llvm.org GIT mirror llvm / release_70 lib / CodeGen / PostRASchedulerList.cpp
release_70

Tree @release_70 (Download .tar.gz)

PostRASchedulerList.cpp @release_70

72f1596
e7e7d0d
 
 
4ee451d
 
e7e7d0d
 
 
 
 
 
 
 
 
 
 
 
 
 
3487771
d04a8d4
2e7be61
d04a8d4
 
343f0c0
3f23744
e7e7d0d
3f23744
21d9003
f6071e1
1525260
ed395c8
2836c28
d04a8d4
4831923
e3a9b4c
6a6190d
e3a9b4c
 
0f38c60
e10deca
e7e7d0d
c25e758
3a5f0d4
e7e7d0d
 
8677f2f
 
2836c28
343f0c0
2e7be61
343f0c0
471850a
5b1b448
471850a
 
 
 
9843a93
2e7be61
21d9003
2e7be61
 
 
2836c28
1f15228
 
 
 
 
 
 
 
 
 
ada0ef8
 
e7e7d0d
6726b6d
86050dc
fa796dd
a70dca1
e7e7d0d
 
c7d081b
21d9003
9f998de
845012e
9146833
c7d081b
3f23744
 
 
 
 
 
 
fadd113
 
690a3cb
fadd113
 
9f998de
7710e25
27d8b83
f7e0423
 
 
 
343f0c0
 
 
6726b6d
343f0c0
c1ae8c9
343f0c0
9001303
343f0c0
 
 
 
 
 
c1ae8c9
 
 
2e7be61
 
 
c1ae8c9
 
480c529
47c1445
 
 
0c79dd7
 
 
d2763f6
 
 
 
 
 
21d9003
2da8bc8
a046b41
 
 
 
2da8bc8
c16fc54
343f0c0
953be89
9e64bbb
 
9f998de
9e64bbb
d2763f6
 
 
47c1445
9f998de
 
 
 
47c1445
 
9f998de
47c1445
480c529
9e64bbb
9f998de
9001303
84b454d
 
c1ae8c9
 
 
a26cd9c
480c529
953be89
c1ae8c9
9f998de
480c529
c1ae8c9
0c79dd7
 
 
557bbe6
 
 
 
9001303
73ba69b
8bacb53
e7e7d0d
cd52a7a
e7e7d0d
1dd8c85
 
94c4904
1dd8c85
 
2da8bc8
a046b41
 
 
 
e9564b2
8e62b29
9f85dcc
eb271ec
2da8bc8
eb271ec
9f85dcc
0c79dd7
6a8c7bf
 
 
 
2da8bc8
5b1b448
fa796dd
5b1b448
4ba8443
2da8bc8
 
 
 
 
 
 
47c1445
 
 
 
d2763f6
 
47c1445
 
 
 
 
0818e78
 
 
 
 
47c1445
 
 
1d03d38
73ba69b
88d2075
73ba69b
 
 
 
 
 
 
77e300e
73ba69b
f7e0423
 
 
 
 
 
 
27d8b83
 
 
 
 
8ce6c58
f7e0423
 
 
343f0c0
d318139
5fa58a5
 
6035518
2da8bc8
9146833
c7d081b
 
fa796dd
5bf7c2a
ddfd137
 
44d2382
27d8b83
 
 
 
 
 
0dad89f
4c3715c
 
5b1b448
 
 
 
 
4c3715c
 
0818e78
e7e7d0d
a046b41
2da8bc8
79ce276
e7e7d0d
ba91007
1f15228
 
 
 
 
 
ca0df55
 
1f15228
 
 
9e64bbb
ba91007
9e64bbb
f711939
 
ba91007
 
 
e16c630
d2763f6
976647d
 
 
e16c630
ba91007
d2763f6
953be89
47c1445
af1d8ca
e16c630
d2763f6
e16c630
f711939
9e64bbb
e16c630
 
47ac0f0
 
ba91007
1274ced
ba91007
d2763f6
953be89
47c1445
af1d8ca
9e64bbb
 
953be89
88a589c
5e41178
0248ff9
343f0c0
e7e7d0d
 
 
9001303
9e64bbb
 
 
953be89
9e64bbb
953be89
9e64bbb
2e7be61
d94a4e5
4ba8443
2e7be61
9e64bbb
 
 
 
