1 //===- ScheduleDAG.cpp - Implement the ScheduleDAG class ------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 /// \file Implements the ScheduleDAG class, which is a base class used by 10 /// scheduling implementation classes. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "llvm/CodeGen/ScheduleDAG.h" 15 #include "llvm/ADT/STLExtras.h" 16 #include "llvm/ADT/SmallVector.h" 17 #include "llvm/ADT/Statistic.h" 18 #include "llvm/ADT/iterator_range.h" 19 #include "llvm/CodeGen/MachineFunction.h" 20 #include "llvm/CodeGen/ScheduleHazardRecognizer.h" 21 #include "llvm/CodeGen/SelectionDAGNodes.h" 22 #include "llvm/CodeGen/TargetInstrInfo.h" 23 #include "llvm/CodeGen/TargetRegisterInfo.h" 24 #include "llvm/CodeGen/TargetSubtargetInfo.h" 25 #include "llvm/Config/llvm-config.h" 26 #include "llvm/Support/CommandLine.h" 27 #include "llvm/Support/Compiler.h" 28 #include "llvm/Support/Debug.h" 29 #include "llvm/Support/raw_ostream.h" 30 #include <algorithm> 31 #include <cassert> 32 #include <iterator> 33 #include <limits> 34 #include <utility> 35 #include <vector> 36 37 using namespace llvm; 38 39 #define DEBUG_TYPE "pre-RA-sched" 40 41 STATISTIC(NumNewPredsAdded, "Number of times a single predecessor was added"); 42 STATISTIC(NumTopoInits, 43 "Number of times the topological order has been recomputed"); 44 45 #ifndef NDEBUG 46 static cl::opt<bool> StressSchedOpt( 47 "stress-sched", cl::Hidden, cl::init(false), 48 cl::desc("Stress test instruction scheduling")); 49 #endif 50 51 void SchedulingPriorityQueue::anchor() {} 52 53 ScheduleDAG::ScheduleDAG(MachineFunction &mf) 54 : TM(mf.getTarget()), TII(mf.getSubtarget().getInstrInfo()), 55 TRI(mf.getSubtarget().getRegisterInfo()), MF(mf), 56 MRI(mf.getRegInfo()) { 57 #ifndef NDEBUG 58 StressSched = StressSchedOpt; 59 #endif 60 } 61 62 ScheduleDAG::~ScheduleDAG() = default; 63 64 void ScheduleDAG::clearDAG() { 65 SUnits.clear(); 66 EntrySU = SUnit(); 67 ExitSU = SUnit(); 68 } 69 70 const MCInstrDesc *ScheduleDAG::getNodeDesc(const SDNode *Node) const { 71 if (!Node || !Node->isMachineOpcode()) return nullptr; 72 return &TII->get(Node->getMachineOpcode()); 73 } 74 75 LLVM_DUMP_METHOD void SDep::dump(const TargetRegisterInfo *TRI) const { 76 switch (getKind()) { 77 case Data: dbgs() << "Data"; break; 78 case Anti: dbgs() << "Anti"; break; 79 case Output: dbgs() << "Out "; break; 80 case Order: dbgs() << "Ord "; break; 81 } 82 83 switch (getKind()) { 84 case Data: 85 dbgs() << " Latency=" << getLatency(); 86 if (TRI && isAssignedRegDep()) 87 dbgs() << " Reg=" << printReg(getReg(), TRI); 88 break; 89 case Anti: 90 case Output: 91 dbgs() << " Latency=" << getLatency(); 92 break; 93 case Order: 94 dbgs() << " Latency=" << getLatency(); 95 switch(Contents.OrdKind) { 96 case Barrier: dbgs() << " Barrier"; break; 97 case MayAliasMem: 98 case MustAliasMem: dbgs() << " Memory"; break; 99 case Artificial: dbgs() << " Artificial"; break; 100 case Weak: dbgs() << " Weak"; break; 101 case Cluster: dbgs() << " Cluster"; break; 102 } 103 break; 104 } 105 } 106 107 bool SUnit::addPred(const SDep &D, bool Required) { 108 // If this node already has this dependence, don't add a redundant one. 109 for (SDep &PredDep : Preds) { 110 // Zero-latency weak edges may be added purely for heuristic ordering. Don't 111 // add them if another kind of edge already exists. 