1 //===- MachineScheduler.cpp - Machine Instruction Scheduler ---------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // MachineScheduler schedules machine instructions after phi elimination. It
11 // preserves LiveIntervals so it can be invoked before register allocation.
13 //===----------------------------------------------------------------------===//
15 #define DEBUG_TYPE "misched"
17 #include "llvm/CodeGen/LiveIntervalAnalysis.h"
18 #include "llvm/CodeGen/MachineScheduler.h"
19 #include "llvm/CodeGen/Passes.h"
20 #include "llvm/CodeGen/RegisterClassInfo.h"
21 #include "llvm/CodeGen/ScheduleDFS.h"
22 #include "llvm/CodeGen/ScheduleHazardRecognizer.h"
23 #include "llvm/Analysis/AliasAnalysis.h"
24 #include "llvm/Support/CommandLine.h"
25 #include "llvm/Support/Debug.h"
26 #include "llvm/Support/ErrorHandling.h"
27 #include "llvm/Support/raw_ostream.h"
28 #include "llvm/ADT/OwningPtr.h"
29 #include "llvm/ADT/PriorityQueue.h"
36 cl::opt<bool> ForceTopDown("misched-topdown", cl::Hidden,
37 cl::desc("Force top-down list scheduling"));
38 cl::opt<bool> ForceBottomUp("misched-bottomup", cl::Hidden,
39 cl::desc("Force bottom-up list scheduling"));
43 static cl::opt<bool> ViewMISchedDAGs("view-misched-dags", cl::Hidden,
44 cl::desc("Pop up a window to show MISched dags after they are processed"));
46 static cl::opt<unsigned> MISchedCutoff("misched-cutoff", cl::Hidden,
47 cl::desc("Stop scheduling after N instructions"), cl::init(~0U));
49 static bool ViewMISchedDAGs = false;
52 // Threshold to very roughly model an out-of-order processor's instruction
53 // buffers. If the actual value of this threshold matters much in practice, then
54 // it can be specified by the machine model. For now, it's an experimental
55 // tuning knob to determine when and if it matters.
56 static cl::opt<unsigned> ILPWindow("ilp-window", cl::Hidden,
57 cl::desc("Allow expected latency to exceed the critical path by N cycles "
58 "before attempting to balance ILP"),
61 // Experimental heuristics
62 static cl::opt<bool> EnableLoadCluster("misched-cluster", cl::Hidden,
63 cl::desc("Enable load clustering."), cl::init(true));
65 // Experimental heuristics
66 static cl::opt<bool> EnableMacroFusion("misched-fusion", cl::Hidden,
67 cl::desc("Enable scheduling for macro fusion."), cl::init(true));
69 //===----------------------------------------------------------------------===//
70 // Machine Instruction Scheduling Pass and Registry
71 //===----------------------------------------------------------------------===//
73 MachineSchedContext::MachineSchedContext():
74 MF(0), MLI(0), MDT(0), PassConfig(0), AA(0), LIS(0) {
75 RegClassInfo = new RegisterClassInfo();
78 MachineSchedContext::~MachineSchedContext() {
83 /// MachineScheduler runs after coalescing and before register allocation.
84 class MachineScheduler : public MachineSchedContext,
85 public MachineFunctionPass {
89 virtual void getAnalysisUsage(AnalysisUsage &AU) const;
91 virtual void releaseMemory() {}
93 virtual bool runOnMachineFunction(MachineFunction&);
95 virtual void print(raw_ostream &O, const Module* = 0) const;
97 static char ID; // Class identification, replacement for typeinfo
101 char MachineScheduler::ID = 0;
103 char &llvm::MachineSchedulerID = MachineScheduler::ID;
105 INITIALIZE_PASS_BEGIN(MachineScheduler, "misched",
106 "Machine Instruction Scheduler", false, false)
107 INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
108 INITIALIZE_PASS_DEPENDENCY(SlotIndexes)
109 INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
110 INITIALIZE_PASS_END(MachineScheduler, "misched",
111 "Machine Instruction Scheduler", false, false)
113 MachineScheduler::MachineScheduler()
114 : MachineFunctionPass(ID) {
115 initializeMachineSchedulerPass(*PassRegistry::getPassRegistry());
118 void MachineScheduler::getAnalysisUsage(AnalysisUsage &AU) const {
119 AU.setPreservesCFG();
120 AU.addRequiredID(MachineDominatorsID);
121 AU.addRequired<MachineLoopInfo>();
122 AU.addRequired<AliasAnalysis>();
123 AU.addRequired<TargetPassConfig>();
124 AU.addRequired<SlotIndexes>();
125 AU.addPreserved<SlotIndexes>();
126 AU.addRequired<LiveIntervals>();
127 AU.addPreserved<LiveIntervals>();
128 MachineFunctionPass::getAnalysisUsage(AU);
131 MachinePassRegistry MachineSchedRegistry::Registry;
133 /// A dummy default scheduler factory indicates whether the scheduler
134 /// is overridden on the command line.
135 static ScheduleDAGInstrs *useDefaultMachineSched(MachineSchedContext *C) {
139 /// MachineSchedOpt allows command line selection of the scheduler.
140 static cl::opt<MachineSchedRegistry::ScheduleDAGCtor, false,
141 RegisterPassParser<MachineSchedRegistry> >
142 MachineSchedOpt("misched",
143 cl::init(&useDefaultMachineSched), cl::Hidden,
144 cl::desc("Machine instruction scheduler to use"));
146 static MachineSchedRegistry
147 DefaultSchedRegistry("default", "Use the target's default scheduler choice.",
148 useDefaultMachineSched);
150 /// Forward declare the standard machine scheduler. This will be used as the
151 /// default scheduler if the target does not set a default.
152 static ScheduleDAGInstrs *createConvergingSched(MachineSchedContext *C);
155 /// Decrement this iterator until reaching the top or a non-debug instr.
156 static MachineBasicBlock::iterator
157 priorNonDebug(MachineBasicBlock::iterator I, MachineBasicBlock::iterator Beg) {
158 assert(I != Beg && "reached the top of the region, cannot decrement");
160 if (!I->isDebugValue())
166 /// If this iterator is a debug value, increment until reaching the End or a
167 /// non-debug instruction.
168 static MachineBasicBlock::iterator
169 nextIfDebug(MachineBasicBlock::iterator I, MachineBasicBlock::iterator End) {
170 for(; I != End; ++I) {
171 if (!I->isDebugValue())
177 /// Top-level MachineScheduler pass driver.
179 /// Visit blocks in function order. Divide each block into scheduling regions
180 /// and visit them bottom-up. Visiting regions bottom-up is not required, but is
181 /// consistent with the DAG builder, which traverses the interior of the
182 /// scheduling regions bottom-up.
184 /// This design avoids exposing scheduling boundaries to the DAG builder,
185 /// simplifying the DAG builder's support for "special" target instructions.
186 /// At the same time the design allows target schedulers to operate across
187 /// scheduling boundaries, for example to bundle the boudary instructions
188 /// without reordering them. This creates complexity, because the target
189 /// scheduler must update the RegionBegin and RegionEnd positions cached by
190 /// ScheduleDAGInstrs whenever adding or removing instructions. A much simpler
191 /// design would be to split blocks at scheduling boundaries, but LLVM has a
192 /// general bias against block splitting purely for implementation simplicity.
193 bool MachineScheduler::runOnMachineFunction(MachineFunction &mf) {
194 DEBUG(dbgs() << "Before MISsched:\n"; mf.print(dbgs()));
196 // Initialize the context of the pass.
198 MLI = &getAnalysis<MachineLoopInfo>();
199 MDT = &getAnalysis<MachineDominatorTree>();
200 PassConfig = &getAnalysis<TargetPassConfig>();
201 AA = &getAnalysis<AliasAnalysis>();
203 LIS = &getAnalysis<LiveIntervals>();
204 const TargetInstrInfo *TII = MF->getTarget().getInstrInfo();
206 RegClassInfo->runOnMachineFunction(*MF);
208 // Select the scheduler, or set the default.
209 MachineSchedRegistry::ScheduleDAGCtor Ctor = MachineSchedOpt;
210 if (Ctor == useDefaultMachineSched) {
211 // Get the default scheduler set by the target.
212 Ctor = MachineSchedRegistry::getDefault();
214 Ctor = createConvergingSched;
215 MachineSchedRegistry::setDefault(Ctor);
218 // Instantiate the selected scheduler.
219 OwningPtr<ScheduleDAGInstrs> Scheduler(Ctor(this));
221 // Visit all machine basic blocks.
223 // TODO: Visit blocks in global postorder or postorder within the bottom-up
224 // loop tree. Then we can optionally compute global RegPressure.
225 for (MachineFunction::iterator MBB = MF->begin(), MBBEnd = MF->end();
226 MBB != MBBEnd; ++MBB) {
228 Scheduler->startBlock(MBB);
230 // Break the block into scheduling regions [I, RegionEnd), and schedule each
231 // region as soon as it is discovered. RegionEnd points the scheduling
232 // boundary at the bottom of the region. The DAG does not include RegionEnd,
233 // but the region does (i.e. the next RegionEnd is above the previous
234 // RegionBegin). If the current block has no terminator then RegionEnd ==
235 // MBB->end() for the bottom region.
237 // The Scheduler may insert instructions during either schedule() or
238 // exitRegion(), even for empty regions. So the local iterators 'I' and
239 // 'RegionEnd' are invalid across these calls.
240 unsigned RemainingInstrs = MBB->size();
241 for(MachineBasicBlock::iterator RegionEnd = MBB->end();
242 RegionEnd != MBB->begin(); RegionEnd = Scheduler->begin()) {
244 // Avoid decrementing RegionEnd for blocks with no terminator.
245 if (RegionEnd != MBB->end()
246 || TII->isSchedulingBoundary(llvm::prior(RegionEnd), MBB, *MF)) {
248 // Count the boundary instruction.
252 // The next region starts above the previous region. Look backward in the
253 // instruction stream until we find the nearest boundary.
254 MachineBasicBlock::iterator I = RegionEnd;
255 for(;I != MBB->begin(); --I, --RemainingInstrs) {
256 if (TII->isSchedulingBoundary(llvm::prior(I), MBB, *MF))
259 // Notify the scheduler of the region, even if we may skip scheduling
260 // it. Perhaps it still needs to be bundled.
261 Scheduler->enterRegion(MBB, I, RegionEnd, RemainingInstrs);
263 // Skip empty scheduling regions (0 or 1 schedulable instructions).
264 if (I == RegionEnd || I == llvm::prior(RegionEnd)) {
265 // Close the current region. Bundle the terminator if needed.
266 // This invalidates 'RegionEnd' and 'I'.
