llvm.org GIT mirror llvm / release_39 lib / Transforms / Utils / LoopUnrollRuntime.cpp
release_39

Tree @release_39 (Download .tar.gz)

LoopUnrollRuntime.cpp @release_39raw · 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
//===-- UnrollLoopRuntime.cpp - Runtime Loop unrolling utilities ----------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements some loop unrolling utilities for loops with run-time
// trip counts.  See LoopUnroll.cpp for unrolling loops with compile-time
// trip counts.
//
// The functions in this file are used to generate extra code when the
// run-time trip count modulo the unroll factor is not 0.  When this is the
// case, we need to generate code to execute these 'left over' iterations.
//
// The current strategy generates an if-then-else sequence prior to the
// unrolled loop to execute the 'left over' iterations before or after the
// unrolled loop.
//
//===----------------------------------------------------------------------===//

#include "llvm/Transforms/Utils/UnrollLoop.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/LoopIterator.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Cloning.h"
#include <algorithm>

using namespace llvm;

#define DEBUG_TYPE "loop-unroll"

STATISTIC(NumRuntimeUnrolled,
          "Number of loops unrolled with run-time trip counts");

/// Connect the unrolling prolog code to the original loop.
/// The unrolling prolog code contains code to execute the
/// 'extra' iterations if the run-time trip count modulo the
/// unroll count is non-zero.
///
/// This function performs the following:
/// - Create PHI nodes at prolog end block to combine values
///   that exit the prolog code and jump around the prolog.
/// - Add a PHI operand to a PHI node at the loop exit block
///   for values that exit the prolog and go around the loop.
/// - Branch around the original loop if the trip count is less
///   than the unroll factor.
///
static void ConnectProlog(Loop *L, Value *BECount, unsigned Count,
                          BasicBlock *PrologExit, BasicBlock *PreHeader,
                          BasicBlock *NewPreHeader, ValueToValueMapTy &VMap,
                          DominatorTree *DT, LoopInfo *LI, bool PreserveLCSSA) {
  BasicBlock *Latch = L->getLoopLatch();
  assert(Latch && "Loop must have a latch");
  BasicBlock *PrologLatch = cast<BasicBlock>(VMap[Latch]);

  // Create a PHI node for each outgoing value from the original loop
  // (which means it is an outgoing value from the prolog code too).
  // The new PHI node is inserted in the prolog end basic block.
  // The new PHI node value is added as an operand of a PHI node in either
  // the loop header or the loop exit block.
  for (BasicBlock *Succ : successors(Latch)) {
    for (Instruction &BBI : *Succ) {
      PHINode *PN = dyn_cast<PHINode>(&BBI);
      // Exit when we passed all PHI nodes.
      if (!PN)
        break;
      // Add a new PHI node to the prolog end block and add the
      // appropriate incoming values.
      PHINode *NewPN = PHINode::Create(PN->getType(), 2, PN->getName() + ".unr",
                                       PrologExit->getFirstNonPHI());
      // Adding a value to the new PHI node from the original loop preheader.
      // This is the value that skips all the prolog code.
      if (L->contains(PN)) {
        NewPN->addIncoming(PN->getIncomingValueForBlock(NewPreHeader),
                           PreHeader);
      } else {
        NewPN->addIncoming(UndefValue::get(PN->getType()), PreHeader);
      }

      Value *V = PN->getIncomingValueForBlock(Latch);
      if (Instruction *I = dyn_cast<Instruction>(V)) {
        if (L->contains(I)) {
          V = VMap.lookup(I);
        }
      }
      // Adding a value to the new PHI node from the last prolog block
      // that was created.
      NewPN->addIncoming(V, PrologLatch);

      // Update the existing PHI node operand with the value from the
      // new PHI node.  How this is done depends on if the existing
      // PHI node is in the original loop block, or the exit block.
      if (L->contains(PN)) {
        PN->setIncomingValue(PN->getBasicBlockIndex(NewPreHeader), NewPN);
      } else {
        PN->addIncoming(NewPN, PrologExit);
      }
    }
  }

  // Create a branch around the original loop, which is taken if there are no
  // iterations remaining to be executed after running the prologue.
  Instruction *InsertPt = PrologExit->getTerminator();
  IRBuilder<> B(InsertPt);

  assert(Count != 0 && "nonsensical Count!");