112 if (!Required && PredDep.getSUnit() == D.getSUnit()) 113 return false; 114 if (PredDep.overlaps(D)) { 115 // Extend the latency if needed. Equivalent to 116 // removePred(PredDep) + addPred(D). 117 if (PredDep.getLatency() < D.getLatency()) { 118 SUnit *PredSU = PredDep.getSUnit(); 119 // Find the corresponding successor in N. 120 SDep ForwardD = PredDep; 121 ForwardD.setSUnit(this); 122 for (SDep &SuccDep : PredSU->Succs) { 123 if (SuccDep == ForwardD) { 124 SuccDep.setLatency(D.getLatency()); 125 break; 126 } 127 } 128 PredDep.setLatency(D.getLatency()); 129 } 130 return false; 131 } 132 } 133 // Now add a corresponding succ to N. 134 SDep P = D; 135 P.setSUnit(this); 136 SUnit *N = D.getSUnit(); 137 // Update the bookkeeping. 138 if (D.getKind() == SDep::Data) { 139 assert(NumPreds < std::numeric_limits<unsigned>::max() && 140 "NumPreds will overflow!"); 141 assert(N->NumSuccs < std::numeric_limits<unsigned>::max() && 142 "NumSuccs will overflow!"); 143 ++NumPreds; 144 ++N->NumSuccs; 145 } 146 if (!N->isScheduled) { 147 if (D.isWeak()) { 148 ++WeakPredsLeft; 149 } 150 else { 151 assert(NumPredsLeft < std::numeric_limits<unsigned>::max() && 152 "NumPredsLeft will overflow!"); 153 ++NumPredsLeft; 154 } 155 } 156 if (!isScheduled) { 157 if (D.isWeak()) { 158 ++N->WeakSuccsLeft; 159 } 160 else { 161 assert(N->NumSuccsLeft < std::numeric_limits<unsigned>::max() && 162 "NumSuccsLeft will overflow!"); 163 ++N->NumSuccsLeft; 164 } 165 } 166 Preds.push_back(D); 167 N->Succs.push_back(P); 168 if (P.getLatency() != 0) { 169 this->setDepthDirty(); 170 N->setHeightDirty(); 171 } 172 return true; 173 } 174 175 void SUnit::removePred(const SDep &D) { 176 // Find the matching predecessor. 177 SmallVectorImpl<SDep>::iterator I = llvm::find(Preds, D); 178 if (I == Preds.end()) 179 return; 180 // Find the corresponding successor in N. 181 SDep P = D; 182 P.setSUnit(this); 183 SUnit *N = D.getSUnit(); 184 SmallVectorImpl<SDep>::iterator Succ = llvm::find(N->Succs, P); 185 assert(Succ != N->Succs.end() && "Mismatching preds / succs lists!"); 186 // Update the bookkeeping. 187 if (P.getKind() == SDep::Data) { 188 assert(NumPreds > 0 && "NumPreds will underflow!"); 189 assert(N->NumSuccs > 0 && "NumSuccs will underflow!"); 190 --NumPreds; 191 --N->NumSuccs; 192 } 193 if (!N->isScheduled) { 194 if (D.isWeak()) { 195 assert(WeakPredsLeft > 0 && "WeakPredsLeft will underflow!"); 196 --WeakPredsLeft; 197 } else { 198 assert(NumPredsLeft > 0 && "NumPredsLeft will underflow!"); 199 --NumPredsLeft; 200 } 201 } 202 if (!isScheduled) { 203 if (D.isWeak()) { 204 assert(WeakSuccsLeft > 0 && "WeakSuccsLeft will underflow!"); 205 --N->WeakSuccsLeft; 206 } else { 207 assert(N->NumSuccsLeft > 0 && "NumSuccsLeft will underflow!"); 208 --N->NumSuccsLeft; 209 } 210 } 211 N->Succs.erase(Succ); 212 Preds.erase(I); 213 if (P.getLatency() != 0) { 214 this->setDepthDirty(); 215 N->setHeightDirty(); 