267 Scheduler->exitRegion();
270 DEBUG(dbgs() << "********** MI Scheduling **********\n");
271 DEBUG(dbgs() << MF->getName()
272 << ":BB#" << MBB->getNumber() << "\n From: " << *I << " To: ";
273 if (RegionEnd != MBB->end()) dbgs() << *RegionEnd;
274 else dbgs() << "End";
275 dbgs() << " Remaining: " << RemainingInstrs << "\n");
277 // Schedule a region: possibly reorder instructions.
278 // This invalidates 'RegionEnd' and 'I'.
279 Scheduler->schedule();
281 // Close the current region.
282 Scheduler->exitRegion();
284 // Scheduling has invalidated the current iterator 'I'. Ask the
285 // scheduler for the top of it's scheduled region.
286 RegionEnd = Scheduler->begin();
288 assert(RemainingInstrs == 0 && "Instruction count mismatch!");
289 Scheduler->finishBlock();
291 Scheduler->finalizeSchedule();
292 DEBUG(LIS->print(dbgs()));
296 void MachineScheduler::print(raw_ostream &O, const Module* m) const {
300 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
301 void ReadyQueue::dump() {
302 dbgs() << Name << ": ";
303 for (unsigned i = 0, e = Queue.size(); i < e; ++i)
304 dbgs() << Queue[i]->NodeNum << " ";
309 //===----------------------------------------------------------------------===//
310 // ScheduleDAGMI - Base class for MachineInstr scheduling with LiveIntervals
312 //===----------------------------------------------------------------------===//
314 bool ScheduleDAGMI::addEdge(SUnit *SuccSU, const SDep &PredDep) {
315 if (SuccSU != &ExitSU) {
316 // Do not use WillCreateCycle, it assumes SD scheduling.
317 // If Pred is reachable from Succ, then the edge creates a cycle.
318 if (Topo.IsReachable(PredDep.getSUnit(), SuccSU))
320 Topo.AddPred(SuccSU, PredDep.getSUnit());
322 SuccSU->addPred(PredDep, /*Required=*/!PredDep.isArtificial());
323 // Return true regardless of whether a new edge needed to be inserted.
327 /// ReleaseSucc - Decrement the NumPredsLeft count of a successor. When
328 /// NumPredsLeft reaches zero, release the successor node.
330 /// FIXME: Adjust SuccSU height based on MinLatency.
331 void ScheduleDAGMI::releaseSucc(SUnit *SU, SDep *SuccEdge) {
332 SUnit *SuccSU = SuccEdge->getSUnit();
334 if (SuccEdge->isWeak()) {
335 --SuccSU->WeakPredsLeft;
336 if (SuccEdge->isCluster())
337 NextClusterSucc = SuccSU;
341 if (SuccSU->NumPredsLeft == 0) {
342 dbgs() << "*** Scheduling failed! ***\n";
344 dbgs() << " has been released too many times!\n";
348 --SuccSU->NumPredsLeft;
349 if (SuccSU->NumPredsLeft == 0 && SuccSU != &ExitSU)
350 SchedImpl->releaseTopNode(SuccSU);
353 /// releaseSuccessors - Call releaseSucc on each of SU's successors.
354 void ScheduleDAGMI::releaseSuccessors(SUnit *SU) {
355 for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
357 releaseSucc(SU, &*I);
361 /// ReleasePred - Decrement the NumSuccsLeft count of a predecessor. When
362 /// NumSuccsLeft reaches zero, release the predecessor node.
364 /// FIXME: Adjust PredSU height based on MinLatency.
365 void ScheduleDAGMI::releasePred(SUnit *SU, SDep *PredEdge) {
366 SUnit *PredSU = PredEdge->getSUnit();
368 if (PredEdge->isWeak()) {
369 --PredSU->WeakSuccsLeft;
370 if (PredEdge->isCluster())
371 NextClusterPred = PredSU;
375 if (PredSU->NumSuccsLeft == 0) {
376 dbgs() << "*** Scheduling failed! ***\n";
378 dbgs() << " has been released too many times!\n";
382 --PredSU->NumSuccsLeft;
383 if (PredSU->NumSuccsLeft == 0 && PredSU != &EntrySU)
384 SchedImpl->releaseBottomNode(PredSU);
387 /// releasePredecessors - Call releasePred on each of SU's predecessors.
388 void ScheduleDAGMI::releasePredecessors(SUnit *SU) {
389 for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
391 releasePred(SU, &*I);
395 void ScheduleDAGMI::moveInstruction(MachineInstr *MI,
396 MachineBasicBlock::iterator InsertPos) {
397 // Advance RegionBegin if the first instruction moves down.
398 if (&*RegionBegin == MI)
401 // Update the instruction stream.
402 BB->splice(InsertPos, BB, MI);
404 // Update LiveIntervals
405 LIS->handleMove(MI, /*UpdateFlags=*/true);
407 // Recede RegionBegin if an instruction moves above the first.
408 if (RegionBegin == InsertPos)
412 bool ScheduleDAGMI::checkSchedLimit() {
414 if (NumInstrsScheduled == MISchedCutoff && MISchedCutoff != ~0U) {
415 CurrentTop = CurrentBottom;
418 ++NumInstrsScheduled;
423 /// enterRegion - Called back from MachineScheduler::runOnMachineFunction after
424 /// crossing a scheduling boundary. [begin, end) includes all instructions in
425 /// the region, including the boundary itself and single-instruction regions
426 /// that don't get scheduled.
427 void ScheduleDAGMI::enterRegion(MachineBasicBlock *bb,
428 MachineBasicBlock::iterator begin,
429 MachineBasicBlock::iterator end,
432 ScheduleDAGInstrs::enterRegion(bb, begin, end, endcount);
434 // For convenience remember the end of the liveness region.
436 (RegionEnd == bb->end()) ? RegionEnd : llvm::next(RegionEnd);
439 // Setup the register pressure trackers for the top scheduled top and bottom
440 // scheduled regions.
441 void ScheduleDAGMI::initRegPressure() {
442 TopRPTracker.init(&MF, RegClassInfo, LIS, BB, RegionBegin);
443 BotRPTracker.init(&MF, RegClassInfo, LIS, BB, LiveRegionEnd);
445 // Close the RPTracker to finalize live ins.
446 RPTracker.closeRegion();
448 DEBUG(RPTracker.getPressure().dump(TRI));
450 // Initialize the live ins and live outs.
451 TopRPTracker.addLiveRegs(RPTracker.getPressure().LiveInRegs);
452 BotRPTracker.addLiveRegs(RPTracker.getPressure().LiveOutRegs);
454 // Close one end of the tracker so we can call
455 // getMaxUpward/DownwardPressureDelta before advancing across any
456 // instructions. This converts currently live regs into live ins/outs.
457 TopRPTracker.closeTop();
458 BotRPTracker.closeBottom();
460 // Account for liveness generated by the region boundary.
461 if (LiveRegionEnd != RegionEnd)
462 BotRPTracker.recede();
464 assert(BotRPTracker.getPos() == RegionEnd && "Can't find the region bottom");
466 // Cache the list of excess pressure sets in this region. This will also track
467 // the max pressure in the scheduled code for these sets.
468 RegionCriticalPSets.clear();
469 std::vector<unsigned> RegionPressure = RPTracker.getPressure().MaxSetPressure;
470 for (unsigned i = 0, e = RegionPressure.size(); i < e; ++i) {
471 unsigned Limit = TRI->getRegPressureSetLimit(i);
472 DEBUG(dbgs() << TRI->getRegPressureSetName(i)
474 << " Actual " << RegionPressure[i] << "\n");
475 if (RegionPressure[i] > Limit)
476 RegionCriticalPSets.push_back(PressureElement(i, 0));
478 DEBUG(dbgs() << "Excess PSets: ";
479 for (unsigned i = 0, e = RegionCriticalPSets.size(); i != e; ++i)
480 dbgs() << TRI->getRegPressureSetName(
481 RegionCriticalPSets[i].PSetID) << " ";
485 // FIXME: When the pressure tracker deals in pressure differences then we won't
486 // iterate over all RegionCriticalPSets[i].
488 updateScheduledPressure(std::vector<unsigned> NewMaxPressure) {
489 for (unsigned i = 0, e = RegionCriticalPSets.size(); i < e; ++i) {
490 unsigned ID = RegionCriticalPSets[i].PSetID;
491 int &MaxUnits = RegionCriticalPSets[i].UnitIncrease;
492 if ((int)NewMaxPressure[ID] > MaxUnits)
493 MaxUnits = NewMaxPressure[ID];
497 /// schedule - Called back from MachineScheduler::runOnMachineFunction
498 /// after setting up the current scheduling region. [RegionBegin, RegionEnd)
499 /// only includes instructions that have DAG nodes, not scheduling boundaries.
501 /// This is a skeletal driver, with all the functionality pushed into helpers,
502 /// so that it can be easilly extended by experimental schedulers. Generally,
503 /// implementing MachineSchedStrategy should be sufficient to implement a new
504 /// scheduling algorithm. However, if a scheduler further subclasses
505 /// ScheduleDAGMI then it will want to override this virtual method in order to
506 /// update any specialized state.
507 void ScheduleDAGMI::schedule() {
508 buildDAGWithRegPressure();
510 Topo.InitDAGTopologicalSorting();
514 DEBUG(for (unsigned su = 0, e = SUnits.size(); su != e; ++su)
515 SUnits[su].dumpAll(this));
517 if (ViewMISchedDAGs) viewGraph();
521 bool IsTopNode = false;
522 while (SUnit *SU = SchedImpl->pickNode(IsTopNode)) {
523 assert(!SU->isScheduled && "Node already scheduled");
524 if (!checkSchedLimit())
527 scheduleMI(SU, IsTopNode);
529 updateQueues(SU, IsTopNode);
531 assert(CurrentTop == CurrentBottom && "Nonempty unscheduled zone.");
536 unsigned BBNum = begin()->getParent()->getNumber();
537 dbgs() << "*** Final schedule for BB#" << BBNum << " ***\n";
543 /// Build the DAG and setup three register pressure trackers.
544 void ScheduleDAGMI::buildDAGWithRegPressure() {
545 // Initialize the register pressure tracker used by buildSchedGraph.
546 RPTracker.init(&MF, RegClassInfo, LIS, BB, LiveRegionEnd);
548 // Account for liveness generate by the region boundary.
549 if (LiveRegionEnd != RegionEnd)
552 // Build the DAG, and compute current register pressure.
553 buildSchedGraph(AA, &RPTracker);
554 if (ViewMISchedDAGs) viewGraph();
556 // Initialize top/bottom trackers after computing region pressure.
560 /// Apply each ScheduleDAGMutation step in order.
561 void ScheduleDAGMI::postprocessDAG() {
562 for (unsigned i = 0, e = Mutations.size(); i < e; ++i) {
563 Mutations[i]->apply(this);
567 // Release all DAG roots for scheduling.
569 // Nodes with unreleased weak edges can still be roots.