  // If BECount <u (Count - 1) then (BECount + 1) % Count == (BECount + 1)
  // This means %xtraiter is (BECount + 1) and all of the iterations of this
  // loop were executed by the prologue.  Note that if BECount <u (Count - 1)
  // then (BECount + 1) cannot unsigned-overflow.
  Value *BrLoopExit =
      B.CreateICmpULT(BECount, ConstantInt::get(BECount->getType(), Count - 1));
  BasicBlock *Exit = L->getUniqueExitBlock();
  assert(Exit && "Loop must have a single exit block only");
  // Split the exit to maintain loop canonicalization guarantees
  SmallVector<BasicBlock*, 4> Preds(predecessors(Exit));
  SplitBlockPredecessors(Exit, Preds, ".unr-lcssa", DT, LI,
                         PreserveLCSSA);
  // Add the branch to the exit block (around the unrolled loop)
  B.CreateCondBr(BrLoopExit, Exit, NewPreHeader);
  InsertPt->eraseFromParent();
}

/// Connect the unrolling epilog code to the original loop.
/// The unrolling epilog code contains code to execute the
/// 'extra' iterations if the run-time trip count modulo the
/// unroll count is non-zero.
///
/// This function performs the following:
/// - Update PHI nodes at the unrolling loop exit and epilog loop exit
/// - Create PHI nodes at the unrolling loop exit to combine
///   values that exit the unrolling loop code and jump around it.
/// - Update PHI operands in the epilog loop by the new PHI nodes
/// - Branch around the epilog loop if extra iters (ModVal) is zero.
///
static void ConnectEpilog(Loop *L, Value *ModVal, BasicBlock *NewExit,
                          BasicBlock *Exit, BasicBlock *PreHeader,
                          BasicBlock *EpilogPreHeader, BasicBlock *NewPreHeader,
                          ValueToValueMapTy &VMap, DominatorTree *DT,
                          LoopInfo *LI, bool PreserveLCSSA)  {
  BasicBlock *Latch = L->getLoopLatch();
  assert(Latch && "Loop must have a latch");
  BasicBlock *EpilogLatch = cast<BasicBlock>(VMap[Latch]);

  // Loop structure should be the following:
  //
  // PreHeader
  // NewPreHeader
  //   Header
  //   ...
  //   Latch
  // NewExit (PN)
  // EpilogPreHeader
  //   EpilogHeader
  //   ...
  //   EpilogLatch
  // Exit (EpilogPN)

  // Update PHI nodes at NewExit and Exit.
  for (Instruction &BBI : *NewExit) {
    PHINode *PN = dyn_cast<PHINode>(&BBI);
    // Exit when we passed all PHI nodes.
    if (!PN)
      break;
    // PN should be used in another PHI located in Exit block as
    // Exit was split by SplitBlockPredecessors into Exit and NewExit
    // Basicaly it should look like:
    // NewExit:
    //   PN = PHI [I, Latch]
    // ...
    // Exit:
    //   EpilogPN = PHI [PN, EpilogPreHeader]
    //
    // There is EpilogPreHeader incoming block instead of NewExit as
    // NewExit was spilt 1 more time to get EpilogPreHeader.
    assert(PN->hasOneUse() && "The phi should have 1 use");
    PHINode *EpilogPN = cast<PHINode> (PN->use_begin()->getUser());
    assert(EpilogPN->getParent() == Exit && "EpilogPN should be in Exit block");

    // Add incoming PreHeader from branch around the Loop
    PN->addIncoming(UndefValue::get(PN->getType()), PreHeader);

    Value *V = PN->getIncomingValueForBlock(Latch);
    Instruction *I = dyn_cast<Instruction>(V);
    if (I && L->contains(I))
      // If value comes from an instruction in the loop add VMap value.
      V = VMap.lookup(I);
    // For the instruction out of the loop, constant or undefined value
    // insert value itself.
    EpilogPN->addIncoming(V, EpilogLatch);

    assert(EpilogPN->getBasicBlockIndex(EpilogPreHeader) >= 0 &&
          "EpilogPN should have EpilogPreHeader incoming block");
    // Change EpilogPreHeader incoming block to NewExit.
    EpilogPN->setIncomingBlock(EpilogPN->getBasicBlockIndex(EpilogPreHeader),
                               NewExit);
    // Now PHIs should look like:
    // NewExit:
    //   PN = PHI [I, Latch], [undef, PreHeader]
    // ...
    // Exit:
    //   EpilogPN = PHI [PN, NewExit], [VMap[I], EpilogLatch]
  }