216 } 217 } 218 219 void SUnit::setDepthDirty() { 220 if (!isDepthCurrent) return; 221 SmallVector<SUnit*, 8> WorkList; 222 WorkList.push_back(this); 223 do { 224 SUnit *SU = WorkList.pop_back_val(); 225 SU->isDepthCurrent = false; 226 for (SDep &SuccDep : SU->Succs) { 227 SUnit *SuccSU = SuccDep.getSUnit(); 228 if (SuccSU->isDepthCurrent) 229 WorkList.push_back(SuccSU); 230 } 231 } while (!WorkList.empty()); 232 } 233 234 void SUnit::setHeightDirty() { 235 if (!isHeightCurrent) return; 236 SmallVector<SUnit*, 8> WorkList; 237 WorkList.push_back(this); 238 do { 239 SUnit *SU = WorkList.pop_back_val(); 240 SU->isHeightCurrent = false; 241 for (SDep &PredDep : SU->Preds) { 242 SUnit *PredSU = PredDep.getSUnit(); 243 if (PredSU->isHeightCurrent) 244 WorkList.push_back(PredSU); 245 } 246 } while (!WorkList.empty()); 247 } 248 249 void SUnit::setDepthToAtLeast(unsigned NewDepth) { 250 if (NewDepth <= getDepth()) 251 return; 252 setDepthDirty(); 253 Depth = NewDepth; 254 isDepthCurrent = true; 255 } 256 257 void SUnit::setHeightToAtLeast(unsigned NewHeight) { 258 if (NewHeight <= getHeight()) 259 return; 260 setHeightDirty(); 261 Height = NewHeight; 262 isHeightCurrent = true; 263 } 264 265 /// Calculates the maximal path from the node to the exit. 266 void SUnit::ComputeDepth() { 267 SmallVector<SUnit*, 8> WorkList; 268 WorkList.push_back(this); 269 do { 270 SUnit *Cur = WorkList.back(); 271 272 bool Done = true; 273 unsigned MaxPredDepth = 0; 274 for (const SDep &PredDep : Cur->Preds) { 275 SUnit *PredSU = PredDep.getSUnit(); 276 if (PredSU->isDepthCurrent) 277 MaxPredDepth = std::max(MaxPredDepth, 278 PredSU->Depth + PredDep.getLatency()); 279 else { 280 Done = false; 281 WorkList.push_back(PredSU); 282 } 283 } 284 285 if (Done) { 286 WorkList.pop_back(); 287 if (MaxPredDepth != Cur->Depth) { 288 Cur->setDepthDirty(); 289 Cur->Depth = MaxPredDepth; 290 } 291 Cur->isDepthCurrent = true; 292 } 293 } while (!WorkList.empty()); 294 } 295 296 /// Calculates the maximal path from the node to the entry. 297 void SUnit::ComputeHeight() { 298 SmallVector<SUnit*, 8> WorkList; 299 WorkList.push_back(this); 300 do { 301 SUnit *Cur = WorkList.back(); 302 303 bool Done = true; 304 unsigned MaxSuccHeight = 0; 305 for (const SDep &SuccDep : Cur->Succs) { 306 SUnit *SuccSU = SuccDep.getSUnit(); 307 if (SuccSU->isHeightCurrent) 308 MaxSuccHeight = std::max(MaxSuccHeight, 309 SuccSU->Height + SuccDep.getLatency()); 310 else { 311 Done = false; 312 WorkList.push_back(SuccSU); 313 } 314 } 315 316 if (Done) { 317 WorkList.pop_back(); 318 if (MaxSuccHeight != Cur->Height) { 319 Cur->setHeightDirty(); 320 Cur->Height = MaxSuccHeight; 321 } 322 Cur->isHeightCurrent = true; 323 } 324 } while (!WorkList.empty()); 325 } 326 327 void SUnit::biasCriticalPath() { 328 if (NumPreds < 2) 329 return; 330 331 SUnit::pred_iterator BestI = Preds.begin(); 332 unsigned MaxDepth = BestI->getSUnit()->getDepth(); 333 for (SUnit::pred_iterator I = std::next(BestI), E = Preds.end(); I != E; 334 ++I) { 335 if (I->getKind() == SDep::Data && I->getSUnit()->getDepth() > MaxDepth) 336 BestI = I; 337 } 338 if (BestI != Preds.begin()) 339 std::swap(*Preds.begin(), *BestI); 340 } 341 342 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP) 343 LLVM_DUMP_METHOD void SUnit::dumpAttributes() const { 344 dbgs() << " # preds left : " << NumPredsLeft << "\n"; 345 dbgs() << " # succs left : " << NumSuccsLeft << "\n"; 346 if (WeakPredsLeft) 347 dbgs() << " # weak preds left : " << WeakPredsLeft << "\n"; 348 if (WeakSuccsLeft) 349 dbgs() << " # weak succs left : " << WeakSuccsLeft << "\n"; 350 dbgs() << " # rdefs left : " << NumRegDefsLeft << "\n"; 351 dbgs() << " Latency : " << Latency << "\n"; 352 dbgs() << " Depth : " << getDepth() << "\n"; 353 dbgs() << " Height : " << getHeight() << "\n"; 354 } 355 356 LLVM_DUMP_METHOD void ScheduleDAG::dumpNodeName(const SUnit &SU) const { 357 if (&SU == &EntrySU) 358 dbgs() << "EntrySU"; 359 else if (&SU == &ExitSU) 360 dbgs() << "ExitSU"; 361 else 362 dbgs() << "SU(" << SU.NodeNum << ")"; 363 } 364 365 LLVM_DUMP_METHOD void ScheduleDAG::dumpNodeAll(const SUnit &SU) const { 366 dumpNode(SU); 367 SU.dumpAttributes(); 368 if (SU.Preds.size() > 0) { 369 dbgs() << " Predecessors:\n"; 370 for (const SDep &Dep : SU.Preds) { 371 dbgs() << " "; 372 dumpNodeName(*Dep.getSUnit()); 373 dbgs() << ": "; 374 Dep.dump(TRI); 375 dbgs() << '\n'; 376 } 377 } 378 if (SU.Succs.size() > 0) { 379 dbgs() << " Successors:\n"; 380 for (const SDep &Dep : SU.Succs) { 381 dbgs() << " "; 382 dumpNodeName(*Dep.getSUnit()); 383 dbgs() << ": "; 384 Dep.dump(TRI); 385 dbgs() << '\n'; 386 } 387 } 388 } 389 #endif 390 391 #ifndef NDEBUG 392 unsigned ScheduleDAG::VerifyScheduledDAG(bool isBottomUp) { 393 bool AnyNotSched = false; 394 unsigned DeadNodes = 0; 395 for (const SUnit &SUnit : SUnits) { 396 if (!SUnit.isScheduled) { 397 if (SUnit.NumPreds == 0 && SUnit.NumSuccs == 0) { 398 ++DeadNodes; 399 continue; 400 } 401 if (!AnyNotSched) 402 dbgs() << "*** Scheduling failed! ***\n"; 403 dumpNode(SUnit); 404 dbgs() << "has not been scheduled!\n"; 405 AnyNotSched = true; 406 } 407 if (SUnit.isScheduled && 408 (isBottomUp ? SUnit.getHeight() : SUnit.getDepth()) > 409 unsigned(std::numeric_limits<int>::max())) { 410 if (!AnyNotSched) 411 dbgs() << "*** Scheduling failed! ***\n"; 412 dumpNode(SUnit); 413 dbgs() << "has an unexpected " 414 << (isBottomUp ? "Height" : "Depth") << " value!\n"; 415 AnyNotSched = true; 416 } 417 if (isBottomUp) { 418 if (SUnit.NumSuccsLeft != 0) { 419 if (!AnyNotSched) 420 dbgs() << "*** Scheduling failed! ***\n"; 421 dumpNode(SUnit); 422 dbgs() << "has successors left!\n"; 423 AnyNotSched = true; 424 } 425 } else { 426 if (SUnit.NumPredsLeft != 0) { 427 if (!AnyNotSched) 428 dbgs() << "*** Scheduling failed! ***\n"; 429 dumpNode(SUnit); 430 dbgs() << "has predecessors left!\n"; 431 AnyNotSched = true; 432 } 433 } 434 } 435 assert(!AnyNotSched); 436 return SUnits.size() - DeadNodes; 437 } 438 #endif 439 440 void ScheduleDAGTopologicalSort::InitDAGTopologicalSorting() { 441 // The idea of the algorithm is taken from 442 // "Online algorithms for managing the topological order of 443 // a directed acyclic graph" by David J. Pearce and Paul H.J. Kelly 444 // This is the MNR algorithm, which was first introduced by 445 // A. Marchetti-Spaccamela, U. Nanni and H. Rohnert in 446 // "Maintaining a topological order under edge insertions". 