570 void ScheduleDAGMI::releaseRoots() {
571 SmallVector<SUnit*, 16> BotRoots;
573 for (std::vector<SUnit>::iterator
574 I = SUnits.begin(), E = SUnits.end(); I != E; ++I) {
576 // A SUnit is ready to top schedule if it has no predecessors.
577 if (!I->NumPredsLeft && SU != &EntrySU)
578 SchedImpl->releaseTopNode(SU);
579 // A SUnit is ready to bottom schedule if it has no successors.
580 if (!I->NumSuccsLeft && SU != &ExitSU)
581 BotRoots.push_back(SU);
583 // Release bottom roots in reverse order so the higher priority nodes appear
584 // first. This is more natural and slightly more efficient.
585 for (SmallVectorImpl<SUnit*>::const_reverse_iterator
586 I = BotRoots.rbegin(), E = BotRoots.rend(); I != E; ++I)
587 SchedImpl->releaseBottomNode(*I);
590 /// Identify DAG roots and setup scheduler queues.
591 void ScheduleDAGMI::initQueues() {
592 NextClusterSucc = NULL;
593 NextClusterPred = NULL;
595 // Initialize the strategy before modifying the DAG.
596 SchedImpl->initialize(this);
598 // Release all DAG roots for scheduling, not including EntrySU/ExitSU.
601 releaseSuccessors(&EntrySU);
602 releasePredecessors(&ExitSU);
604 SchedImpl->registerRoots();
606 // Advance past initial DebugValues.
607 assert(TopRPTracker.getPos() == RegionBegin && "bad initial Top tracker");
608 CurrentTop = nextIfDebug(RegionBegin, RegionEnd);
609 TopRPTracker.setPos(CurrentTop);
611 CurrentBottom = RegionEnd;
614 /// Move an instruction and update register pressure.
615 void ScheduleDAGMI::scheduleMI(SUnit *SU, bool IsTopNode) {
616 // Move the instruction to its new location in the instruction stream.
617 MachineInstr *MI = SU->getInstr();
620 assert(SU->isTopReady() && "node still has unscheduled dependencies");
621 if (&*CurrentTop == MI)
622 CurrentTop = nextIfDebug(++CurrentTop, CurrentBottom);
624 moveInstruction(MI, CurrentTop);
625 TopRPTracker.setPos(MI);
628 // Update top scheduled pressure.
629 TopRPTracker.advance();
630 assert(TopRPTracker.getPos() == CurrentTop && "out of sync");
631 updateScheduledPressure(TopRPTracker.getPressure().MaxSetPressure);
634 assert(SU->isBottomReady() && "node still has unscheduled dependencies");
635 MachineBasicBlock::iterator priorII =
636 priorNonDebug(CurrentBottom, CurrentTop);
638 CurrentBottom = priorII;
640 if (&*CurrentTop == MI) {
641 CurrentTop = nextIfDebug(++CurrentTop, priorII);
642 TopRPTracker.setPos(CurrentTop);
644 moveInstruction(MI, CurrentBottom);
647 // Update bottom scheduled pressure.
648 BotRPTracker.recede();
649 assert(BotRPTracker.getPos() == CurrentBottom && "out of sync");
650 updateScheduledPressure(BotRPTracker.getPressure().MaxSetPressure);
654 /// Update scheduler queues after scheduling an instruction.
655 void ScheduleDAGMI::updateQueues(SUnit *SU, bool IsTopNode) {
656 // Release dependent instructions for scheduling.
658 releaseSuccessors(SU);
660 releasePredecessors(SU);
662 SU->isScheduled = true;
664 // Notify the scheduling strategy after updating the DAG.
665 SchedImpl->schedNode(SU, IsTopNode);
668 /// Reinsert any remaining debug_values, just like the PostRA scheduler.
669 void ScheduleDAGMI::placeDebugValues() {
670 // If first instruction was a DBG_VALUE then put it back.
672 BB->splice(RegionBegin, BB, FirstDbgValue);
673 RegionBegin = FirstDbgValue;
676 for (std::vector<std::pair<MachineInstr *, MachineInstr *> >::iterator
677 DI = DbgValues.end(), DE = DbgValues.begin(); DI != DE; --DI) {
678 std::pair<MachineInstr *, MachineInstr *> P = *prior(DI);
679 MachineInstr *DbgValue = P.first;
680 MachineBasicBlock::iterator OrigPrevMI = P.second;
681 if (&*RegionBegin == DbgValue)
683 BB->splice(++OrigPrevMI, BB, DbgValue);
684 if (OrigPrevMI == llvm::prior(RegionEnd))
685 RegionEnd = DbgValue;
688 FirstDbgValue = NULL;
691 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
692 void ScheduleDAGMI::dumpSchedule() const {
693 for (MachineBasicBlock::iterator MI = begin(), ME = end(); MI != ME; ++MI) {
694 if (SUnit *SU = getSUnit(&(*MI)))
697 dbgs() << "Missing SUnit\n";
702 //===----------------------------------------------------------------------===//
703 // LoadClusterMutation - DAG post-processing to cluster loads.
704 //===----------------------------------------------------------------------===//
707 /// \brief Post-process the DAG to create cluster edges between neighboring
709 class LoadClusterMutation : public ScheduleDAGMutation {
714 LoadInfo(SUnit *su, unsigned reg, unsigned ofs)
715 : SU(su), BaseReg(reg), Offset(ofs) {}
717 static bool LoadInfoLess(const LoadClusterMutation::LoadInfo &LHS,
718 const LoadClusterMutation::LoadInfo &RHS);
720 const TargetInstrInfo *TII;
721 const TargetRegisterInfo *TRI;
723 LoadClusterMutation(const TargetInstrInfo *tii,
724 const TargetRegisterInfo *tri)
725 : TII(tii), TRI(tri) {}
727 virtual void apply(ScheduleDAGMI *DAG);
729 void clusterNeighboringLoads(ArrayRef<SUnit*> Loads, ScheduleDAGMI *DAG);
733 bool LoadClusterMutation::LoadInfoLess(
734 const LoadClusterMutation::LoadInfo &LHS,
735 const LoadClusterMutation::LoadInfo &RHS) {
736 if (LHS.BaseReg != RHS.BaseReg)
737 return LHS.BaseReg < RHS.BaseReg;
738 return LHS.Offset < RHS.Offset;
741 void LoadClusterMutation::clusterNeighboringLoads(ArrayRef<SUnit*> Loads,
742 ScheduleDAGMI *DAG) {
743 SmallVector<LoadClusterMutation::LoadInfo,32> LoadRecords;
744 for (unsigned Idx = 0, End = Loads.size(); Idx != End; ++Idx) {
745 SUnit *SU = Loads[Idx];
748 if (TII->getLdStBaseRegImmOfs(SU->getInstr(), BaseReg, Offset, TRI))
749 LoadRecords.push_back(LoadInfo(SU, BaseReg, Offset));
751 if (LoadRecords.size() < 2)
753 std::sort(LoadRecords.begin(), LoadRecords.end(), LoadInfoLess);
754 unsigned ClusterLength = 1;
755 for (unsigned Idx = 0, End = LoadRecords.size(); Idx < (End - 1); ++Idx) {
756 if (LoadRecords[Idx].BaseReg != LoadRecords[Idx+1].BaseReg) {
761 SUnit *SUa = LoadRecords[Idx].SU;
762 SUnit *SUb = LoadRecords[Idx+1].SU;
763 if (TII->shouldClusterLoads(SUa->getInstr(), SUb->getInstr(), ClusterLength)
764 && DAG->addEdge(SUb, SDep(SUa, SDep::Cluster))) {
766 DEBUG(dbgs() << "Cluster loads SU(" << SUa->NodeNum << ") - SU("
767 << SUb->NodeNum << ")\n");
768 // Copy successor edges from SUa to SUb. Interleaving computation
769 // dependent on SUa can prevent load combining due to register reuse.
770 // Predecessor edges do not need to be copied from SUb to SUa since nearby
771 // loads should have effectively the same inputs.
772 for (SUnit::const_succ_iterator
773 SI = SUa->Succs.begin(), SE = SUa->Succs.end(); SI != SE; ++SI) {
774 if (SI->getSUnit() == SUb)
776 DEBUG(dbgs() << " Copy Succ SU(" << SI->getSUnit()->NodeNum << ")\n");
777 DAG->addEdge(SI->getSUnit(), SDep(SUb, SDep::Artificial));
786 /// \brief Callback from DAG postProcessing to create cluster edges for loads.
787 void LoadClusterMutation::apply(ScheduleDAGMI *DAG) {
788 // Map DAG NodeNum to store chain ID.
789 DenseMap<unsigned, unsigned> StoreChainIDs;
790 // Map each store chain to a set of dependent loads.
791 SmallVector<SmallVector<SUnit*,4>, 32> StoreChainDependents;
792 for (unsigned Idx = 0, End = DAG->SUnits.size(); Idx != End; ++Idx) {
793 SUnit *SU = &DAG->SUnits[Idx];
794 if (!SU->getInstr()->mayLoad())
796 unsigned ChainPredID = DAG->SUnits.size();
797 for (SUnit::const_pred_iterator
798 PI = SU->Preds.begin(), PE = SU->Preds.end(); PI != PE; ++PI) {
800 ChainPredID = PI->getSUnit()->NodeNum;
804 // Check if this chain-like pred has been seen
805 // before. ChainPredID==MaxNodeID for loads at the top of the schedule.
806 unsigned NumChains = StoreChainDependents.size();
807 std::pair<DenseMap<unsigned, unsigned>::iterator, bool> Result =
808 StoreChainIDs.insert(std::make_pair(ChainPredID, NumChains));
810 StoreChainDependents.resize(NumChains + 1);
811 StoreChainDependents[Result.first->second].push_back(SU);
813 // Iterate over the store chains.
814 for (unsigned Idx = 0, End = StoreChainDependents.size(); Idx != End; ++Idx)
815 clusterNeighboringLoads(StoreChainDependents[Idx], DAG);
818 //===----------------------------------------------------------------------===//
819 // MacroFusion - DAG post-processing to encourage fusion of macro ops.
820 //===----------------------------------------------------------------------===//
823 /// \brief Post-process the DAG to create cluster edges between instructions
824 /// that may be fused by the processor into a single operation.
825 class MacroFusion : public ScheduleDAGMutation {
826 const TargetInstrInfo *TII;
828 MacroFusion(const TargetInstrInfo *tii): TII(tii) {}
830 virtual void apply(ScheduleDAGMI *DAG);
834 /// \brief Callback from DAG postProcessing to create cluster edges to encourage
835 /// fused operations.
836 void MacroFusion::apply(ScheduleDAGMI *DAG) {
837 // For now, assume targets can only fuse with the branch.