  // Create PHI nodes at NewExit (from the unrolling loop Latch and PreHeader).
  // Update corresponding PHI nodes in epilog loop.
  for (BasicBlock *Succ : successors(Latch)) {
    // Skip this as we already updated phis in exit blocks.
    if (!L->contains(Succ))
      continue;
    for (Instruction &BBI : *Succ) {
      PHINode *PN = dyn_cast<PHINode>(&BBI);
      // Exit when we passed all PHI nodes.
      if (!PN)
        break;
      // Add new PHI nodes to the loop exit block and update epilog
      // PHIs with the new PHI values.
      PHINode *NewPN = PHINode::Create(PN->getType(), 2, PN->getName() + ".unr",
                                       NewExit->getFirstNonPHI());
      // Adding a value to the new PHI node from the unrolling loop preheader.
      NewPN->addIncoming(PN->getIncomingValueForBlock(NewPreHeader), PreHeader);
      // Adding a value to the new PHI node from the unrolling loop latch.
      NewPN->addIncoming(PN->getIncomingValueForBlock(Latch), Latch);

      // Update the existing PHI node operand with the value from the new PHI
      // node.  Corresponding instruction in epilog loop should be PHI.
      PHINode *VPN = cast<PHINode>(VMap[&BBI]);
      VPN->setIncomingValue(VPN->getBasicBlockIndex(EpilogPreHeader), NewPN);
    }
  }

  Instruction *InsertPt = NewExit->getTerminator();
  IRBuilder<> B(InsertPt);
  Value *BrLoopExit = B.CreateIsNotNull(ModVal, "lcmp.mod");
  assert(Exit && "Loop must have a single exit block only");
  // Split the exit to maintain loop canonicalization guarantees
  SmallVector<BasicBlock*, 4> Preds(predecessors(Exit));
  SplitBlockPredecessors(Exit, Preds, ".epilog-lcssa", DT, LI,
                         PreserveLCSSA);
  // Add the branch to the exit block (around the unrolling loop)
  B.CreateCondBr(BrLoopExit, EpilogPreHeader, Exit);
  InsertPt->eraseFromParent();
}

/// Create a clone of the blocks in a loop and connect them together.
/// If CreateRemainderLoop is false, loop structure will not be cloned,
/// otherwise a new loop will be created including all cloned blocks, and the
/// iterator of it switches to count NewIter down to 0.
/// The cloned blocks should be inserted between InsertTop and InsertBot.
/// If loop structure is cloned InsertTop should be new preheader, InsertBot
/// new loop exit.
///
static void CloneLoopBlocks(Loop *L, Value *NewIter,
                            const bool CreateRemainderLoop,
                            const bool UseEpilogRemainder,
                            BasicBlock *InsertTop, BasicBlock *InsertBot,
                            BasicBlock *Preheader,
                            std::vector<BasicBlock *> &NewBlocks,
                            LoopBlocksDFS &LoopBlocks, ValueToValueMapTy &VMap,
                            LoopInfo *LI) {
  StringRef suffix = UseEpilogRemainder ? "epil" : "prol";
  BasicBlock *Header = L->getHeader();
  BasicBlock *Latch = L->getLoopLatch();
  Function *F = Header->getParent();
  LoopBlocksDFS::RPOIterator BlockBegin = LoopBlocks.beginRPO();
  LoopBlocksDFS::RPOIterator BlockEnd = LoopBlocks.endRPO();
  Loop *NewLoop = nullptr;
  Loop *ParentLoop = L->getParentLoop();
  if (CreateRemainderLoop) {
    NewLoop = new Loop();
    if (ParentLoop)
      ParentLoop->addChildLoop(NewLoop);
    else
      LI->addTopLevelLoop(NewLoop);
  }

  // For each block in the original loop, create a new copy,
  // and update the value map with the newly created values.
  for (LoopBlocksDFS::RPOIterator BB = BlockBegin; BB != BlockEnd; ++BB) {
    BasicBlock *NewBB = CloneBasicBlock(*BB, VMap, "." + suffix, F);
    NewBlocks.push_back(NewBB);

    if (NewLoop)
      NewLoop->addBasicBlockToLoop(NewBB, *LI);
    else if (ParentLoop)
      ParentLoop->addBasicBlockToLoop(NewBB, *LI);