953be89
c9a5b9e
953be89
343f0c0
4ba8443
9001303
68675c6
 
9001303
557bbe6
21d9003
 
 
 
 
 
47c1445
953be89
9001303
2e7be61
21d9003
 
 
05a0bf3
 
0818e78
 
 
 
 
 
d94a4e5
343f0c0
557bbe6
343f0c0
 
 
9e64bbb
 
 
a26cd9c
4ba8443
cf46b5a
9e64bbb
 
 
 
953be89
4ba8443
2e7be61
9e64bbb
 
953be89
9e64bbb
 
0c79dd7
 
 
 
 
 
343f0c0
 
 
 
 
ae692f2
557bbe6
54e4c36
c277ab0
cf6b613
ae692f2
 
 
343f0c0
c277ab0
e1b2129
343f0c0
e1b2129
4ba8443
343f0c0
 
c277ab0
 
89fd437
15ab359
 
 
 
 
 
 
 
 
9001303
9e64bbb
 
 
343f0c0
9e64bbb
 
 
557bbe6
9e64bbb
4de099d
557bbe6
4de099d
343f0c0
 
 
 
 
557bbe6
0818e78
 
9001303
343f0c0
9001303
4de099d
557bbe6
343f0c0
557bbe6
343f0c0
953be89
343f0c0
 
8bacb53
 
0818e78
8bacb53
4ba8443
8bacb53
 
 
343f0c0
 
557bbe6
343f0c0
9001303
4de099d
 
 
 
 
 
9e64bbb
557bbe6
9e64bbb
557bbe6
343f0c0
 
ae692f2
343f0c0
 
 
 
9e64bbb
2ffb0ce
 
be441c0
2ffb0ce
343f0c0
 
2836c28
343f0c0
 
 
 
3f23744
343f0c0
557bbe6
343f0c0
 
 
 
 
557bbe6
 
343f0c0
c93d837
0818e78
 
c93d837
4ba8443
2836c28
 
 
 
 
2da8bc8
2836c28
8bacb53
 
7710e25
 
 
 
8bacb53
 
 
 
 
 
 
2836c28
 
 
 
 
 
 
 
8bacb53
 
 
 
 
0818e78
 
8bacb53
 
 
 
 
4ba8443
8bacb53
 
2836c28
 
 
 
 
 
4de099d
343f0c0
8bacb53
 
 
 
 
4de099d
557bbe6
2836c28
be441c0
cf9aa28
0818e78
 
cf9aa28
 
 
 
2836c28
be441c0
0818e78
2ffb0ce
 
 
 
0818e78
2ffb0ce
557bbe6
2ffb0ce
 
 
 
8bacb53
2ffb0ce
 
2836c28
be441c0
343f0c0
 
 
 
4c72720
 
 
 
 
 
 
 
343f0c0
84b454d
 
 
68675c6
84b454d
 
 
68675c6
84b454d
 
 
 
68675c6
84b454d
 
68675c6
84b454d
 
 
 
d628f19
84b454d
 
 
 
 
d628f19
84b454d
 
 
 
 
4ba8443
84b454d
  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
//===----- SchedulePostRAList.cpp - list scheduler ------------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This implements a top-down list scheduler, using standard algorithms.
// The basic approach uses a priority queue of available nodes to schedule.
// One at a time, nodes are taken from the priority queue (thus in priority
// order), checked for legality to schedule, and emitted if legal.
//
// Nodes may not be legal to schedule either due to structural hazards (e.g.
// pipeline or resource constraints) or because an input to the instruction has
// not completed execution.
//
//===----------------------------------------------------------------------===//

#include "AggressiveAntiDepBreaker.h"
#include "AntiDepBreaker.h"
#include "CriticalAntiDepBreaker.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/CodeGen/LatencyPriorityQueue.h"
#include "llvm/CodeGen/MachineDominators.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineLoopInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/RegisterClassInfo.h"
#include "llvm/CodeGen/ScheduleDAGInstrs.h"
#include "llvm/CodeGen/ScheduleHazardRecognizer.h"
#include "llvm/CodeGen/SchedulerRegistry.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetPassConfig.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/Config/llvm-config.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;

#define DEBUG_TYPE "post-RA-sched"