447 // 448 // Short description of the algorithm: 449 // 450 // Topological ordering, ord, of a DAG maps each node to a topological 451 // index so that for all edges X->Y it is the case that ord(X) < ord(Y). 452 // 453 // This means that if there is a path from the node X to the node Z, 454 // then ord(X) < ord(Z). 455 // 456 // This property can be used to check for reachability of nodes: 457 // if Z is reachable from X, then an insertion of the edge Z->X would 458 // create a cycle. 459 // 460 // The algorithm first computes a topological ordering for the DAG by 461 // initializing the Index2Node and Node2Index arrays and then tries to keep 462 // the ordering up-to-date after edge insertions by reordering the DAG. 463 // 464 // On insertion of the edge X->Y, the algorithm first marks by calling DFS 465 // the nodes reachable from Y, and then shifts them using Shift to lie 466 // immediately after X in Index2Node. 467 468 // Cancel pending updates, mark as valid. 469 Dirty = false; 470 Updates.clear(); 471 472 unsigned DAGSize = SUnits.size(); 473 std::vector<SUnit*> WorkList; 474 WorkList.reserve(DAGSize); 475 476 Index2Node.resize(DAGSize); 477 Node2Index.resize(DAGSize); 478 479 // Initialize the data structures. 480 if (ExitSU) 481 WorkList.push_back(ExitSU); 482 for (SUnit &SU : SUnits) { 483 int NodeNum = SU.NodeNum; 484 unsigned Degree = SU.Succs.size(); 485 // Temporarily use the Node2Index array as scratch space for degree counts. 486 Node2Index[NodeNum] = Degree; 487 488 // Is it a node without dependencies? 489 if (Degree == 0) { 490 assert(SU.Succs.empty() && "SUnit should have no successors"); 491 // Collect leaf nodes. 492 WorkList.push_back(&SU); 493 } 494 } 495 496 int Id = DAGSize; 497 while (!WorkList.empty()) { 498 SUnit *SU = WorkList.back(); 499 WorkList.pop_back(); 500 if (SU->NodeNum < DAGSize) 501 Allocate(SU->NodeNum, --Id); 502 for (const SDep &PredDep : SU->Preds) { 503 SUnit *SU = PredDep.getSUnit(); 504 if (SU->NodeNum < DAGSize && !--Node2Index[SU->NodeNum]) 505 // If all dependencies of the node are processed already, 506 // then the node can be computed now. 507 WorkList.push_back(SU); 508 } 509 } 510 511 Visited.resize(DAGSize); 512 NumTopoInits++; 513 514 #ifndef NDEBUG 515 // Check correctness of the ordering 516 for (SUnit &SU : SUnits) { 517 for (const SDep &PD : SU.Preds) { 518 assert(Node2Index[SU.NodeNum] > Node2Index[PD.getSUnit()->NodeNum] && 519 "Wrong topological sorting"); 520 } 521 } 522 #endif 523 } 524 525 void ScheduleDAGTopologicalSort::FixOrder() { 526 // Recompute from scratch after new nodes have been added. 527 if (Dirty) { 528 InitDAGTopologicalSorting(); 529 return; 530 } 531 532 // Otherwise apply updates one-by-one. 533 for (auto &U : Updates) 534 AddPred(U.first, U.second); 535 Updates.clear(); 536 } 537 538 void ScheduleDAGTopologicalSort::AddPredQueued(SUnit *Y, SUnit *X) { 539 // Recomputing the order from scratch is likely more efficient than applying 540 // updates one-by-one for too many updates. The current cut-off is arbitrarily 541 // chosen. 