    VMap[*BB] = NewBB;
    if (Header == *BB) {
      // For the first block, add a CFG connection to this newly
      // created block.
      InsertTop->getTerminator()->setSuccessor(0, NewBB);
    }

    if (Latch == *BB) {
      // For the last block, if CreateRemainderLoop is false, create a direct
      // jump to InsertBot. If not, create a loop back to cloned head.
      VMap.erase((*BB)->getTerminator());
      BasicBlock *FirstLoopBB = cast<BasicBlock>(VMap[Header]);
      BranchInst *LatchBR = cast<BranchInst>(NewBB->getTerminator());
      IRBuilder<> Builder(LatchBR);
      if (!CreateRemainderLoop) {
        Builder.CreateBr(InsertBot);
      } else {
        PHINode *NewIdx = PHINode::Create(NewIter->getType(), 2,
                                          suffix + ".iter",
                                          FirstLoopBB->getFirstNonPHI());
        Value *IdxSub =
            Builder.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
                              NewIdx->getName() + ".sub");
        Value *IdxCmp =
            Builder.CreateIsNotNull(IdxSub, NewIdx->getName() + ".cmp");
        Builder.CreateCondBr(IdxCmp, FirstLoopBB, InsertBot);
        NewIdx->addIncoming(NewIter, InsertTop);
        NewIdx->addIncoming(IdxSub, NewBB);
      }
      LatchBR->eraseFromParent();
    }
  }

  // Change the incoming values to the ones defined in the preheader or
  // cloned loop.
  for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
    PHINode *NewPHI = cast<PHINode>(VMap[&*I]);
    if (!CreateRemainderLoop) {
      if (UseEpilogRemainder) {
        unsigned idx = NewPHI->getBasicBlockIndex(Preheader);
        NewPHI->setIncomingBlock(idx, InsertTop);
        NewPHI->removeIncomingValue(Latch, false);
      } else {
        VMap[&*I] = NewPHI->getIncomingValueForBlock(Preheader);
        cast<BasicBlock>(VMap[Header])->getInstList().erase(NewPHI);
      }
    } else {
      unsigned idx = NewPHI->getBasicBlockIndex(Preheader);
      NewPHI->setIncomingBlock(idx, InsertTop);
      BasicBlock *NewLatch = cast<BasicBlock>(VMap[Latch]);
      idx = NewPHI->getBasicBlockIndex(Latch);
      Value *InVal = NewPHI->getIncomingValue(idx);
      NewPHI->setIncomingBlock(idx, NewLatch);
      if (Value *V = VMap.lookup(InVal))
        NewPHI->setIncomingValue(idx, V);
    }
  }
  if (NewLoop) {
    // Add unroll disable metadata to disable future unrolling for this loop.
    SmallVector<Metadata *, 4> MDs;
    // Reserve first location for self reference to the LoopID metadata node.
    MDs.push_back(nullptr);
    MDNode *LoopID = NewLoop->getLoopID();
    if (LoopID) {
      // First remove any existing loop unrolling metadata.
      for (unsigned i = 1, ie = LoopID->getNumOperands(); i < ie; ++i) {
        bool IsUnrollMetadata = false;
        MDNode *MD = dyn_cast<MDNode>(LoopID->getOperand(i));
        if (MD) {
          const MDString *S = dyn_cast<MDString>(MD->getOperand(0));
          IsUnrollMetadata = S && S->getString().startswith("llvm.loop.unroll.");
        }
        if (!IsUnrollMetadata)
          MDs.push_back(LoopID->getOperand(i));
      }
    }

    LLVMContext &Context = NewLoop->getHeader()->getContext();
    SmallVector<Metadata *, 1> DisableOperands;
    DisableOperands.push_back(MDString::get(Context, "llvm.loop.unroll.disable"));
    MDNode *DisableNode = MDNode::get(Context, DisableOperands);
    MDs.push_back(DisableNode);

    MDNode *NewLoopID = MDNode::get(Context, MDs);
    // Set operand 0 to refer to the loop id itself.
    NewLoopID->replaceOperandWith(0, NewLoopID);
    NewLoop->setLoopID(NewLoopID);
  }
}