STATISTIC(NumNoops, "Number of noops inserted");
STATISTIC(NumStalls, "Number of pipeline stalls");
STATISTIC(NumFixedAnti, "Number of fixed anti-dependencies");

// Post-RA scheduling is enabled with
// TargetSubtargetInfo.enablePostRAScheduler(). This flag can be used to
// override the target.
static cl::opt<bool>
EnablePostRAScheduler("post-RA-scheduler",
                       cl::desc("Enable scheduling after register allocation"),
                       cl::init(false), cl::Hidden);
static cl::opt<std::string>
EnableAntiDepBreaking("break-anti-dependencies",
                      cl::desc("Break post-RA scheduling anti-dependencies: "
                               "\"critical\", \"all\", or \"none\""),
                      cl::init("none"), cl::Hidden);

// If DebugDiv > 0 then only schedule MBB with (ID % DebugDiv) == DebugMod
static cl::opt<int>
DebugDiv("postra-sched-debugdiv",
                      cl::desc("Debug control MBBs that are scheduled"),
                      cl::init(0), cl::Hidden);
static cl::opt<int>
DebugMod("postra-sched-debugmod",
                      cl::desc("Debug control MBBs that are scheduled"),
                      cl::init(0), cl::Hidden);

AntiDepBreaker::~AntiDepBreaker() { }

namespace {
  class PostRAScheduler : public MachineFunctionPass {
    const TargetInstrInfo *TII;
    RegisterClassInfo RegClassInfo;

  public:
    static char ID;
    PostRAScheduler() : MachineFunctionPass(ID) {}

    void getAnalysisUsage(AnalysisUsage &AU) const override {
      AU.setPreservesCFG();
      AU.addRequired<AAResultsWrapperPass>();
      AU.addRequired<TargetPassConfig>();
      AU.addRequired<MachineDominatorTree>();
      AU.addPreserved<MachineDominatorTree>();
      AU.addRequired<MachineLoopInfo>();
      AU.addPreserved<MachineLoopInfo>();
      MachineFunctionPass::getAnalysisUsage(AU);
    }

    MachineFunctionProperties getRequiredProperties() const override {
      return MachineFunctionProperties().set(
          MachineFunctionProperties::Property::NoVRegs);
    }

    bool runOnMachineFunction(MachineFunction &Fn) override;

  private:
    bool enablePostRAScheduler(
        const TargetSubtargetInfo &ST, CodeGenOpt::Level OptLevel,
        TargetSubtargetInfo::AntiDepBreakMode &Mode,
        TargetSubtargetInfo::RegClassVector &CriticalPathRCs) const;
  };
  char PostRAScheduler::ID = 0;

  class SchedulePostRATDList : public ScheduleDAGInstrs {
    /// AvailableQueue - The priority queue to use for the available SUnits.
    ///
    LatencyPriorityQueue AvailableQueue;

    /// PendingQueue - This contains all of the instructions whose operands have
    /// been issued, but their results are not ready yet (due to the latency of
    /// the operation).  Once the operands becomes available, the instruction is
    /// added to the AvailableQueue.
    std::vector<SUnit*> PendingQueue;

    /// HazardRec - The hazard recognizer to use.
    ScheduleHazardRecognizer *HazardRec;

    /// AntiDepBreak - Anti-dependence breaking object, or NULL if none
    AntiDepBreaker *AntiDepBreak;

    /// AA - AliasAnalysis for making memory reference queries.
    AliasAnalysis *AA;

    /// The schedule. Null SUnit*'s represent noop instructions.
    std::vector<SUnit*> Sequence;

    /// Ordered list of DAG postprocessing steps.
    std::vector<std::unique_ptr<ScheduleDAGMutation>> Mutations;

    /// The index in BB of RegionEnd.
    ///
    /// This is the instruction number from the top of the current block, not
    /// the SlotIndex. It is only used by the AntiDepBreaker.
    unsigned EndIndex;

  public:
    SchedulePostRATDList(
        MachineFunction &MF, MachineLoopInfo &MLI, AliasAnalysis *AA,
        const RegisterClassInfo &,
        TargetSubtargetInfo::AntiDepBreakMode AntiDepMode,
        SmallVectorImpl<const TargetRegisterClass *> &CriticalPathRCs);

    ~SchedulePostRATDList() override;

