Visited;
  SmallVector<MachineBasicBlock*, 4> Worklist(SMBB->succ_begin(),
                                              SMBB->succ_end());

  while (!Worklist.empty()) {
    MachineBasicBlock *MBB = Worklist.pop_back_val();

    if (MBB == EMBB || !Visited.insert(MBB).second)
      continue;
    for(auto &Term : MBB->terminators())
      if (TII->isKillTerminator(Term.getOpcode()))
        return false;

    Worklist.append(MBB->succ_begin(), MBB->succ_end());
  }

  return true;
}

void SILowerControlFlow::emitIf(MachineInstr &MI) {
  MachineBasicBlock &MBB = *MI.getParent();
  const DebugLoc &DL = MI.getDebugLoc();
  MachineBasicBlock::iterator I(&MI);

  MachineOperand &SaveExec = MI.getOperand(0);
  MachineOperand &Cond = MI.getOperand(1);
  assert(SaveExec.getSubReg() == AMDGPU::NoSubRegister &&
         Cond.getSubReg() == AMDGPU::NoSubRegister);

  Register SaveExecReg = SaveExec.getReg();

  MachineOperand &ImpDefSCC = MI.getOperand(4);
  assert(ImpDefSCC.getReg() == AMDGPU::SCC && ImpDefSCC.isDef());

  // If there is only one use of save exec register and that use is SI_END_CF,
  // we can optimize SI_IF by returning the full saved exec mask instead of
  // just cleared bits.
  bool SimpleIf = isSimpleIf(MI, MRI, TII);

  // Add an implicit def of exec to discourage scheduling VALU after this which
  // will interfere with trying to form s_and_saveexec_b64 later.
  Register CopyReg = SimpleIf ? SaveExecReg
                       : MRI->createVirtualRegister(BoolRC);
  MachineInstr *CopyExec =
    BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), CopyReg)
    .addReg(Exec)
    .addReg(Exec, RegState::ImplicitDefine);

  unsigned Tmp = MRI->createVirtualRegister(BoolRC);

  MachineInstr *And =
    BuildMI(MBB, I, DL, TII->get(AndOpc), Tmp)
    .addReg(CopyReg)
    .add(Cond);

  setImpSCCDefDead(*And, true);

  MachineInstr *Xor = nullptr;
  if (!SimpleIf) {
    Xor =
      BuildMI(MBB, I, DL, TII->get(XorOpc), SaveExecReg)
      .addReg(Tmp)
      .addReg(CopyReg);
    setImpSCCDefDead(*Xor, ImpDefSCC.isDead());
  }

  // Use a copy that is a terminator to get correct spill code placement it with
  // fast regalloc.
  MachineInstr *SetExec =
    BuildMI(MBB, I, DL, TII->get(MovTermOpc), Exec)
    .addReg(Tmp, RegState::Kill);

  // Insert a pseudo terminator to help keep the verifier happy. This will also
  // be used later when inserting skips.
  MachineInstr *NewBr = BuildMI(MBB, I, DL, TII->get(AMDGPU::SI_MASK_BRANCH))
                            .add(MI.getOperand(2));

  if (!LIS) {
    MI.eraseFromParent();
    return;
  }

  LIS->InsertMachineInstrInMaps(*CopyExec);

  // Replace with and so we don't need to fix the live interval for condition
  // register.
  LIS->ReplaceMachineInstrInMaps(MI, *And);

  if (!SimpleIf)
    LIS->InsertMachineInstrInMaps(*Xor);
  LIS->InsertMachineInstrInMaps(*SetExec);
  LIS->InsertMachineInstrInMaps(*NewBr);

  LIS->removeAllRegUnitsForPhysReg(AMDGPU::EXEC);
  MI.eraseFromParent();