838 MachineInstr *Branch = DAG->ExitSU.getInstr();
842 for (unsigned Idx = DAG->SUnits.size(); Idx > 0;) {
843 SUnit *SU = &DAG->SUnits[--Idx];
844 if (!TII->shouldScheduleAdjacent(SU->getInstr(), Branch))
847 // Create a single weak edge from SU to ExitSU. The only effect is to cause
848 // bottom-up scheduling to heavily prioritize the clustered SU. There is no
849 // need to copy predecessor edges from ExitSU to SU, since top-down
850 // scheduling cannot prioritize ExitSU anyway. To defer top-down scheduling
851 // of SU, we could create an artificial edge from the deepest root, but it
852 // hasn't been needed yet.
853 bool Success = DAG->addEdge(&DAG->ExitSU, SDep(SU, SDep::Cluster));
855 assert(Success && "No DAG nodes should be reachable from ExitSU");
857 DEBUG(dbgs() << "Macro Fuse SU(" << SU->NodeNum << ")\n");
862 //===----------------------------------------------------------------------===//
863 // ConvergingScheduler - Implementation of the standard MachineSchedStrategy.
864 //===----------------------------------------------------------------------===//
867 /// ConvergingScheduler shrinks the unscheduled zone using heuristics to balance
869 class ConvergingScheduler : public MachineSchedStrategy {
871 /// Represent the type of SchedCandidate found within a single queue.
872 /// pickNodeBidirectional depends on these listed by decreasing priority.
874 NoCand, SingleExcess, SingleCritical, Cluster,
875 ResourceReduce, ResourceDemand, BotHeightReduce, BotPathReduce,
876 TopDepthReduce, TopPathReduce, SingleMax, MultiPressure, NextDefUse,
880 static const char *getReasonStr(ConvergingScheduler::CandReason Reason);
883 /// Policy for scheduling the next instruction in the candidate's zone.
886 unsigned ReduceResIdx;
887 unsigned DemandResIdx;
889 CandPolicy(): ReduceLatency(false), ReduceResIdx(0), DemandResIdx(0) {}
892 /// Status of an instruction's critical resource consumption.
893 struct SchedResourceDelta {
894 // Count critical resources in the scheduled region required by SU.
895 unsigned CritResources;
897 // Count critical resources from another region consumed by SU.
898 unsigned DemandedResources;
900 SchedResourceDelta(): CritResources(0), DemandedResources(0) {}
902 bool operator==(const SchedResourceDelta &RHS) const {
903 return CritResources == RHS.CritResources
904 && DemandedResources == RHS.DemandedResources;
906 bool operator!=(const SchedResourceDelta &RHS) const {
907 return !operator==(RHS);
911 /// Store the state used by ConvergingScheduler heuristics, required for the
912 /// lifetime of one invocation of pickNode().
913 struct SchedCandidate {
916 // The best SUnit candidate.
919 // The reason for this candidate.
922 // Register pressure values for the best candidate.
923 RegPressureDelta RPDelta;
925 // Critical resource consumption of the best candidate.
926 SchedResourceDelta ResDelta;
928 SchedCandidate(const CandPolicy &policy)
929 : Policy(policy), SU(NULL), Reason(NoCand) {}
931 bool isValid() const { return SU; }
933 // Copy the status of another candidate without changing policy.
934 void setBest(SchedCandidate &Best) {
935 assert(Best.Reason != NoCand && "uninitialized Sched candidate");
937 Reason = Best.Reason;
938 RPDelta = Best.RPDelta;
939 ResDelta = Best.ResDelta;
942 void initResourceDelta(const ScheduleDAGMI *DAG,
943 const TargetSchedModel *SchedModel);
946 /// Summarize the unscheduled region.
947 struct SchedRemainder {
948 // Critical path through the DAG in expected latency.
949 unsigned CriticalPath;
951 // Unscheduled resources
952 SmallVector<unsigned, 16> RemainingCounts;
953 // Critical resource for the unscheduled zone.
955 // Number of micro-ops left to schedule.
956 unsigned RemainingMicroOps;
957 // Is the unscheduled zone resource limited.
958 bool IsResourceLimited;
960 unsigned MaxRemainingCount;
964 RemainingCounts.clear();
966 RemainingMicroOps = 0;
967 IsResourceLimited = false;
968 MaxRemainingCount = 0;
971 SchedRemainder() { reset(); }
973 void init(ScheduleDAGMI *DAG, const TargetSchedModel *SchedModel);
976 /// Each Scheduling boundary is associated with ready queues. It tracks the
977 /// current cycle in the direction of movement, and maintains the state
978 /// of "hazards" and other interlocks at the current cycle.
979 struct SchedBoundary {
981 const TargetSchedModel *SchedModel;
984 ReadyQueue Available;
988 // For heuristics, keep a list of the nodes that immediately depend on the
989 // most recently scheduled node.
990 SmallPtrSet<const SUnit*, 8> NextSUs;
992 ScheduleHazardRecognizer *HazardRec;
997 /// MinReadyCycle - Cycle of the soonest available instruction.
998 unsigned MinReadyCycle;
1000 // The expected latency of the critical path in this scheduled zone.
1001 unsigned ExpectedLatency;
1003 // Resources used in the scheduled zone beyond this boundary.
1004 SmallVector<unsigned, 16> ResourceCounts;
1006 // Cache the critical resources ID in this scheduled zone.
1007 unsigned CritResIdx;
1009 // Is the scheduled region resource limited vs. latency limited.
1010 bool IsResourceLimited;
1012 unsigned ExpectedCount;
1014 // Policy flag: attempt to find ILP until expected latency is covered.
1015 bool ShouldIncreaseILP;
1018 // Remember the greatest min operand latency.
1019 unsigned MaxMinLatency;
1025 CheckPending = false;
1030 MinReadyCycle = UINT_MAX;
1031 ExpectedLatency = 0;
1032 ResourceCounts.resize(1);
1033 assert(!ResourceCounts[0] && "nonzero count for bad resource");
1035 IsResourceLimited = false;
1037 ShouldIncreaseILP = false;
1041 // Reserve a zero-count for invalid CritResIdx.
1042 ResourceCounts.resize(1);
1045 /// Pending queues extend the ready queues with the same ID and the
1046 /// PendingFlag set.
1047 SchedBoundary(unsigned ID, const Twine &Name):
1048 DAG(0), SchedModel(0), Rem(0), Available(ID, Name+".A"),
1049 Pending(ID << ConvergingScheduler::LogMaxQID, Name+".P") {
1053 ~SchedBoundary() { delete HazardRec; }
1055 void init(ScheduleDAGMI *dag, const TargetSchedModel *smodel,
1056 SchedRemainder *rem);
1058 bool isTop() const {
1059 return Available.getID() == ConvergingScheduler::TopQID;
1062 unsigned getUnscheduledLatency(SUnit *SU) const {
1064 return SU->getHeight();
1065 return SU->getDepth();
1068 unsigned getCriticalCount() const {
1069 return ResourceCounts[CritResIdx];
1072 bool checkHazard(SUnit *SU);
1074 void checkILPPolicy();
1076 void releaseNode(SUnit *SU, unsigned ReadyCycle);
1080 void countResource(unsigned PIdx, unsigned Cycles);
1082 void bumpNode(SUnit *SU);
1084 void releasePending();
1086 void removeReady(SUnit *SU);
1088 SUnit *pickOnlyChoice();
1093 const TargetSchedModel *SchedModel;
1094 const TargetRegisterInfo *TRI;
1096 // State of the top and bottom scheduled instruction boundaries.
1102 /// SUnit::NodeQueueId: 0 (none), 1 (top), 2 (bot), 3 (both)
1109 ConvergingScheduler():
1110 DAG(0), SchedModel(0), TRI(0), Top(TopQID, "TopQ"), Bot(BotQID, "BotQ") {}
1112 virtual void initialize(ScheduleDAGMI *dag);
1114 virtual SUnit *pickNode(bool &IsTopNode);
1116 virtual void schedNode(SUnit *SU, bool IsTopNode);
1118 virtual void releaseTopNode(SUnit *SU);
1120 virtual void releaseBottomNode(SUnit *SU);
1122 virtual void registerRoots();
1126 ConvergingScheduler::SchedBoundary &CriticalZone,
1127 ConvergingScheduler::SchedCandidate &CriticalCand,
1128 ConvergingScheduler::SchedBoundary &OppositeZone,
1129 ConvergingScheduler::SchedCandidate &OppositeCand);
1131 void checkResourceLimits(ConvergingScheduler::SchedCandidate &TopCand,
1132 ConvergingScheduler::SchedCandidate &BotCand);
1134 void tryCandidate(SchedCandidate &Cand,
1135 SchedCandidate &TryCand,
1136 SchedBoundary &Zone,
1137 const RegPressureTracker &RPTracker,
1138 RegPressureTracker &TempTracker);
1140 SUnit *pickNodeBidirectional(bool &IsTopNode);
1142 void pickNodeFromQueue(SchedBoundary &Zone,
1143 const RegPressureTracker &RPTracker,
1144 SchedCandidate &Candidate);
1147 void traceCandidate(const SchedCandidate &Cand, const SchedBoundary &Zone);
1152 void ConvergingScheduler::SchedRemainder::
1153 init(ScheduleDAGMI *DAG, const TargetSchedModel *SchedModel) {
1155 if (!SchedModel->hasInstrSchedModel())
1157 RemainingCounts.resize(SchedModel->getNumProcResourceKinds());
1158 for (std::vector<SUnit>::iterator
1159 I = DAG->SUnits.begin(), E = DAG->SUnits.end(); I != E; ++I) {
1160 const MCSchedClassDesc *SC = DAG->getSchedClass(&*I);
1161 RemainingMicroOps += SchedModel->getNumMicroOps(I->getInstr(), SC);
1162 for (TargetSchedModel::ProcResIter
1163 PI = SchedModel->getWriteProcResBegin(SC),
1164 PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
1165 unsigned PIdx = PI->ProcResourceIdx;
1166 unsigned Factor = SchedModel->getResourceFactor(PIdx);
1167 RemainingCounts[PIdx] += (Factor * PI->Cycles);
1172 void ConvergingScheduler::SchedBoundary::
1173 init(ScheduleDAGMI *dag, const TargetSchedModel *smodel, SchedRemainder *rem) {
1176 SchedModel = smodel;
1178 if (SchedModel->hasInstrSchedModel())
1179 ResourceCounts.resize(SchedModel->getNumProcResourceKinds());
1182 void ConvergingScheduler::initialize(ScheduleDAGMI *dag) {
1184 SchedModel = DAG->getSchedModel();
1186 Rem.init(DAG, SchedModel);
1187 Top.init(DAG, SchedModel, &Rem);
1188 Bot.init(DAG, SchedModel, &Rem);
1190 // Initialize resource counts.
1192 // Initialize the HazardRecognizers. If itineraries don't exist, are empty, or
1193 // are disabled, then these HazardRecs will be disabled.