/// Insert code in the prolog/epilog code when unrolling a loop with a
/// run-time trip-count.
///
/// This method assumes that the loop unroll factor is total number
/// of loop bodies in the loop after unrolling. (Some folks refer
/// to the unroll factor as the number of *extra* copies added).
/// We assume also that the loop unroll factor is a power-of-two. So, after
/// unrolling the loop, the number of loop bodies executed is 2,
/// 4, 8, etc.  Note - LLVM converts the if-then-sequence to a switch
/// instruction in SimplifyCFG.cpp.  Then, the backend decides how code for
/// the switch instruction is generated.
///
/// ***Prolog case***
///        extraiters = tripcount % loopfactor
///        if (extraiters == 0) jump Loop:
///        else jump Prol:
/// Prol:  LoopBody;
///        extraiters -= 1                 // Omitted if unroll factor is 2.
///        if (extraiters != 0) jump Prol: // Omitted if unroll factor is 2.
///        if (tripcount < loopfactor) jump End:
/// Loop:
/// ...
/// End:
///
/// ***Epilog case***
///        extraiters = tripcount % loopfactor
///        if (tripcount < loopfactor) jump LoopExit:
///        unroll_iters = tripcount - extraiters
/// Loop:  LoopBody; (executes unroll_iter times);
///        unroll_iter -= 1
///        if (unroll_iter != 0) jump Loop:
/// LoopExit:
///        if (extraiters == 0) jump EpilExit:
/// Epil:  LoopBody; (executes extraiters times)
///        extraiters -= 1                 // Omitted if unroll factor is 2.
///        if (extraiters != 0) jump Epil: // Omitted if unroll factor is 2.
/// EpilExit:

bool llvm::UnrollRuntimeLoopRemainder(Loop *L, unsigned Count,
                                      bool AllowExpensiveTripCount,
                                      bool UseEpilogRemainder,
                                      LoopInfo *LI, ScalarEvolution *SE,
                                      DominatorTree *DT, bool PreserveLCSSA) {
  // for now, only unroll loops that contain a single exit
  if (!L->getExitingBlock())
    return false;

  // Make sure the loop is in canonical form, and there is a single
  // exit block only.
  if (!L->isLoopSimplifyForm())
    return false;
  BasicBlock *Exit = L->getUniqueExitBlock(); // successor out of loop
  if (!Exit)
    return false;

  // Use Scalar Evolution to compute the trip count. This allows more loops to
  // be unrolled than relying on induction var simplification.
  if (!SE)
    return false;

  // Only unroll loops with a computable trip count, and the trip count needs
  // to be an int value (allowing a pointer type is a TODO item).
  const SCEV *BECountSC = SE->getBackedgeTakenCount(L);
  if (isa<SCEVCouldNotCompute>(BECountSC) ||
      !BECountSC->getType()->isIntegerTy())
    return false;

  unsigned BEWidth = cast<IntegerType>(BECountSC->getType())->getBitWidth();

  // Add 1 since the backedge count doesn't include the first loop iteration.
  const SCEV *TripCountSC =
      SE->getAddExpr(BECountSC, SE->getConstant(BECountSC->getType(), 1));
  if (isa<SCEVCouldNotCompute>(TripCountSC))
    return false;

  BasicBlock *Header = L->getHeader();
  BasicBlock *PreHeader = L->getLoopPreheader();
  BranchInst *PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
  const DataLayout &DL = Header->getModule()->getDataLayout();
  SCEVExpander Expander(*SE, DL, "loop-unroll");
  if (!AllowExpensiveTripCount &&
      Expander.isHighCostExpansion(TripCountSC, L, PreHeaderBR))
    return false;

  // This constraint lets us deal with an overflowing trip count easily; see the
  // comment on ModVal below.
  if (Log2_32(Count) > BEWidth)
    return false;

  // If this loop is nested, then the loop unroller changes the code in the
  // parent loop, so the Scalar Evolution pass needs to be run again.
  if (Loop *ParentLoop = L->getParentLoop())
    SE->forgetLoop(ParentLoop);

  BasicBlock *Latch = L->getLoopLatch();

  // Loop structure is the following:
  //
  // PreHeader
  //   Header
  //   ...
  //   Latch
  // Exit