  // FIXME: Is there a better way of adjusting the liveness? It shouldn't be
  // hard to add another def here but I'm not sure how to correctly update the
  // valno.
  LIS->removeInterval(SaveExecReg);
  LIS->createAndComputeVirtRegInterval(SaveExecReg);
  LIS->createAndComputeVirtRegInterval(Tmp);
  if (!SimpleIf)
    LIS->createAndComputeVirtRegInterval(CopyReg);
}

void SILowerControlFlow::emitElse(MachineInstr &MI) {
  MachineBasicBlock &MBB = *MI.getParent();
  const DebugLoc &DL = MI.getDebugLoc();

  Register DstReg = MI.getOperand(0).getReg();
  assert(MI.getOperand(0).getSubReg() == AMDGPU::NoSubRegister);

  bool ExecModified = MI.getOperand(3).getImm() != 0;
  MachineBasicBlock::iterator Start = MBB.begin();

  // We are running before TwoAddressInstructions, and si_else's operands are
  // tied. In order to correctly tie the registers, split this into a copy of
  // the src like it does.
  Register CopyReg = MRI->createVirtualRegister(BoolRC);
  MachineInstr *CopyExec =
    BuildMI(MBB, Start, DL, TII->get(AMDGPU::COPY), CopyReg)
      .add(MI.getOperand(1)); // Saved EXEC

  // This must be inserted before phis and any spill code inserted before the
  // else.
  Register SaveReg = ExecModified ?
    MRI->createVirtualRegister(BoolRC) : DstReg;
  MachineInstr *OrSaveExec =
    BuildMI(MBB, Start, DL, TII->get(OrSaveExecOpc), SaveReg)
    .addReg(CopyReg);

  MachineBasicBlock *DestBB = MI.getOperand(2).getMBB();

  MachineBasicBlock::iterator ElsePt(MI);

  if (ExecModified) {
    MachineInstr *And =
      BuildMI(MBB, ElsePt, DL, TII->get(AndOpc), DstReg)
      .addReg(Exec)
      .addReg(SaveReg);

    if (LIS)
      LIS->InsertMachineInstrInMaps(*And);
  }

  MachineInstr *Xor =
    BuildMI(MBB, ElsePt, DL, TII->get(XorTermrOpc), Exec)
    .addReg(Exec)
    .addReg(DstReg);

  MachineInstr *Branch =
    BuildMI(MBB, ElsePt, DL, TII->get(AMDGPU::SI_MASK_BRANCH))
    .addMBB(DestBB);

  if (!LIS) {
    MI.eraseFromParent();
    return;
  }

  LIS->RemoveMachineInstrFromMaps(MI);
  MI.eraseFromParent();

  LIS->InsertMachineInstrInMaps(*CopyExec);
  LIS->InsertMachineInstrInMaps(*OrSaveExec);

  LIS->InsertMachineInstrInMaps(*Xor);
  LIS->InsertMachineInstrInMaps(*Branch);

  // src reg is tied to dst reg.
  LIS->removeInterval(DstReg);
  LIS->createAndComputeVirtRegInterval(DstReg);
  LIS->createAndComputeVirtRegInterval(CopyReg);
  if (ExecModified)
    LIS->createAndComputeVirtRegInterval(SaveReg);

  // Let this be recomputed.
  LIS->removeAllRegUnitsForPhysReg(AMDGPU::EXEC);
}

void SILowerControlFlow::emitIfBreak(MachineInstr &MI) {
  MachineBasicBlock &MBB = *MI.getParent();
  const DebugLoc &DL = MI.getDebugLoc();
  auto Dst = MI.getOperand(0).getReg();

  // Skip ANDing with exec if the break condition is already masked by exec
  // because it is a V_CMP in the same basic block. (We know the break
  // condition operand was an i1 in IR, so if it is a VALU instruction it must
  // be one with a carry-out.)
  bool SkipAnding = false;
  if (MI.getOperand(1).isReg()) {
    if (MachineInstr *Def = MRI->getUniqueVRegDef(MI.getOperand(1).getReg())) {
      SkipAnding = Def->getParent() == MI.getParent()
          && SIInstrInfo::isVALU(*Def);
    }
  }