1194 const InstrItineraryData *Itin = SchedModel->getInstrItineraries();
1195 const TargetMachine &TM = DAG->MF.getTarget();
1196 Top.HazardRec = TM.getInstrInfo()->CreateTargetMIHazardRecognizer(Itin, DAG);
1197 Bot.HazardRec = TM.getInstrInfo()->CreateTargetMIHazardRecognizer(Itin, DAG);
1199 assert((!ForceTopDown || !ForceBottomUp) &&
1200 "-misched-topdown incompatible with -misched-bottomup");
1203 void ConvergingScheduler::releaseTopNode(SUnit *SU) {
1204 if (SU->isScheduled)
1207 for (SUnit::succ_iterator I = SU->Preds.begin(), E = SU->Preds.end();
1209 unsigned PredReadyCycle = I->getSUnit()->TopReadyCycle;
1210 unsigned MinLatency = I->getMinLatency();
1212 Top.MaxMinLatency = std::max(MinLatency, Top.MaxMinLatency);
1214 if (SU->TopReadyCycle < PredReadyCycle + MinLatency)
1215 SU->TopReadyCycle = PredReadyCycle + MinLatency;
1217 Top.releaseNode(SU, SU->TopReadyCycle);
1220 void ConvergingScheduler::releaseBottomNode(SUnit *SU) {
1221 if (SU->isScheduled)
1224 assert(SU->getInstr() && "Scheduled SUnit must have instr");
1226 for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
1230 unsigned SuccReadyCycle = I->getSUnit()->BotReadyCycle;
1231 unsigned MinLatency = I->getMinLatency();
1233 Bot.MaxMinLatency = std::max(MinLatency, Bot.MaxMinLatency);
1235 if (SU->BotReadyCycle < SuccReadyCycle + MinLatency)
1236 SU->BotReadyCycle = SuccReadyCycle + MinLatency;
1238 Bot.releaseNode(SU, SU->BotReadyCycle);
1241 void ConvergingScheduler::registerRoots() {
1242 Rem.CriticalPath = DAG->ExitSU.getDepth();
1243 // Some roots may not feed into ExitSU. Check all of them in case.
1244 for (std::vector<SUnit*>::const_iterator
1245 I = Bot.Available.begin(), E = Bot.Available.end(); I != E; ++I) {
1246 if ((*I)->getDepth() > Rem.CriticalPath)
1247 Rem.CriticalPath = (*I)->getDepth();
1249 DEBUG(dbgs() << "Critical Path: " << Rem.CriticalPath << '\n');
1252 /// Does this SU have a hazard within the current instruction group.
1254 /// The scheduler supports two modes of hazard recognition. The first is the
1255 /// ScheduleHazardRecognizer API. It is a fully general hazard recognizer that
1256 /// supports highly complicated in-order reservation tables
1257 /// (ScoreboardHazardRecognizer) and arbitraty target-specific logic.
1259 /// The second is a streamlined mechanism that checks for hazards based on
1260 /// simple counters that the scheduler itself maintains. It explicitly checks
1261 /// for instruction dispatch limitations, including the number of micro-ops that
1262 /// can dispatch per cycle.
1264 /// TODO: Also check whether the SU must start a new group.
1265 bool ConvergingScheduler::SchedBoundary::checkHazard(SUnit *SU) {
1266 if (HazardRec->isEnabled())
1267 return HazardRec->getHazardType(SU) != ScheduleHazardRecognizer::NoHazard;
1269 unsigned uops = SchedModel->getNumMicroOps(SU->getInstr());
1270 if ((IssueCount > 0) && (IssueCount + uops > SchedModel->getIssueWidth())) {
1271 DEBUG(dbgs() << " SU(" << SU->NodeNum << ") uops="
1272 << SchedModel->getNumMicroOps(SU->getInstr()) << '\n');
1278 /// If expected latency is covered, disable ILP policy.
1279 void ConvergingScheduler::SchedBoundary::checkILPPolicy() {
1280 if (ShouldIncreaseILP
1281 && (IsResourceLimited || ExpectedLatency <= CurrCycle)) {
1282 ShouldIncreaseILP = false;
1283 DEBUG(dbgs() << "Disable ILP: " << Available.getName() << '\n');
1287 void ConvergingScheduler::SchedBoundary::releaseNode(SUnit *SU,
1288 unsigned ReadyCycle) {
1290 if (ReadyCycle < MinReadyCycle)
1291 MinReadyCycle = ReadyCycle;
1293 // Check for interlocks first. For the purpose of other heuristics, an
1294 // instruction that cannot issue appears as if it's not in the ReadyQueue.
1295 if (ReadyCycle > CurrCycle || checkHazard(SU))
1300 // Record this node as an immediate dependent of the scheduled node.
1303 // If CriticalPath has been computed, then check if the unscheduled nodes
1304 // exceed the ILP window. Before registerRoots, CriticalPath==0.
1305 if (Rem->CriticalPath && (ExpectedLatency + getUnscheduledLatency(SU)
1306 > Rem->CriticalPath + ILPWindow)) {
1307 ShouldIncreaseILP = true;
1308 DEBUG(dbgs() << "Increase ILP: " << Available.getName() << " "
1309 << ExpectedLatency << " + " << getUnscheduledLatency(SU) << '\n');
1313 /// Move the boundary of scheduled code by one cycle.
1314 void ConvergingScheduler::SchedBoundary::bumpCycle() {
1315 unsigned Width = SchedModel->getIssueWidth();
1316 IssueCount = (IssueCount <= Width) ? 0 : IssueCount - Width;
1318 unsigned NextCycle = CurrCycle + 1;
1319 assert(MinReadyCycle < UINT_MAX && "MinReadyCycle uninitialized");
1320 if (MinReadyCycle > NextCycle) {
1322 NextCycle = MinReadyCycle;
1325 if (!HazardRec->isEnabled()) {
1326 // Bypass HazardRec virtual calls.
1327 CurrCycle = NextCycle;
1330 // Bypass getHazardType calls in case of long latency.
1331 for (; CurrCycle != NextCycle; ++CurrCycle) {
1333 HazardRec->AdvanceCycle();
1335 HazardRec->RecedeCycle();
1338 CheckPending = true;
1339 IsResourceLimited = getCriticalCount() > std::max(ExpectedLatency, CurrCycle);
1341 DEBUG(dbgs() << " *** " << Available.getName() << " cycle "
1342 << CurrCycle << '\n');
1345 /// Add the given processor resource to this scheduled zone.
1346 void ConvergingScheduler::SchedBoundary::countResource(unsigned PIdx,
1348 unsigned Factor = SchedModel->getResourceFactor(PIdx);
1349 DEBUG(dbgs() << " " << SchedModel->getProcResource(PIdx)->Name
1350 << " +(" << Cycles << "x" << Factor
1351 << ") / " << SchedModel->getLatencyFactor() << '\n');
1353 unsigned Count = Factor * Cycles;
1354 ResourceCounts[PIdx] += Count;
1355 assert(Rem->RemainingCounts[PIdx] >= Count && "resource double counted");
1356 Rem->RemainingCounts[PIdx] -= Count;
1358 // Reset MaxRemainingCount for sanity.
1359 Rem->MaxRemainingCount = 0;
1361 // Check if this resource exceeds the current critical resource by a full
1362 // cycle. If so, it becomes the critical resource.
1363 if ((int)(ResourceCounts[PIdx] - ResourceCounts[CritResIdx])
1364 >= (int)SchedModel->getLatencyFactor()) {
1366 DEBUG(dbgs() << " *** Critical resource "
1367 << SchedModel->getProcResource(PIdx)->Name << " x"
1368 << ResourceCounts[PIdx] << '\n');
1372 /// Move the boundary of scheduled code by one SUnit.
1373 void ConvergingScheduler::SchedBoundary::bumpNode(SUnit *SU) {
1374 // Update the reservation table.
1375 if (HazardRec->isEnabled()) {
1376 if (!isTop() && SU->isCall) {
1377 // Calls are scheduled with their preceding instructions. For bottom-up
1378 // scheduling, clear the pipeline state before emitting.
1381 HazardRec->EmitInstruction(SU);
1383 // Update resource counts and critical resource.
1384 if (SchedModel->hasInstrSchedModel()) {
1385 const MCSchedClassDesc *SC = DAG->getSchedClass(SU);
1386 Rem->RemainingMicroOps -= SchedModel->getNumMicroOps(SU->getInstr(), SC);
1387 for (TargetSchedModel::ProcResIter
1388 PI = SchedModel->getWriteProcResBegin(SC),
1389 PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
1390 countResource(PI->ProcResourceIdx, PI->Cycles);
1394 if (SU->getDepth() > ExpectedLatency)
1395 ExpectedLatency = SU->getDepth();
1398 if (SU->getHeight() > ExpectedLatency)
1399 ExpectedLatency = SU->getHeight();
1402 IsResourceLimited = getCriticalCount() > std::max(ExpectedLatency, CurrCycle);
1404 // Check the instruction group dispatch limit.
1405 // TODO: Check if this SU must end a dispatch group.
1406 IssueCount += SchedModel->getNumMicroOps(SU->getInstr());
1408 // checkHazard prevents scheduling multiple instructions per cycle that exceed
1409 // issue width. However, we commonly reach the maximum. In this case
1410 // opportunistically bump the cycle to avoid uselessly checking everything in
1411 // the readyQ. Furthermore, a single instruction may produce more than one
1412 // cycle's worth of micro-ops.
1413 if (IssueCount >= SchedModel->getIssueWidth()) {
1414 DEBUG(dbgs() << " *** Max instrs at cycle " << CurrCycle << '\n');
1419 /// Release pending ready nodes in to the available queue. This makes them
1420 /// visible to heuristics.
1421 void ConvergingScheduler::SchedBoundary::releasePending() {
1422 // If the available queue is empty, it is safe to reset MinReadyCycle.
1423 if (Available.empty())
1424 MinReadyCycle = UINT_MAX;
1426 // Check to see if any of the pending instructions are ready to issue. If
1427 // so, add them to the available queue.
1428 for (unsigned i = 0, e = Pending.size(); i != e; ++i) {
1429 SUnit *SU = *(Pending.begin()+i);
1430 unsigned ReadyCycle = isTop() ? SU->TopReadyCycle : SU->BotReadyCycle;
1432 if (ReadyCycle < MinReadyCycle)
1433 MinReadyCycle = ReadyCycle;
1435 if (ReadyCycle > CurrCycle)
1438 if (checkHazard(SU))
1442 Pending.remove(Pending.begin()+i);
1445 DEBUG(if (!Pending.empty()) Pending.dump());
1446 CheckPending = false;
1449 /// Remove SU from the ready set for this boundary.