    /// startBlock - Initialize register live-range state for scheduling in
    /// this block.
    ///
    void startBlock(MachineBasicBlock *BB) override;

    // Set the index of RegionEnd within the current BB.
    void setEndIndex(unsigned EndIdx) { EndIndex = EndIdx; }

    /// Initialize the scheduler state for the next scheduling region.
    void enterRegion(MachineBasicBlock *bb,
                     MachineBasicBlock::iterator begin,
                     MachineBasicBlock::iterator end,
                     unsigned regioninstrs) override;

    /// Notify that the scheduler has finished scheduling the current region.
    void exitRegion() override;

    /// Schedule - Schedule the instruction range using list scheduling.
    ///
    void schedule() override;

    void EmitSchedule();

    /// Observe - Update liveness information to account for the current
    /// instruction, which will not be scheduled.
    ///
    void Observe(MachineInstr &MI, unsigned Count);

    /// finishBlock - Clean up register live-range state.
    ///
    void finishBlock() override;

  private:
    /// Apply each ScheduleDAGMutation step in order.
    void postprocessDAG();

    void ReleaseSucc(SUnit *SU, SDep *SuccEdge);
    void ReleaseSuccessors(SUnit *SU);
    void ScheduleNodeTopDown(SUnit *SU, unsigned CurCycle);
    void ListScheduleTopDown();

    void dumpSchedule() const;
    void emitNoop(unsigned CurCycle);
  };
}

char &llvm::PostRASchedulerID = PostRAScheduler::ID;

INITIALIZE_PASS(PostRAScheduler, DEBUG_TYPE,
                "Post RA top-down list latency scheduler", false, false)

SchedulePostRATDList::SchedulePostRATDList(
    MachineFunction &MF, MachineLoopInfo &MLI, AliasAnalysis *AA,
    const RegisterClassInfo &RCI,
    TargetSubtargetInfo::AntiDepBreakMode AntiDepMode,
    SmallVectorImpl<const TargetRegisterClass *> &CriticalPathRCs)
    : ScheduleDAGInstrs(MF, &MLI), AA(AA), EndIndex(0) {

  const InstrItineraryData *InstrItins =
      MF.getSubtarget().getInstrItineraryData();
  HazardRec =
      MF.getSubtarget().getInstrInfo()->CreateTargetPostRAHazardRecognizer(
          InstrItins, this);
  MF.getSubtarget().getPostRAMutations(Mutations);

  assert((AntiDepMode == TargetSubtargetInfo::ANTIDEP_NONE ||
          MRI.tracksLiveness()) &&
         "Live-ins must be accurate for anti-dependency breaking");
  AntiDepBreak =
    ((AntiDepMode == TargetSubtargetInfo::ANTIDEP_ALL) ?
     (AntiDepBreaker *)new AggressiveAntiDepBreaker(MF, RCI, CriticalPathRCs) :
     ((AntiDepMode == TargetSubtargetInfo::ANTIDEP_CRITICAL) ?
      (AntiDepBreaker *)new CriticalAntiDepBreaker(MF, RCI) : nullptr));
}

SchedulePostRATDList::~SchedulePostRATDList() {
  delete HazardRec;
  delete AntiDepBreak;
}

/// Initialize state associated with the next scheduling region.
void SchedulePostRATDList::enterRegion(MachineBasicBlock *bb,
                 MachineBasicBlock::iterator begin,
                 MachineBasicBlock::iterator end,
                 unsigned regioninstrs) {
  ScheduleDAGInstrs::enterRegion(bb, begin, end, regioninstrs);
  Sequence.clear();
}

/// Print the schedule before exiting the region.
void SchedulePostRATDList::exitRegion() {
  LLVM_DEBUG({
    dbgs() << "*** Final schedule ***\n";
    dumpSchedule();
    dbgs() << '\n';
  });
  ScheduleDAGInstrs::exitRegion();
}

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
/// dumpSchedule - dump the scheduled Sequence.
LLVM_DUMP_METHOD void SchedulePostRATDList::dumpSchedule() const {
  for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
    if (SUnit *SU = Sequence[i])
      SU->dump(this);
    else
      dbgs() << "**** NOOP ****\n";
  }
}
#endif

bool PostRAScheduler::enablePostRAScheduler(
    const TargetSubtargetInfo &ST,
    CodeGenOpt::Level OptLevel,
    TargetSubtargetInfo::AntiDepBreakMode &Mode,
    TargetSubtargetInfo::RegClassVector &CriticalPathRCs) const {
  Mode = ST.getAntiDepBreakMode();
  ST.getCriticalPathRCs(CriticalPathRCs);