  // AND the break condition operand with exec, then OR that into the "loop
  // exit" mask.
  MachineInstr *And = nullptr, *Or = nullptr;
  if (!SkipAnding) {
    And = BuildMI(MBB, &MI, DL, TII->get(AndOpc), Dst)
             .addReg(Exec)
             .add(MI.getOperand(1));
    Or = BuildMI(MBB, &MI, DL, TII->get(OrOpc), Dst)
             .addReg(Dst)
             .add(MI.getOperand(2));
  } else
    Or = BuildMI(MBB, &MI, DL, TII->get(OrOpc), Dst)
             .add(MI.getOperand(1))
             .add(MI.getOperand(2));

  if (LIS) {
    if (And)
      LIS->InsertMachineInstrInMaps(*And);
    LIS->ReplaceMachineInstrInMaps(MI, *Or);
  }

  MI.eraseFromParent();
}

void SILowerControlFlow::emitLoop(MachineInstr &MI) {
  MachineBasicBlock &MBB = *MI.getParent();
  const DebugLoc &DL = MI.getDebugLoc();

  MachineInstr *AndN2 =
      BuildMI(MBB, &MI, DL, TII->get(Andn2TermOpc), Exec)
          .addReg(Exec)
          .add(MI.getOperand(0));

  MachineInstr *Branch =
      BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_CBRANCH_EXECNZ))
          .add(MI.getOperand(1));

  if (LIS) {
    LIS->ReplaceMachineInstrInMaps(MI, *AndN2);
    LIS->InsertMachineInstrInMaps(*Branch);
  }

  MI.eraseFromParent();
}

void SILowerControlFlow::emitEndCf(MachineInstr &MI) {
  MachineBasicBlock &MBB = *MI.getParent();
  const DebugLoc &DL = MI.getDebugLoc();

  MachineBasicBlock::iterator InsPt = MBB.begin();
  MachineInstr *NewMI =
      BuildMI(MBB, InsPt, DL, TII->get(OrOpc), Exec)
          .addReg(Exec)
          .add(MI.getOperand(0));

  if (LIS)
    LIS->ReplaceMachineInstrInMaps(MI, *NewMI);

  MI.eraseFromParent();

  if (LIS)
    LIS->handleMove(*NewMI);
}

// Returns replace operands for a logical operation, either single result
// for exec or two operands if source was another equivalent operation.
void SILowerControlFlow::findMaskOperands(MachineInstr &MI, unsigned OpNo,
       SmallVectorImpl<MachineOperand> &Src) const {
  MachineOperand &Op = MI.getOperand(OpNo);
  if (!Op.isReg() || !TargetRegisterInfo::isVirtualRegister(Op.getReg())) {
    Src.push_back(Op);
    return;
  }

  MachineInstr *Def = MRI->getUniqueVRegDef(Op.getReg());
  if (!Def || Def->getParent() != MI.getParent() ||
      !(Def->isFullCopy() || (Def->getOpcode() == MI.getOpcode())))
    return;

  // Make sure we do not modify exec between def and use.
  // A copy with implcitly defined exec inserted earlier is an exclusion, it
  // does not really modify exec.
  for (auto I = Def->getIterator(); I != MI.getIterator(); ++I)
    if (I->modifiesRegister(AMDGPU::EXEC, TRI) &&
        !(I->isCopy() && I->getOperand(0).getReg() != Exec))
      return;

  for (const auto &SrcOp : Def->explicit_operands())
    if (SrcOp.isReg() && SrcOp.isUse() &&
        (TargetRegisterInfo::isVirtualRegister(SrcOp.getReg()) ||
        SrcOp.getReg() == Exec))
      Src.push_back(SrcOp);
}