1450 void ConvergingScheduler::SchedBoundary::removeReady(SUnit *SU) {
1451 if (Available.isInQueue(SU))
1452 Available.remove(Available.find(SU));
1454 assert(Pending.isInQueue(SU) && "bad ready count");
1455 Pending.remove(Pending.find(SU));
1459 /// If this queue only has one ready candidate, return it. As a side effect,
1460 /// defer any nodes that now hit a hazard, and advance the cycle until at least
1461 /// one node is ready. If multiple instructions are ready, return NULL.
1462 SUnit *ConvergingScheduler::SchedBoundary::pickOnlyChoice() {
1466 if (IssueCount > 0) {
1467 // Defer any ready instrs that now have a hazard.
1468 for (ReadyQueue::iterator I = Available.begin(); I != Available.end();) {
1469 if (checkHazard(*I)) {
1471 I = Available.remove(I);
1477 for (unsigned i = 0; Available.empty(); ++i) {
1478 assert(i <= (HazardRec->getMaxLookAhead() + MaxMinLatency) &&
1479 "permanent hazard"); (void)i;
1483 if (Available.size() == 1)
1484 return *Available.begin();
1488 /// Record the candidate policy for opposite zones with different critical
1491 /// If the CriticalZone is latency limited, don't force a policy for the
1492 /// candidates here. Instead, When releasing each candidate, releaseNode
1493 /// compares the region's critical path to the candidate's height or depth and
1494 /// the scheduled zone's expected latency then sets ShouldIncreaseILP.
1495 void ConvergingScheduler::balanceZones(
1496 ConvergingScheduler::SchedBoundary &CriticalZone,
1497 ConvergingScheduler::SchedCandidate &CriticalCand,
1498 ConvergingScheduler::SchedBoundary &OppositeZone,
1499 ConvergingScheduler::SchedCandidate &OppositeCand) {
1501 if (!CriticalZone.IsResourceLimited)
1504 SchedRemainder *Rem = CriticalZone.Rem;
1506 // If the critical zone is overconsuming a resource relative to the
1507 // remainder, try to reduce it.
1508 unsigned RemainingCritCount =
1509 Rem->RemainingCounts[CriticalZone.CritResIdx];
1510 if ((int)(Rem->MaxRemainingCount - RemainingCritCount)
1511 > (int)SchedModel->getLatencyFactor()) {
1512 CriticalCand.Policy.ReduceResIdx = CriticalZone.CritResIdx;
1513 DEBUG(dbgs() << "Balance " << CriticalZone.Available.getName() << " reduce "
1514 << SchedModel->getProcResource(CriticalZone.CritResIdx)->Name
1517 // If the other zone is underconsuming a resource relative to the full zone,
1518 // try to increase it.
1519 unsigned OppositeCount =
1520 OppositeZone.ResourceCounts[CriticalZone.CritResIdx];
1521 if ((int)(OppositeZone.ExpectedCount - OppositeCount)
1522 > (int)SchedModel->getLatencyFactor()) {
1523 OppositeCand.Policy.DemandResIdx = CriticalZone.CritResIdx;
1524 DEBUG(dbgs() << "Balance " << OppositeZone.Available.getName() << " demand "
1525 << SchedModel->getProcResource(OppositeZone.CritResIdx)->Name
1530 /// Determine if the scheduled zones exceed resource limits or critical path and
1531 /// set each candidate's ReduceHeight policy accordingly.
1532 void ConvergingScheduler::checkResourceLimits(
1533 ConvergingScheduler::SchedCandidate &TopCand,
1534 ConvergingScheduler::SchedCandidate &BotCand) {
1536 Bot.checkILPPolicy();
1537 Top.checkILPPolicy();
1538 if (Bot.ShouldIncreaseILP)
1539 BotCand.Policy.ReduceLatency = true;
1540 if (Top.ShouldIncreaseILP)
1541 TopCand.Policy.ReduceLatency = true;
1543 // Handle resource-limited regions.
1544 if (Top.IsResourceLimited && Bot.IsResourceLimited
1545 && Top.CritResIdx == Bot.CritResIdx) {
1546 // If the scheduled critical resource in both zones is no longer the
1547 // critical remaining resource, attempt to reduce resource height both ways.
1548 if (Top.CritResIdx != Rem.CritResIdx) {
1549 TopCand.Policy.ReduceResIdx = Top.CritResIdx;
1550 BotCand.Policy.ReduceResIdx = Bot.CritResIdx;
1551 DEBUG(dbgs() << "Reduce scheduled "
1552 << SchedModel->getProcResource(Top.CritResIdx)->Name << '\n');
1556 // Handle latency-limited regions.
1557 if (!Top.IsResourceLimited && !Bot.IsResourceLimited) {
1558 // If the total scheduled expected latency exceeds the region's critical
1559 // path then reduce latency both ways.
1561 // Just because a zone is not resource limited does not mean it is latency
1562 // limited. Unbuffered resource, such as max micro-ops may cause CurrCycle
1563 // to exceed expected latency.
1564 if ((Top.ExpectedLatency + Bot.ExpectedLatency >= Rem.CriticalPath)
1565 && (Rem.CriticalPath > Top.CurrCycle + Bot.CurrCycle)) {
1566 TopCand.Policy.ReduceLatency = true;
1567 BotCand.Policy.ReduceLatency = true;
1568 DEBUG(dbgs() << "Reduce scheduled latency " << Top.ExpectedLatency
1569 << " + " << Bot.ExpectedLatency << '\n');
1573 // The critical resource is different in each zone, so request balancing.
1575 // Compute the cost of each zone.
1576 Rem.MaxRemainingCount = std::max(
1577 Rem.RemainingMicroOps * SchedModel->getMicroOpFactor(),
1578 Rem.RemainingCounts[Rem.CritResIdx]);
1579 Top.ExpectedCount = std::max(Top.ExpectedLatency, Top.CurrCycle);
1580 Top.ExpectedCount = std::max(
1581 Top.getCriticalCount(),
1582 Top.ExpectedCount * SchedModel->getLatencyFactor());
1583 Bot.ExpectedCount = std::max(Bot.ExpectedLatency, Bot.CurrCycle);
1584 Bot.ExpectedCount = std::max(
1585 Bot.getCriticalCount(),
1586 Bot.ExpectedCount * SchedModel->getLatencyFactor());
1588 balanceZones(Top, TopCand, Bot, BotCand);
1589 balanceZones(Bot, BotCand, Top, TopCand);
1592 void ConvergingScheduler::SchedCandidate::
1593 initResourceDelta(const ScheduleDAGMI *DAG,
1594 const TargetSchedModel *SchedModel) {
1595 if (!Policy.ReduceResIdx && !Policy.DemandResIdx)
1598 const MCSchedClassDesc *SC = DAG->getSchedClass(SU);
1599 for (TargetSchedModel::ProcResIter
1600 PI = SchedModel->getWriteProcResBegin(SC),
1601 PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
1602 if (PI->ProcResourceIdx == Policy.ReduceResIdx)
1603 ResDelta.CritResources += PI->Cycles;
1604 if (PI->ProcResourceIdx == Policy.DemandResIdx)
1605 ResDelta.DemandedResources += PI->Cycles;
1609 /// Return true if this heuristic determines order.
1610 static bool tryLess(unsigned TryVal, unsigned CandVal,
1611 ConvergingScheduler::SchedCandidate &TryCand,
1612 ConvergingScheduler::SchedCandidate &Cand,
1613 ConvergingScheduler::CandReason Reason) {
1614 if (TryVal < CandVal) {
1615 TryCand.Reason = Reason;
1618 if (TryVal > CandVal) {
1619 if (Cand.Reason > Reason)
1620 Cand.Reason = Reason;
1626 static bool tryGreater(unsigned TryVal, unsigned CandVal,
1627 ConvergingScheduler::SchedCandidate &TryCand,
1628 ConvergingScheduler::SchedCandidate &Cand,
1629 ConvergingScheduler::CandReason Reason) {
1630 if (TryVal > CandVal) {
1631 TryCand.Reason = Reason;
1634 if (TryVal < CandVal) {
1635 if (Cand.Reason > Reason)
1636 Cand.Reason = Reason;
1642 static unsigned getWeakLeft(const SUnit *SU, bool isTop) {
1643 return (isTop) ? SU->WeakPredsLeft : SU->WeakSuccsLeft;
1646 /// Apply a set of heursitics to a new candidate. Heuristics are currently
1647 /// hierarchical. This may be more efficient than a graduated cost model because
1648 /// we don't need to evaluate all aspects of the model for each node in the
1649 /// queue. But it's really done to make the heuristics easier to debug and
1650 /// statistically analyze.
1652 /// \param Cand provides the policy and current best candidate.
1653 /// \param TryCand refers to the next SUnit candidate, otherwise uninitialized.
1654 /// \param Zone describes the scheduled zone that we are extending.
1655 /// \param RPTracker describes reg pressure within the scheduled zone.
1656 /// \param TempTracker is a scratch pressure tracker to reuse in queries.
1657 void ConvergingScheduler::tryCandidate(SchedCandidate &Cand,
1658 SchedCandidate &TryCand,
1659 SchedBoundary &Zone,
1660 const RegPressureTracker &RPTracker,
1661 RegPressureTracker &TempTracker) {
1663 // Always initialize TryCand's RPDelta.
1664 TempTracker.getMaxPressureDelta(TryCand.SU->getInstr(), TryCand.RPDelta,
1665 DAG->getRegionCriticalPSets(),
1666 DAG->getRegPressure().MaxSetPressure);
1668 // Initialize the candidate if needed.
1669 if (!Cand.isValid()) {
1670 TryCand.Reason = NodeOrder;
1673 // Avoid exceeding the target's limit.
1674 if (tryLess(TryCand.RPDelta.Excess.UnitIncrease,
1675 Cand.RPDelta.Excess.UnitIncrease, TryCand, Cand, SingleExcess))
1677 if (Cand.Reason == SingleExcess)
1678 Cand.Reason = MultiPressure;
1680 // Avoid increasing the max critical pressure in the scheduled region.
1681 if (tryLess(TryCand.RPDelta.CriticalMax.UnitIncrease,
1682 Cand.RPDelta.CriticalMax.UnitIncrease,
1683 TryCand, Cand, SingleCritical))
1685 if (Cand.Reason == SingleCritical)
1686 Cand.Reason = MultiPressure;
1688 // Keep clustered nodes together to encourage downstream peephole
1689 // optimizations which may reduce resource requirements.
1691 // This is a best effort to set things up for a post-RA pass. Optimizations
1692 // like generating loads of multiple registers should ideally be done within
1693 // the scheduler pass by combining the loads during DAG postprocessing.
1694 const SUnit *NextClusterSU =
1695 Zone.isTop() ? DAG->getNextClusterSucc() : DAG->getNextClusterPred();
1696 if (tryGreater(TryCand.SU == NextClusterSU, Cand.SU == NextClusterSU,
1697 TryCand, Cand, Cluster))
1699 // Currently, weak edges are for clustering, so we hard-code that reason.