  // Check for explicit enable/disable of post-ra scheduling.
  if (EnablePostRAScheduler.getPosition() > 0)
    return EnablePostRAScheduler;

  return ST.enablePostRAScheduler() &&
         OptLevel >= ST.getOptLevelToEnablePostRAScheduler();
}

bool PostRAScheduler::runOnMachineFunction(MachineFunction &Fn) {
  if (skipFunction(Fn.getFunction()))
    return false;

  TII = Fn.getSubtarget().getInstrInfo();
  MachineLoopInfo &MLI = getAnalysis<MachineLoopInfo>();
  AliasAnalysis *AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();
  TargetPassConfig *PassConfig = &getAnalysis<TargetPassConfig>();

  RegClassInfo.runOnMachineFunction(Fn);

  TargetSubtargetInfo::AntiDepBreakMode AntiDepMode =
    TargetSubtargetInfo::ANTIDEP_NONE;
  SmallVector<const TargetRegisterClass*, 4> CriticalPathRCs;

  // Check that post-RA scheduling is enabled for this target.
  // This may upgrade the AntiDepMode.
  if (!enablePostRAScheduler(Fn.getSubtarget(), PassConfig->getOptLevel(),
                             AntiDepMode, CriticalPathRCs))
    return false;

  // Check for antidep breaking override...
  if (EnableAntiDepBreaking.getPosition() > 0) {
    AntiDepMode = (EnableAntiDepBreaking == "all")
      ? TargetSubtargetInfo::ANTIDEP_ALL
      : ((EnableAntiDepBreaking == "critical")
         ? TargetSubtargetInfo::ANTIDEP_CRITICAL
         : TargetSubtargetInfo::ANTIDEP_NONE);
  }

  LLVM_DEBUG(dbgs() << "PostRAScheduler\n");

  SchedulePostRATDList Scheduler(Fn, MLI, AA, RegClassInfo, AntiDepMode,
                                 CriticalPathRCs);

  // Loop over all of the basic blocks
  for (auto &MBB : Fn) {
#ifndef NDEBUG
    // If DebugDiv > 0 then only schedule MBB with (ID % DebugDiv) == DebugMod
    if (DebugDiv > 0) {
      static int bbcnt = 0;
      if (bbcnt++ % DebugDiv != DebugMod)
        continue;
      dbgs() << "*** DEBUG scheduling " << Fn.getName() << ":"
             << printMBBReference(MBB) << " ***\n";
    }
#endif

    // Initialize register live-range state for scheduling in this block.
    Scheduler.startBlock(&MBB);

    // Schedule each sequence of instructions not interrupted by a label
    // or anything else that effectively needs to shut down scheduling.
    MachineBasicBlock::iterator Current = MBB.end();
    unsigned Count = MBB.size(), CurrentCount = Count;
    for (MachineBasicBlock::iterator I = Current; I != MBB.begin();) {
      MachineInstr &MI = *std::prev(I);
      --Count;
      // Calls are not scheduling boundaries before register allocation, but
      // post-ra we don't gain anything by scheduling across calls since we
      // don't need to worry about register pressure.
      if (MI.isCall() || TII->isSchedulingBoundary(MI, &MBB, Fn)) {
        Scheduler.enterRegion(&MBB, I, Current, CurrentCount - Count);
        Scheduler.setEndIndex(CurrentCount);
        Scheduler.schedule();
        Scheduler.exitRegion();
        Scheduler.EmitSchedule();
        Current = &MI;
        CurrentCount = Count;
        Scheduler.Observe(MI, CurrentCount);
      }
      I = MI;
      if (MI.isBundle())
        Count -= MI.getBundleSize();
    }
    assert(Count == 0 && "Instruction count mismatch!");
    assert((MBB.begin() == Current || CurrentCount != 0) &&
           "Instruction count mismatch!");
    Scheduler.enterRegion(&MBB, MBB.begin(), Current, CurrentCount);
    Scheduler.setEndIndex(CurrentCount);
    Scheduler.schedule();
    Scheduler.exitRegion();
    Scheduler.EmitSchedule();