542 Dirty = Dirty || Updates.size() > 10; 543 544 if (Dirty) 545 return; 546 547 Updates.emplace_back(Y, X); 548 } 549 550 void ScheduleDAGTopologicalSort::AddPred(SUnit *Y, SUnit *X) { 551 int UpperBound, LowerBound; 552 LowerBound = Node2Index[Y->NodeNum]; 553 UpperBound = Node2Index[X->NodeNum]; 554 bool HasLoop = false; 555 // Is Ord(X) < Ord(Y) ? 556 if (LowerBound < UpperBound) { 557 // Update the topological order. 558 Visited.reset(); 559 DFS(Y, UpperBound, HasLoop); 560 assert(!HasLoop && "Inserted edge creates a loop!"); 561 // Recompute topological indexes. 562 Shift(Visited, LowerBound, UpperBound); 563 } 564 565 NumNewPredsAdded++; 566 } 567 568 void ScheduleDAGTopologicalSort::RemovePred(SUnit *M, SUnit *N) { 569 // InitDAGTopologicalSorting(); 570 } 571 572 void ScheduleDAGTopologicalSort::DFS(const SUnit *SU, int UpperBound, 573 bool &HasLoop) { 574 std::vector<const SUnit*> WorkList; 575 WorkList.reserve(SUnits.size()); 576 577 WorkList.push_back(SU); 578 do { 579 SU = WorkList.back(); 580 WorkList.pop_back(); 581 Visited.set(SU->NodeNum); 582 for (const SDep &SuccDep : llvm::reverse(SU->Succs)) { 583 unsigned s = SuccDep.getSUnit()->NodeNum; 584 // Edges to non-SUnits are allowed but ignored (e.g. ExitSU). 585 if (s >= Node2Index.size()) 586 continue; 587 if (Node2Index[s] == UpperBound) { 588 HasLoop = true; 589 return; 590 } 591 // Visit successors if not already and in affected region. 592 if (!Visited.test(s) && Node2Index[s] < UpperBound) { 593 WorkList.push_back(SuccDep.getSUnit()); 594 } 595 } 596 } while (!WorkList.empty()); 597 } 598 599 std::vector<int> ScheduleDAGTopologicalSort::GetSubGraph(const SUnit &StartSU, 600 const SUnit &TargetSU, 601 bool &Success) { 602 std::vector<const SUnit*> WorkList; 603 int LowerBound = Node2Index[StartSU.NodeNum]; 604 int UpperBound = Node2Index[TargetSU.NodeNum]; 605 bool Found = false; 606 BitVector VisitedBack; 607 std::vector<int> Nodes; 608 609 if (LowerBound > UpperBound) { 610 Success = false; 611 return Nodes; 612 } 613 614 WorkList.reserve(SUnits.size()); 615 Visited.reset(); 616 617 // Starting from StartSU, visit all successors up 618 // to UpperBound. 619 WorkList.push_back(&StartSU); 620 do { 621 const SUnit *SU = WorkList.back(); 622 WorkList.pop_back(); 623 for (const SDep &SD : llvm::reverse(SU->Succs)) { 624 const SUnit *Succ = SD.getSUnit(); 625 unsigned s = Succ->NodeNum; 626 // Edges to non-SUnits are allowed but ignored (e.g. ExitSU). 627 if (Succ->isBoundaryNode()) 628 continue; 629 if (Node2Index[s] == UpperBound) { 630 Found = true; 631 continue; 632 } 633 // Visit successors if not already and in affected region. 634 if (!Visited.test(s) && Node2Index[s] < UpperBound) { 635 Visited.set(s); 636 WorkList.push_back(Succ); 637 } 638 } 639 } while (!WorkList.empty()); 640 641 if (!Found) { 642 Success = false; 643 return Nodes; 644 } 645 646 WorkList.clear(); 647 VisitedBack.resize(SUnits.size()); 648 Found = false; 649 650 // Starting from TargetSU, visit all predecessors up 651 // to LowerBound. SUs that are visited by the two 652 // passes are added to Nodes. 