1700 // However, deferring the current TryCand will not change Cand's reason.
1701 CandReason OrigReason = Cand.Reason;
1702 if (tryLess(getWeakLeft(TryCand.SU, Zone.isTop()),
1703 getWeakLeft(Cand.SU, Zone.isTop()),
1704 TryCand, Cand, Cluster)) {
1705 Cand.Reason = OrigReason;
1708 // Avoid critical resource consumption and balance the schedule.
1709 TryCand.initResourceDelta(DAG, SchedModel);
1710 if (tryLess(TryCand.ResDelta.CritResources, Cand.ResDelta.CritResources,
1711 TryCand, Cand, ResourceReduce))
1713 if (tryGreater(TryCand.ResDelta.DemandedResources,
1714 Cand.ResDelta.DemandedResources,
1715 TryCand, Cand, ResourceDemand))
1718 // Avoid serializing long latency dependence chains.
1719 if (Cand.Policy.ReduceLatency) {
1721 if (Cand.SU->getDepth() * SchedModel->getLatencyFactor()
1722 > Zone.ExpectedCount) {
1723 if (tryLess(TryCand.SU->getDepth(), Cand.SU->getDepth(),
1724 TryCand, Cand, TopDepthReduce))
1727 if (tryGreater(TryCand.SU->getHeight(), Cand.SU->getHeight(),
1728 TryCand, Cand, TopPathReduce))
1732 if (Cand.SU->getHeight() * SchedModel->getLatencyFactor()
1733 > Zone.ExpectedCount) {
1734 if (tryLess(TryCand.SU->getHeight(), Cand.SU->getHeight(),
1735 TryCand, Cand, BotHeightReduce))
1738 if (tryGreater(TryCand.SU->getDepth(), Cand.SU->getDepth(),
1739 TryCand, Cand, BotPathReduce))
1744 // Avoid increasing the max pressure of the entire region.
1745 if (tryLess(TryCand.RPDelta.CurrentMax.UnitIncrease,
1746 Cand.RPDelta.CurrentMax.UnitIncrease, TryCand, Cand, SingleMax))
1748 if (Cand.Reason == SingleMax)
1749 Cand.Reason = MultiPressure;
1751 // Prefer immediate defs/users of the last scheduled instruction. This is a
1752 // nice pressure avoidance strategy that also conserves the processor's
1753 // register renaming resources and keeps the machine code readable.
1754 if (tryGreater(Zone.NextSUs.count(TryCand.SU), Zone.NextSUs.count(Cand.SU),
1755 TryCand, Cand, NextDefUse))
1758 // Fall through to original instruction order.
1759 if ((Zone.isTop() && TryCand.SU->NodeNum < Cand.SU->NodeNum)
1760 || (!Zone.isTop() && TryCand.SU->NodeNum > Cand.SU->NodeNum)) {
1761 TryCand.Reason = NodeOrder;
1765 /// pickNodeFromQueue helper that returns true if the LHS reg pressure effect is
1766 /// more desirable than RHS from scheduling standpoint.
1767 static bool compareRPDelta(const RegPressureDelta &LHS,
1768 const RegPressureDelta &RHS) {
1769 // Compare each component of pressure in decreasing order of importance
1770 // without checking if any are valid. Invalid PressureElements are assumed to
1771 // have UnitIncrease==0, so are neutral.
1773 // Avoid increasing the max critical pressure in the scheduled region.
1774 if (LHS.Excess.UnitIncrease != RHS.Excess.UnitIncrease) {
1775 DEBUG(dbgs() << "RP excess top - bot: "
1776 << (LHS.Excess.UnitIncrease - RHS.Excess.UnitIncrease) << '\n');
1777 return LHS.Excess.UnitIncrease < RHS.Excess.UnitIncrease;
1779 // Avoid increasing the max critical pressure in the scheduled region.
1780 if (LHS.CriticalMax.UnitIncrease != RHS.CriticalMax.UnitIncrease) {
1781 DEBUG(dbgs() << "RP critical top - bot: "
1782 << (LHS.CriticalMax.UnitIncrease - RHS.CriticalMax.UnitIncrease)
1784 return LHS.CriticalMax.UnitIncrease < RHS.CriticalMax.UnitIncrease;
1786 // Avoid increasing the max pressure of the entire region.
1787 if (LHS.CurrentMax.UnitIncrease != RHS.CurrentMax.UnitIncrease) {
1788 DEBUG(dbgs() << "RP current top - bot: "
1789 << (LHS.CurrentMax.UnitIncrease - RHS.CurrentMax.UnitIncrease)
1791 return LHS.CurrentMax.UnitIncrease < RHS.CurrentMax.UnitIncrease;
1797 const char *ConvergingScheduler::getReasonStr(
1798 ConvergingScheduler::CandReason Reason) {
1800 case NoCand: return "NOCAND ";
1801 case SingleExcess: return "REG-EXCESS";
1802 case SingleCritical: return "REG-CRIT ";
1803 case Cluster: return "CLUSTER ";
1804 case SingleMax: return "REG-MAX ";
1805 case MultiPressure: return "REG-MULTI ";
1806 case ResourceReduce: return "RES-REDUCE";
1807 case ResourceDemand: return "RES-DEMAND";
1808 case TopDepthReduce: return "TOP-DEPTH ";
1809 case TopPathReduce: return "TOP-PATH ";
1810 case BotHeightReduce:return "BOT-HEIGHT";
1811 case BotPathReduce: return "BOT-PATH ";
1812 case NextDefUse: return "DEF-USE ";
1813 case NodeOrder: return "ORDER ";
1815 llvm_unreachable("Unknown reason!");
1818 void ConvergingScheduler::traceCandidate(const SchedCandidate &Cand,
1819 const SchedBoundary &Zone) {
1820 const char *Label = getReasonStr(Cand.Reason);
1822 unsigned ResIdx = 0;
1823 unsigned Latency = 0;
1824 switch (Cand.Reason) {
1828 P = Cand.RPDelta.Excess;
1830 case SingleCritical:
1831 P = Cand.RPDelta.CriticalMax;
1834 P = Cand.RPDelta.CurrentMax;
1836 case ResourceReduce:
1837 ResIdx = Cand.Policy.ReduceResIdx;
1839 case ResourceDemand:
1840 ResIdx = Cand.Policy.DemandResIdx;
1842 case TopDepthReduce:
1843 Latency = Cand.SU->getDepth();
1846 Latency = Cand.SU->getHeight();
1848 case BotHeightReduce:
1849 Latency = Cand.SU->getHeight();
1852 Latency = Cand.SU->getDepth();
1855 dbgs() << Label << " " << Zone.Available.getName() << " ";
1857 dbgs() << TRI->getRegPressureSetName(P.PSetID) << ":" << P.UnitIncrease
1862 dbgs() << SchedModel->getProcResource(ResIdx)->Name << " ";
1866 dbgs() << Latency << " cycles ";
1873 /// Pick the best candidate from the top queue.
1875 /// TODO: getMaxPressureDelta results can be mostly cached for each SUnit during
1876 /// DAG building. To adjust for the current scheduling location we need to
1877 /// maintain the number of vreg uses remaining to be top-scheduled.
1878 void ConvergingScheduler::pickNodeFromQueue(SchedBoundary &Zone,
1879 const RegPressureTracker &RPTracker,
1880 SchedCandidate &Cand) {
1881 ReadyQueue &Q = Zone.Available;
1885 // getMaxPressureDelta temporarily modifies the tracker.
1886 RegPressureTracker &TempTracker = const_cast<RegPressureTracker&>(RPTracker);
1888 for (ReadyQueue::iterator I = Q.begin(), E = Q.end(); I != E; ++I) {
1890 SchedCandidate TryCand(Cand.Policy);
1892 tryCandidate(Cand, TryCand, Zone, RPTracker, TempTracker);
1893 if (TryCand.Reason != NoCand) {
1894 // Initialize resource delta if needed in case future heuristics query it.
1895 if (TryCand.ResDelta == SchedResourceDelta())
1896 TryCand.initResourceDelta(DAG, SchedModel);
1897 Cand.setBest(TryCand);
1898 DEBUG(traceCandidate(Cand, Zone));
1904 static void tracePick(const ConvergingScheduler::SchedCandidate &Cand,
1906 DEBUG(dbgs() << "Pick " << (IsTop ? "top" : "bot")
1907 << " SU(" << Cand.SU->NodeNum << ") "
1908 << ConvergingScheduler::getReasonStr(Cand.Reason) << '\n');
1911 /// Pick the best candidate node from either the top or bottom queue.
1912 SUnit *ConvergingScheduler::pickNodeBidirectional(bool &IsTopNode) {
1913 // Schedule as far as possible in the direction of no choice. This is most
1914 // efficient, but also provides the best heuristics for CriticalPSets.
1915 if (SUnit *SU = Bot.pickOnlyChoice()) {
1919 if (SUnit *SU = Top.pickOnlyChoice()) {
1923 CandPolicy NoPolicy;
1924 SchedCandidate BotCand(NoPolicy);
1925 SchedCandidate TopCand(NoPolicy);
1926 checkResourceLimits(TopCand, BotCand);
1928 // Prefer bottom scheduling when heuristics are silent.
1929 pickNodeFromQueue(Bot, DAG->getBotRPTracker(), BotCand);
1930 assert(BotCand.Reason != NoCand && "failed to find the first candidate");
1932 // If either Q has a single candidate that provides the least increase in
1933 // Excess pressure, we can immediately schedule from that Q.
1935 // RegionCriticalPSets summarizes the pressure within the scheduled region and
1936 // affects picking from either Q. If scheduling in one direction must
1937 // increase pressure for one of the excess PSets, then schedule in that
1938 // direction first to provide more freedom in the other direction.
1939 if (BotCand.Reason == SingleExcess || BotCand.Reason == SingleCritical) {
1941 tracePick(BotCand, IsTopNode);
1944 // Check if the top Q has a better candidate.
1945 pickNodeFromQueue(Top, DAG->getTopRPTracker(), TopCand);
1946 assert(TopCand.Reason != NoCand && "failed to find the first candidate");
1948 // If either Q has a single candidate that minimizes pressure above the
1949 // original region's pressure pick it.
1950 if (TopCand.Reason <= SingleMax || BotCand.Reason <= SingleMax) {
1951 if (TopCand.Reason < BotCand.Reason) {
1953 tracePick(TopCand, IsTopNode);
1957 tracePick(BotCand, IsTopNode);
1960 // Check for a salient pressure difference and pick the best from either side.
1961 if (compareRPDelta(TopCand.RPDelta, BotCand.RPDelta)) {
1963 tracePick(TopCand, IsTopNode);
1966 // Otherwise prefer the bottom candidate, in node order if all else failed.