  BasicBlock *NewPreHeader;
  BasicBlock *NewExit = nullptr;
  BasicBlock *PrologExit = nullptr;
  BasicBlock *EpilogPreHeader = nullptr;
  BasicBlock *PrologPreHeader = nullptr;

  if (UseEpilogRemainder) {
    // If epilog remainder
    // Split PreHeader to insert a branch around loop for unrolling.
    NewPreHeader = SplitBlock(PreHeader, PreHeader->getTerminator(), DT, LI);
    NewPreHeader->setName(PreHeader->getName() + ".new");
    // Split Exit to create phi nodes from branch above.
    SmallVector<BasicBlock*, 4> Preds(predecessors(Exit));
    NewExit = SplitBlockPredecessors(Exit, Preds, ".unr-lcssa",
                                     DT, LI, PreserveLCSSA);
    // Split NewExit to insert epilog remainder loop.
    EpilogPreHeader = SplitBlock(NewExit, NewExit->getTerminator(), DT, LI);
    EpilogPreHeader->setName(Header->getName() + ".epil.preheader");
  } else {
    // If prolog remainder
    // Split the original preheader twice to insert prolog remainder loop
    PrologPreHeader = SplitEdge(PreHeader, Header, DT, LI);
    PrologPreHeader->setName(Header->getName() + ".prol.preheader");
    PrologExit = SplitBlock(PrologPreHeader, PrologPreHeader->getTerminator(),
                            DT, LI);
    PrologExit->setName(Header->getName() + ".prol.loopexit");
    // Split PrologExit to get NewPreHeader.
    NewPreHeader = SplitBlock(PrologExit, PrologExit->getTerminator(), DT, LI);
    NewPreHeader->setName(PreHeader->getName() + ".new");
  }
  // Loop structure should be the following:
  //  Epilog             Prolog
  //
  // PreHeader         PreHeader
  // *NewPreHeader     *PrologPreHeader
  //   Header          *PrologExit
  //   ...             *NewPreHeader
  //   Latch             Header
  // *NewExit            ...
  // *EpilogPreHeader    Latch
  // Exit              Exit

  // Calculate conditions for branch around loop for unrolling
  // in epilog case and around prolog remainder loop in prolog case.
  // Compute the number of extra iterations required, which is:
  //  extra iterations = run-time trip count % loop unroll factor
  PreHeaderBR = cast<BranchInst>(PreHeader->getTerminator());
  Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
                                            PreHeaderBR);
  Value *BECount = Expander.expandCodeFor(BECountSC, BECountSC->getType(),
                                          PreHeaderBR);
  IRBuilder<> B(PreHeaderBR);
  Value *ModVal;
  // Calculate ModVal = (BECount + 1) % Count.
  // Note that TripCount is BECount + 1.
  if (isPowerOf2_32(Count)) {
    // When Count is power of 2 we don't BECount for epilog case, however we'll
    // need it for a branch around unrolling loop for prolog case.
    ModVal = B.CreateAnd(TripCount, Count - 1, "xtraiter");
    //  1. There are no iterations to be run in the prolog/epilog loop.
    // OR
    //  2. The addition computing TripCount overflowed.
    //
    // If (2) is true, we know that TripCount really is (1 << BEWidth) and so
    // the number of iterations that remain to be run in the original loop is a
    // multiple Count == (1 << Log2(Count)) because Log2(Count) <= BEWidth (we
    // explicitly check this above).
  } else {
    // As (BECount + 1) can potentially unsigned overflow we count
    // (BECount % Count) + 1 which is overflow safe as BECount % Count < Count.
    Value *ModValTmp = B.CreateURem(BECount,
                                    ConstantInt::get(BECount->getType(),
                                                     Count));
    Value *ModValAdd = B.CreateAdd(ModValTmp,
                                   ConstantInt::get(ModValTmp->getType(), 1));
    // At that point (BECount % Count) + 1 could be equal to Count.
    // To handle this case we need to take mod by Count one more time.
    ModVal = B.CreateURem(ModValAdd,
                          ConstantInt::get(BECount->getType(), Count),
                          "xtraiter");
  }
  Value *BranchVal =
      UseEpilogRemainder ? B.CreateICmpULT(BECount,
                                           ConstantInt::get(BECount->getType(),
                                                            Count - 1)) :
                           B.CreateIsNotNull(ModVal, "lcmp.mod");
  BasicBlock *RemainderLoop = UseEpilogRemainder ? NewExit : PrologPreHeader;
  BasicBlock *UnrollingLoop = UseEpilogRemainder ? NewPreHeader : PrologExit;
  // Branch to either remainder (extra iterations) loop or unrolling loop.
  B.CreateCondBr(BranchVal, RemainderLoop, UnrollingLoop);
  PreHeaderBR->eraseFromParent();
  Function *F = Header->getParent();
  // Get an ordered list of blocks in the loop to help with the ordering of the
  // cloned blocks in the prolog/epilog code
  LoopBlocksDFS LoopBlocks(L);
  LoopBlocks.perform(LI);