653 WorkList.push_back(&TargetSU); 654 do { 655 const SUnit *SU = WorkList.back(); 656 WorkList.pop_back(); 657 for (const SDep &SD : llvm::reverse(SU->Preds)) { 658 const SUnit *Pred = SD.getSUnit(); 659 unsigned s = Pred->NodeNum; 660 // Edges to non-SUnits are allowed but ignored (e.g. EntrySU). 661 if (Pred->isBoundaryNode()) 662 continue; 663 if (Node2Index[s] == LowerBound) { 664 Found = true; 665 continue; 666 } 667 if (!VisitedBack.test(s) && Visited.test(s)) { 668 VisitedBack.set(s); 669 WorkList.push_back(Pred); 670 Nodes.push_back(s); 671 } 672 } 673 } while (!WorkList.empty()); 674 675 assert(Found && "Error in SUnit Graph!"); 676 Success = true; 677 return Nodes; 678 } 679 680 void ScheduleDAGTopologicalSort::Shift(BitVector& Visited, int LowerBound, 681 int UpperBound) { 682 std::vector<int> L; 683 int shift = 0; 684 int i; 685 686 for (i = LowerBound; i <= UpperBound; ++i) { 687 // w is node at topological index i. 688 int w = Index2Node[i]; 689 if (Visited.test(w)) { 690 // Unmark. 691 Visited.reset(w); 692 L.push_back(w); 693 shift = shift + 1; 694 } else { 695 Allocate(w, i - shift); 696 } 697 } 698 699 for (unsigned LI : L) { 700 Allocate(LI, i - shift); 701 i = i + 1; 702 } 703 } 704 705 bool ScheduleDAGTopologicalSort::WillCreateCycle(SUnit *TargetSU, SUnit *SU) { 706 FixOrder(); 707 // Is SU reachable from TargetSU via successor edges? 708 if (IsReachable(SU, TargetSU)) 709 return true; 710 for (const SDep &PredDep : TargetSU->Preds) 711 if (PredDep.isAssignedRegDep() && 712 IsReachable(SU, PredDep.getSUnit())) 713 return true; 714 return false; 715 } 716 717 void ScheduleDAGTopologicalSort::AddSUnitWithoutPredecessors(const SUnit *SU) { 718 assert(SU->NodeNum == Index2Node.size() && "Node cannot be added at the end"); 719 assert(SU->NumPreds == 0 && "Can only add SU's with no predecessors"); 720 Node2Index.push_back(Index2Node.size()); 721 Index2Node.push_back(SU->NodeNum); 722 Visited.resize(Node2Index.size()); 723 } 724 725 bool ScheduleDAGTopologicalSort::IsReachable(const SUnit *SU, 726 const SUnit *TargetSU) { 727 assert(TargetSU != nullptr && "Invalid target SUnit"); 728 assert(SU != nullptr && "Invalid SUnit"); 729 FixOrder(); 730 // If insertion of the edge SU->TargetSU would create a cycle 731 // then there is a path from TargetSU to SU. 732 int UpperBound, LowerBound; 733 LowerBound = Node2Index[TargetSU->NodeNum]; 734 UpperBound = Node2Index[SU->NodeNum]; 735 bool HasLoop = false; 736 // Is Ord(TargetSU) < Ord(SU) ? 737 if (LowerBound < UpperBound) { 738 Visited.reset(); 739 // There may be a path from TargetSU to SU. Check for it. 740 DFS(TargetSU, UpperBound, HasLoop); 741 } 742 return HasLoop; 743 } 744 745 void ScheduleDAGTopologicalSort::Allocate(int n, int index) { 746 Node2Index[n] = index; 747 Index2Node[index] = n; 748 } 749 750 ScheduleDAGTopologicalSort:: 751 ScheduleDAGTopologicalSort(std::vector<SUnit> &sunits, SUnit *exitsu) 752 : SUnits(sunits), ExitSU(exitsu) {} 753 754 ScheduleHazardRecognizer::~ScheduleHazardRecognizer() = default; 755