    // Clean up register live-range state.
    Scheduler.finishBlock();

    // Update register kills
    Scheduler.fixupKills(MBB);
  }

  return true;
}

/// StartBlock - Initialize register live-range state for scheduling in
/// this block.
///
void SchedulePostRATDList::startBlock(MachineBasicBlock *BB) {
  // Call the superclass.
  ScheduleDAGInstrs::startBlock(BB);

  // Reset the hazard recognizer and anti-dep breaker.
  HazardRec->Reset();
  if (AntiDepBreak)
    AntiDepBreak->StartBlock(BB);
}

/// Schedule - Schedule the instruction range using list scheduling.
///
void SchedulePostRATDList::schedule() {
  // Build the scheduling graph.
  buildSchedGraph(AA);

  if (AntiDepBreak) {
    unsigned Broken =
      AntiDepBreak->BreakAntiDependencies(SUnits, RegionBegin, RegionEnd,
                                          EndIndex, DbgValues);

    if (Broken != 0) {
      // We made changes. Update the dependency graph.
      // Theoretically we could update the graph in place:
      // When a live range is changed to use a different register, remove
      // the def's anti-dependence *and* output-dependence edges due to
      // that register, and add new anti-dependence and output-dependence
      // edges based on the next live range of the register.
      ScheduleDAG::clearDAG();
      buildSchedGraph(AA);

      NumFixedAnti += Broken;
    }
  }

  postprocessDAG();

  LLVM_DEBUG(dbgs() << "********** List Scheduling **********\n");
  LLVM_DEBUG(for (const SUnit &SU
                  : SUnits) {
    SU.dumpAll(this);
    dbgs() << '\n';
  });

  AvailableQueue.initNodes(SUnits);
  ListScheduleTopDown();
  AvailableQueue.releaseState();
}

/// Observe - Update liveness information to account for the current
/// instruction, which will not be scheduled.
///
void SchedulePostRATDList::Observe(MachineInstr &MI, unsigned Count) {
  if (AntiDepBreak)
    AntiDepBreak->Observe(MI, Count, EndIndex);
}

/// FinishBlock - Clean up register live-range state.
///
void SchedulePostRATDList::finishBlock() {
  if (AntiDepBreak)
    AntiDepBreak->FinishBlock();

  // Call the superclass.
  ScheduleDAGInstrs::finishBlock();
}

/// Apply each ScheduleDAGMutation step in order.
void SchedulePostRATDList::postprocessDAG() {
  for (auto &M : Mutations)
    M->apply(this);
}

//===----------------------------------------------------------------------===//
//  Top-Down Scheduling
//===----------------------------------------------------------------------===//

/// ReleaseSucc - Decrement the NumPredsLeft count of a successor. Add it to
/// the PendingQueue if the count reaches zero.
void SchedulePostRATDList::ReleaseSucc(SUnit *SU, SDep *SuccEdge) {
  SUnit *SuccSU = SuccEdge->getSUnit();

  if (SuccEdge->isWeak()) {
    --SuccSU->WeakPredsLeft;
    return;
  }
#ifndef NDEBUG
  if (SuccSU->NumPredsLeft == 0) {
    dbgs() << "*** Scheduling failed! ***\n";
    SuccSU->dump(this);
    dbgs() << " has been released too many times!\n";
    llvm_unreachable(nullptr);
  }
#endif
  --SuccSU->NumPredsLeft;

  // Standard scheduler algorithms will recompute the depth of the successor
  // here as such:
  //   SuccSU->setDepthToAtLeast(SU->getDepth() + SuccEdge->getLatency());
  //
  // However, we lazily compute node depth instead. Note that
  // ScheduleNodeTopDown has already updated the depth of this node which causes
  // all descendents to be marked dirty. Setting the successor depth explicitly
  // here would cause depth to be recomputed for all its ancestors. If the
  // successor is not yet ready (because of a transitively redundant edge) then
  // this causes depth computation to be quadratic in the size of the DAG.