  //
  // For each extra loop iteration, create a copy of the loop's basic blocks
  // and generate a condition that branches to the copy depending on the
  // number of 'left over' iterations.
  //
  std::vector<BasicBlock *> NewBlocks;
  ValueToValueMapTy VMap;

  // For unroll factor 2 remainder loop will have 1 iterations.
  // Do not create 1 iteration loop.
  bool CreateRemainderLoop = (Count != 2);

  // Clone all the basic blocks in the loop. If Count is 2, we don't clone
  // the loop, otherwise we create a cloned loop to execute the extra
  // iterations. This function adds the appropriate CFG connections.
  BasicBlock *InsertBot = UseEpilogRemainder ? Exit : PrologExit;
  BasicBlock *InsertTop = UseEpilogRemainder ? EpilogPreHeader : PrologPreHeader;
  CloneLoopBlocks(L, ModVal, CreateRemainderLoop, UseEpilogRemainder, InsertTop,
                  InsertBot, NewPreHeader, NewBlocks, LoopBlocks, VMap, LI);

  // Insert the cloned blocks into the function.
  F->getBasicBlockList().splice(InsertBot->getIterator(),
                                F->getBasicBlockList(),
                                NewBlocks[0]->getIterator(),
                                F->end());

  // Loop structure should be the following:
  //  Epilog             Prolog
  //
  // PreHeader         PreHeader
  // NewPreHeader      PrologPreHeader
  //   Header            PrologHeader
  //   ...               ...
  //   Latch             PrologLatch
  // NewExit           PrologExit
  // EpilogPreHeader   NewPreHeader
  //   EpilogHeader      Header
  //   ...               ...
  //   EpilogLatch       Latch
  // Exit              Exit

  // Rewrite the cloned instruction operands to use the values created when the
  // clone is created.
  for (BasicBlock *BB : NewBlocks) {
    for (Instruction &I : *BB) {
      RemapInstruction(&I, VMap,
                       RF_NoModuleLevelChanges | RF_IgnoreMissingLocals);
    }
  }

  if (UseEpilogRemainder) {
    // Connect the epilog code to the original loop and update the
    // PHI functions.
    ConnectEpilog(L, ModVal, NewExit, Exit, PreHeader,
                  EpilogPreHeader, NewPreHeader, VMap, DT, LI,
                  PreserveLCSSA);

    // Update counter in loop for unrolling.
    // I should be multiply of Count.
    IRBuilder<> B2(NewPreHeader->getTerminator());
    Value *TestVal = B2.CreateSub(TripCount, ModVal, "unroll_iter");
    BranchInst *LatchBR = cast<BranchInst>(Latch->getTerminator());
    B2.SetInsertPoint(LatchBR);
    PHINode *NewIdx = PHINode::Create(TestVal->getType(), 2, "niter",
                                      Header->getFirstNonPHI());
    Value *IdxSub =
        B2.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
                     NewIdx->getName() + ".nsub");
    Value *IdxCmp;
    if (LatchBR->getSuccessor(0) == Header)
      IdxCmp = B2.CreateIsNotNull(IdxSub, NewIdx->getName() + ".ncmp");
    else
      IdxCmp = B2.CreateIsNull(IdxSub, NewIdx->getName() + ".ncmp");
    NewIdx->addIncoming(TestVal, NewPreHeader);
    NewIdx->addIncoming(IdxSub, Latch);
    LatchBR->setCondition(IdxCmp);
  } else {
    // Connect the prolog code to the original loop and update the
    // PHI functions.
    ConnectProlog(L, BECount, Count, PrologExit, PreHeader, NewPreHeader,
                  VMap, DT, LI, PreserveLCSSA);
  }
  NumRuntimeUnrolled++;
  return true;
}