// Search and combine pairs of equivalent instructions, like
// S_AND_B64 x, (S_AND_B64 x, y) => S_AND_B64 x, y
// S_OR_B64  x, (S_OR_B64  x, y) => S_OR_B64  x, y
// One of the operands is exec mask.
void SILowerControlFlow::combineMasks(MachineInstr &MI) {
  assert(MI.getNumExplicitOperands() == 3);
  SmallVector<MachineOperand, 4> Ops;
  unsigned OpToReplace = 1;
  findMaskOperands(MI, 1, Ops);
  if (Ops.size() == 1) OpToReplace = 2; // First operand can be exec or its copy
  findMaskOperands(MI, 2, Ops);
  if (Ops.size() != 3) return;

  unsigned UniqueOpndIdx;
  if (Ops[0].isIdenticalTo(Ops[1])) UniqueOpndIdx = 2;
  else if (Ops[0].isIdenticalTo(Ops[2])) UniqueOpndIdx = 1;
  else if (Ops[1].isIdenticalTo(Ops[2])) UniqueOpndIdx = 1;
  else return;

  unsigned Reg = MI.getOperand(OpToReplace).getReg();
  MI.RemoveOperand(OpToReplace);
  MI.addOperand(Ops[UniqueOpndIdx]);
  if (MRI->use_empty(Reg))
    MRI->getUniqueVRegDef(Reg)->eraseFromParent();
}

bool SILowerControlFlow::runOnMachineFunction(MachineFunction &MF) {
  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
  TII = ST.getInstrInfo();
  TRI = &TII->getRegisterInfo();

  // This doesn't actually need LiveIntervals, but we can preserve them.
  LIS = getAnalysisIfAvailable<LiveIntervals>();
  MRI = &MF.getRegInfo();
  BoolRC = TRI->getBoolRC();

  if (ST.isWave32()) {
    AndOpc = AMDGPU::S_AND_B32;
    OrOpc = AMDGPU::S_OR_B32;
    XorOpc = AMDGPU::S_XOR_B32;
    MovTermOpc = AMDGPU::S_MOV_B32_term;
    Andn2TermOpc = AMDGPU::S_ANDN2_B32_term;
    XorTermrOpc = AMDGPU::S_XOR_B32_term;
    OrSaveExecOpc = AMDGPU::S_OR_SAVEEXEC_B32;
    Exec = AMDGPU::EXEC_LO;
  } else {
    AndOpc = AMDGPU::S_AND_B64;
    OrOpc = AMDGPU::S_OR_B64;
    XorOpc = AMDGPU::S_XOR_B64;
    MovTermOpc = AMDGPU::S_MOV_B64_term;
    Andn2TermOpc = AMDGPU::S_ANDN2_B64_term;
    XorTermrOpc = AMDGPU::S_XOR_B64_term;
    OrSaveExecOpc = AMDGPU::S_OR_SAVEEXEC_B64;
    Exec = AMDGPU::EXEC;
  }

  MachineFunction::iterator NextBB;
  for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
       BI != BE; BI = NextBB) {
    NextBB = std::next(BI);
    MachineBasicBlock &MBB = *BI;

    MachineBasicBlock::iterator I, Next, Last;

    for (I = MBB.begin(), Last = MBB.end(); I != MBB.end(); I = Next) {
      Next = std::next(I);
      MachineInstr &MI = *I;

      switch (MI.getOpcode()) {
      case AMDGPU::SI_IF:
        emitIf(MI);
        break;

      case AMDGPU::SI_ELSE:
        emitElse(MI);
        break;

      case AMDGPU::SI_IF_BREAK:
        emitIfBreak(MI);
        break;

      case AMDGPU::SI_LOOP:
        emitLoop(MI);
        break;

      case AMDGPU::SI_END_CF:
        emitEndCf(MI);
        break;

      case AMDGPU::S_AND_B64:
      case AMDGPU::S_OR_B64:
      case AMDGPU::S_AND_B32:
      case AMDGPU::S_OR_B32:
        // Cleanup bit manipulations on exec mask
        combineMasks(MI);
        Last = I;
        continue;

      default:
        Last = I;
        continue;
      }

      // Replay newly inserted code to combine masks
      Next = (Last == MBB.end()) ? MBB.begin() : Last;
    }
  }

  return true;
}