  // If all the node's predecessors are scheduled, this node is ready
  // to be scheduled. Ignore the special ExitSU node.
  if (SuccSU->NumPredsLeft == 0 && SuccSU != &ExitSU)
    PendingQueue.push_back(SuccSU);
}

/// ReleaseSuccessors - Call ReleaseSucc on each of SU's successors.
void SchedulePostRATDList::ReleaseSuccessors(SUnit *SU) {
  for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
       I != E; ++I) {
    ReleaseSucc(SU, &*I);
  }
}

/// ScheduleNodeTopDown - Add the node to the schedule. Decrement the pending
/// count of its successors. If a successor pending count is zero, add it to
/// the Available queue.
void SchedulePostRATDList::ScheduleNodeTopDown(SUnit *SU, unsigned CurCycle) {
  LLVM_DEBUG(dbgs() << "*** Scheduling [" << CurCycle << "]: ");
  LLVM_DEBUG(SU->dump(this));

  Sequence.push_back(SU);
  assert(CurCycle >= SU->getDepth() &&
         "Node scheduled above its depth!");
  SU->setDepthToAtLeast(CurCycle);

  ReleaseSuccessors(SU);
  SU->isScheduled = true;
  AvailableQueue.scheduledNode(SU);
}

/// emitNoop - Add a noop to the current instruction sequence.
void SchedulePostRATDList::emitNoop(unsigned CurCycle) {
  LLVM_DEBUG(dbgs() << "*** Emitting noop in cycle " << CurCycle << '\n');
  HazardRec->EmitNoop();
  Sequence.push_back(nullptr);   // NULL here means noop
  ++NumNoops;
}

/// ListScheduleTopDown - The main loop of list scheduling for top-down
/// schedulers.
void SchedulePostRATDList::ListScheduleTopDown() {
  unsigned CurCycle = 0;

  // We're scheduling top-down but we're visiting the regions in
  // bottom-up order, so we don't know the hazards at the start of a
  // region. So assume no hazards (this should usually be ok as most
  // blocks are a single region).
  HazardRec->Reset();

  // Release any successors of the special Entry node.
  ReleaseSuccessors(&EntrySU);

  // Add all leaves to Available queue.
  for (unsigned i = 0, e = SUnits.size(); i != e; ++i) {
    // It is available if it has no predecessors.
    if (!SUnits[i].NumPredsLeft && !SUnits[i].isAvailable) {
      AvailableQueue.push(&SUnits[i]);
      SUnits[i].isAvailable = true;
    }
  }

  // In any cycle where we can't schedule any instructions, we must
  // stall or emit a noop, depending on the target.
  bool CycleHasInsts = false;

  // While Available queue is not empty, grab the node with the highest
  // priority. If it is not ready put it back.  Schedule the node.
  std::vector<SUnit*> NotReady;
  Sequence.reserve(SUnits.size());
  while (!AvailableQueue.empty() || !PendingQueue.empty()) {
    // Check to see if any of the pending instructions are ready to issue.  If
    // so, add them to the available queue.
    unsigned MinDepth = ~0u;
    for (unsigned i = 0, e = PendingQueue.size(); i != e; ++i) {
      if (PendingQueue[i]->getDepth() <= CurCycle) {
        AvailableQueue.push(PendingQueue[i]);
        PendingQueue[i]->isAvailable = true;
        PendingQueue[i] = PendingQueue.back();
        PendingQueue.pop_back();
        --i; --e;
      } else if (PendingQueue[i]->getDepth() < MinDepth)
        MinDepth = PendingQueue[i]->getDepth();
    }

    LLVM_DEBUG(dbgs() << "\n*** Examining Available\n";
               AvailableQueue.dump(this));

    SUnit *FoundSUnit = nullptr, *NotPreferredSUnit = nullptr;
    bool HasNoopHazards = false;
    while (!AvailableQueue.empty()) {
      SUnit *CurSUnit = AvailableQueue.pop();

      ScheduleHazardRecognizer::HazardType HT =
        HazardRec->getHazardType(CurSUnit, 0/*no stalls*/);
      if (HT == ScheduleHazardRecognizer::NoHazard) {
        if (HazardRec->ShouldPreferAnother(CurSUnit)) {
          if (!NotPreferredSUnit) {
            // If this is the first non-preferred node for this cycle, then
            // record it and continue searching for a preferred node. If this
            // is not the first non-preferred node, then treat it as though
            // there had been a hazard.
            NotPreferredSUnit = CurSUnit;
            continue;
          }
        } else {
          FoundSUnit = CurSUnit;
          break;
        }
      }