1967 if (TopCand.Reason < BotCand.Reason) {
1969 tracePick(TopCand, IsTopNode);
1973 tracePick(BotCand, IsTopNode);
1977 /// Pick the best node to balance the schedule. Implements MachineSchedStrategy.
1978 SUnit *ConvergingScheduler::pickNode(bool &IsTopNode) {
1979 if (DAG->top() == DAG->bottom()) {
1980 assert(Top.Available.empty() && Top.Pending.empty() &&
1981 Bot.Available.empty() && Bot.Pending.empty() && "ReadyQ garbage");
1987 SU = Top.pickOnlyChoice();
1989 CandPolicy NoPolicy;
1990 SchedCandidate TopCand(NoPolicy);
1991 pickNodeFromQueue(Top, DAG->getTopRPTracker(), TopCand);
1992 assert(TopCand.Reason != NoCand && "failed to find the first candidate");
1997 else if (ForceBottomUp) {
1998 SU = Bot.pickOnlyChoice();
2000 CandPolicy NoPolicy;
2001 SchedCandidate BotCand(NoPolicy);
2002 pickNodeFromQueue(Bot, DAG->getBotRPTracker(), BotCand);
2003 assert(BotCand.Reason != NoCand && "failed to find the first candidate");
2009 SU = pickNodeBidirectional(IsTopNode);
2011 } while (SU->isScheduled);
2013 if (SU->isTopReady())
2014 Top.removeReady(SU);
2015 if (SU->isBottomReady())
2016 Bot.removeReady(SU);
2018 DEBUG(dbgs() << "*** " << (IsTopNode ? "Top" : "Bottom")
2019 << " Scheduling Instruction in cycle "
2020 << (IsTopNode ? Top.CurrCycle : Bot.CurrCycle) << '\n';
2025 /// Update the scheduler's state after scheduling a node. This is the same node
2026 /// that was just returned by pickNode(). However, ScheduleDAGMI needs to update
2027 /// it's state based on the current cycle before MachineSchedStrategy does.
2028 void ConvergingScheduler::schedNode(SUnit *SU, bool IsTopNode) {
2030 SU->TopReadyCycle = Top.CurrCycle;
2034 SU->BotReadyCycle = Bot.CurrCycle;
2039 /// Create the standard converging machine scheduler. This will be used as the
2040 /// default scheduler if the target does not set a default.
2041 static ScheduleDAGInstrs *createConvergingSched(MachineSchedContext *C) {
2042 assert((!ForceTopDown || !ForceBottomUp) &&
2043 "-misched-topdown incompatible with -misched-bottomup");
2044 ScheduleDAGMI *DAG = new ScheduleDAGMI(C, new ConvergingScheduler());
2045 // Register DAG post-processors.
2046 if (EnableLoadCluster)
2047 DAG->addMutation(new LoadClusterMutation(DAG->TII, DAG->TRI));
2048 if (EnableMacroFusion)
2049 DAG->addMutation(new MacroFusion(DAG->TII));
2052 static MachineSchedRegistry
2053 ConvergingSchedRegistry("converge", "Standard converging scheduler.",
2054 createConvergingSched);
2056 //===----------------------------------------------------------------------===//
2057 // ILP Scheduler. Currently for experimental analysis of heuristics.
2058 //===----------------------------------------------------------------------===//
2061 /// \brief Order nodes by the ILP metric.
2063 SchedDFSResult *DFSResult;
2064 BitVector *ScheduledTrees;
2067 ILPOrder(SchedDFSResult *dfs, BitVector *schedtrees, bool MaxILP)
2068 : DFSResult(dfs), ScheduledTrees(schedtrees), MaximizeILP(MaxILP) {}
2070 /// \brief Apply a less-than relation on node priority.
2072 /// (Return true if A comes after B in the Q.)
2073 bool operator()(const SUnit *A, const SUnit *B) const {
2074 unsigned SchedTreeA = DFSResult->getSubtreeID(A);
2075 unsigned SchedTreeB = DFSResult->getSubtreeID(B);
2076 if (SchedTreeA != SchedTreeB) {
2077 // Unscheduled trees have lower priority.
2078 if (ScheduledTrees->test(SchedTreeA) != ScheduledTrees->test(SchedTreeB))
2079 return ScheduledTrees->test(SchedTreeB);
2081 // Trees with shallower connections have have lower priority.
2082 if (DFSResult->getSubtreeLevel(SchedTreeA)
2083 != DFSResult->getSubtreeLevel(SchedTreeB)) {
2084 return DFSResult->getSubtreeLevel(SchedTreeA)
2085 < DFSResult->getSubtreeLevel(SchedTreeB);
2089 return DFSResult->getILP(A) < DFSResult->getILP(B);
2091 return DFSResult->getILP(A) > DFSResult->getILP(B);
2095 /// \brief Schedule based on the ILP metric.
2096 class ILPScheduler : public MachineSchedStrategy {
2097 /// In case all subtrees are eventually connected to a common root through
2098 /// data dependence (e.g. reduction), place an upper limit on their size.
2100 /// FIXME: A subtree limit is generally good, but in the situation commented
2101 /// above, where multiple similar subtrees feed a common root, we should
2102 /// only split at a point where the resulting subtrees will be balanced.
2103 /// (a motivating test case must be found).
2104 static const unsigned SubtreeLimit = 16;
2106 SchedDFSResult DFSResult;
2107 BitVector ScheduledTrees;
2110 std::vector<SUnit*> ReadyQ;
2112 ILPScheduler(bool MaximizeILP)
2113 : DFSResult(/*BottomUp=*/true, SubtreeLimit),
2114 Cmp(&DFSResult, &ScheduledTrees, MaximizeILP) {}
2116 virtual void initialize(ScheduleDAGMI *DAG) {
2119 DFSResult.resize(DAG->SUnits.size());
2120 ScheduledTrees.clear();
2123 virtual void registerRoots() {
2124 DFSResult.compute(ReadyQ);
2125 ScheduledTrees.resize(DFSResult.getNumSubtrees());
2126 // Restore the heap in ReadyQ with the updated DFS results.
2127 std::make_heap(ReadyQ.begin(), ReadyQ.end(), Cmp);
2130 /// Implement MachineSchedStrategy interface.
2131 /// -----------------------------------------
2133 /// Callback to select the highest priority node from the ready Q.
2134 virtual SUnit *pickNode(bool &IsTopNode) {
2135 if (ReadyQ.empty()) return NULL;
2136 pop_heap(ReadyQ.begin(), ReadyQ.end(), Cmp);
2137 SUnit *SU = ReadyQ.back();
2140 DEBUG(dbgs() << "*** Scheduling " << "SU(" << SU->NodeNum << "): "
2142 << " ILP: " << DFSResult.getILP(SU)
2143 << " Tree: " << DFSResult.getSubtreeID(SU) << " @"
2144 << DFSResult.getSubtreeLevel(DFSResult.getSubtreeID(SU))<< '\n');
2148 /// Callback after a node is scheduled. Mark a newly scheduled tree, notify
2149 /// DFSResults, and resort the priority Q.
2150 virtual void schedNode(SUnit *SU, bool IsTopNode) {
2151 assert(!IsTopNode && "SchedDFSResult needs bottom-up");
2152 if (!ScheduledTrees.test(DFSResult.getSubtreeID(SU))) {
2153 ScheduledTrees.set(DFSResult.getSubtreeID(SU));
2154 DFSResult.scheduleTree(DFSResult.getSubtreeID(SU));
2155 std::make_heap(ReadyQ.begin(), ReadyQ.end(), Cmp);
2159 virtual void releaseTopNode(SUnit *) { /*only called for top roots*/ }
2161 virtual void releaseBottomNode(SUnit *SU) {
2162 ReadyQ.push_back(SU);
2163 std::push_heap(ReadyQ.begin(), ReadyQ.end(), Cmp);
2168 static ScheduleDAGInstrs *createILPMaxScheduler(MachineSchedContext *C) {
2169 return new ScheduleDAGMI(C, new ILPScheduler(true));
2171 static ScheduleDAGInstrs *createILPMinScheduler(MachineSchedContext *C) {
2172 return new ScheduleDAGMI(C, new ILPScheduler(false));
2174 static MachineSchedRegistry ILPMaxRegistry(
2175 "ilpmax", "Schedule bottom-up for max ILP", createILPMaxScheduler);
2176 static MachineSchedRegistry ILPMinRegistry(
2177 "ilpmin", "Schedule bottom-up for min ILP", createILPMinScheduler);
2179 //===----------------------------------------------------------------------===//
2180 // Machine Instruction Shuffler for Correctness Testing
2181 //===----------------------------------------------------------------------===//
2185 /// Apply a less-than relation on the node order, which corresponds to the
2186 /// instruction order prior to scheduling. IsReverse implements greater-than.
2187 template<bool IsReverse>
2189 bool operator()(SUnit *A, SUnit *B) const {
2191 return A->NodeNum > B->NodeNum;
2193 return A->NodeNum < B->NodeNum;
2197 /// Reorder instructions as much as possible.
2198 class InstructionShuffler : public MachineSchedStrategy {
2202 // Using a less-than relation (SUnitOrder<false>) for the TopQ priority
2203 // gives nodes with a higher number higher priority causing the latest
2204 // instructions to be scheduled first.
2205 PriorityQueue<SUnit*, std::vector<SUnit*>, SUnitOrder<false> >
2207 // When scheduling bottom-up, use greater-than as the queue priority.
2208 PriorityQueue<SUnit*, std::vector<SUnit*>, SUnitOrder<true> >
2211 InstructionShuffler(bool alternate, bool topdown)
2212 : IsAlternating(alternate), IsTopDown(topdown) {}
2214 virtual void initialize(ScheduleDAGMI *) {
2219 /// Implement MachineSchedStrategy interface.
2220 /// -----------------------------------------
2222 virtual SUnit *pickNode(bool &IsTopNode) {
2226 if (TopQ.empty()) return NULL;
2229 } while (SU->isScheduled);
2234 if (BottomQ.empty()) return NULL;
2237 } while (SU->isScheduled);
2241 IsTopDown = !IsTopDown;
2245 virtual void schedNode(SUnit *SU, bool IsTopNode) {}
2247 virtual void releaseTopNode(SUnit *SU) {
2250 virtual void releaseBottomNode(SUnit *SU) {
2256 static ScheduleDAGInstrs *createInstructionShuffler(MachineSchedContext *C) {
2257 bool Alternate = !ForceTopDown && !ForceBottomUp;
2258 bool TopDown = !ForceBottomUp;
2259 assert((TopDown || !ForceTopDown) &&
2260 "-misched-topdown incompatible with -misched-bottomup");
2261 return new ScheduleDAGMI(C, new InstructionShuffler(Alternate, TopDown));
2263 static MachineSchedRegistry ShufflerRegistry(
2264 "shuffle", "Shuffle machine instructions alternating directions",
2265 createInstructionShuffler);