      // Remember if this is a noop hazard.
      HasNoopHazards |= HT == ScheduleHazardRecognizer::NoopHazard;

      NotReady.push_back(CurSUnit);
    }

    // If we have a non-preferred node, push it back onto the available list.
    // If we did not find a preferred node, then schedule this first
    // non-preferred node.
    if (NotPreferredSUnit) {
      if (!FoundSUnit) {
        LLVM_DEBUG(
            dbgs() << "*** Will schedule a non-preferred instruction...\n");
        FoundSUnit = NotPreferredSUnit;
      } else {
        AvailableQueue.push(NotPreferredSUnit);
      }

      NotPreferredSUnit = nullptr;
    }

    // Add the nodes that aren't ready back onto the available list.
    if (!NotReady.empty()) {
      AvailableQueue.push_all(NotReady);
      NotReady.clear();
    }

    // If we found a node to schedule...
    if (FoundSUnit) {
      // If we need to emit noops prior to this instruction, then do so.
      unsigned NumPreNoops = HazardRec->PreEmitNoops(FoundSUnit);
      for (unsigned i = 0; i != NumPreNoops; ++i)
        emitNoop(CurCycle);

      // ... schedule the node...
      ScheduleNodeTopDown(FoundSUnit, CurCycle);
      HazardRec->EmitInstruction(FoundSUnit);
      CycleHasInsts = true;
      if (HazardRec->atIssueLimit()) {
        LLVM_DEBUG(dbgs() << "*** Max instructions per cycle " << CurCycle
                          << '\n');
        HazardRec->AdvanceCycle();
        ++CurCycle;
        CycleHasInsts = false;
      }
    } else {
      if (CycleHasInsts) {
        LLVM_DEBUG(dbgs() << "*** Finished cycle " << CurCycle << '\n');
        HazardRec->AdvanceCycle();
      } else if (!HasNoopHazards) {
        // Otherwise, we have a pipeline stall, but no other problem,
        // just advance the current cycle and try again.
        LLVM_DEBUG(dbgs() << "*** Stall in cycle " << CurCycle << '\n');
        HazardRec->AdvanceCycle();
        ++NumStalls;
      } else {
        // Otherwise, we have no instructions to issue and we have instructions
        // that will fault if we don't do this right.  This is the case for
        // processors without pipeline interlocks and other cases.
        emitNoop(CurCycle);
      }

      ++CurCycle;
      CycleHasInsts = false;
    }
  }

#ifndef NDEBUG
  unsigned ScheduledNodes = VerifyScheduledDAG(/*isBottomUp=*/false);
  unsigned Noops = 0;
  for (unsigned i = 0, e = Sequence.size(); i != e; ++i)
    if (!Sequence[i])
      ++Noops;
  assert(Sequence.size() - Noops == ScheduledNodes &&
         "The number of nodes scheduled doesn't match the expected number!");
#endif // NDEBUG
}

// EmitSchedule - Emit the machine code in scheduled order.
void SchedulePostRATDList::EmitSchedule() {
  RegionBegin = RegionEnd;

  // If first instruction was a DBG_VALUE then put it back.
  if (FirstDbgValue)
    BB->splice(RegionEnd, BB, FirstDbgValue);

  // Then re-insert them according to the given schedule.
  for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
    if (SUnit *SU = Sequence[i])
      BB->splice(RegionEnd, BB, SU->getInstr());
    else
      // Null SUnit* is a noop.
      TII->insertNoop(*BB, RegionEnd);

    // Update the Begin iterator, as the first instruction in the block
    // may have been scheduled later.
    if (i == 0)
      RegionBegin = std::prev(RegionEnd);
  }

  // Reinsert any remaining debug_values.
  for (std::vector<std::pair<MachineInstr *, MachineInstr *> >::iterator
         DI = DbgValues.end(), DE = DbgValues.begin(); DI != DE; --DI) {
    std::pair<MachineInstr *, MachineInstr *> P = *std::prev(DI);
    MachineInstr *DbgValue = P.first;
    MachineBasicBlock::iterator OrigPrivMI = P.second;
    BB->splice(++OrigPrivMI, BB, DbgValue);
  }
  DbgValues.clear();
  FirstDbgValue = nullptr;
}