1 //===- MachineSink.cpp - Sinking for machine instructions -----------------===// 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 // This pass moves instructions into successor blocks when possible, so that 10 // they aren't executed on paths where their results aren't needed. 11 // 12 // This pass is not intended to be a replacement or a complete alternative 13 // for an LLVM-IR-level sinking pass. It is only designed to sink simple 14 // constructs that are not exposed before lowering and instruction selection. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm/ADT/SetVector.h" 19 #include "llvm/ADT/SmallSet.h" 20 #include "llvm/ADT/SmallVector.h" 21 #include "llvm/ADT/SparseBitVector.h" 22 #include "llvm/ADT/Statistic.h" 23 #include "llvm/Analysis/AliasAnalysis.h" 24 #include "llvm/CodeGen/MachineBasicBlock.h" 25 #include "llvm/CodeGen/MachineBlockFrequencyInfo.h" 26 #include "llvm/CodeGen/MachineBranchProbabilityInfo.h" 27 #include "llvm/CodeGen/MachineDominators.h" 28 #include "llvm/CodeGen/MachineFunction.h" 29 #include "llvm/CodeGen/MachineFunctionPass.h" 30 #include "llvm/CodeGen/MachineInstr.h" 31 #include "llvm/CodeGen/MachineLoopInfo.h" 32 #include "llvm/CodeGen/MachineOperand.h" 33 #include "llvm/CodeGen/MachinePostDominators.h" 34 #include "llvm/CodeGen/MachineRegisterInfo.h" 35 #include "llvm/CodeGen/TargetInstrInfo.h" 36 #include "llvm/CodeGen/TargetRegisterInfo.h" 37 #include "llvm/CodeGen/TargetSubtargetInfo.h" 38 #include "llvm/IR/BasicBlock.h" 39 #include "llvm/IR/LLVMContext.h" 40 #include "llvm/IR/DebugInfoMetadata.h" 41 #include "llvm/Pass.h" 42 #include "llvm/Support/BranchProbability.h" 43 #include "llvm/Support/CommandLine.h" 44 #include "llvm/Support/Debug.h" 45 #include "llvm/Support/raw_ostream.h" 46 #include <algorithm> 47 #include <cassert> 48 #include <cstdint> 49 #include <map> 50 #include <utility> 51 #include <vector> 52 53 using namespace llvm; 54 55 #define DEBUG_TYPE "machine-sink" 56 57 static cl::opt<bool> 58 SplitEdges("machine-sink-split", 59 cl::desc("Split critical edges during machine sinking"), 60 cl::init(true), cl::Hidden); 61 62 static cl::opt<bool> 63 UseBlockFreqInfo("machine-sink-bfi", 64 cl::desc("Use block frequency info to find successors to sink"), 65 cl::init(true), cl::Hidden); 66 67 static cl::opt<unsigned> SplitEdgeProbabilityThreshold( 68 "machine-sink-split-probability-threshold", 69 cl::desc( 70 "Percentage threshold for splitting single-instruction critical edge. " 71 "If the branch threshold is higher than this threshold, we allow " 72 "speculative execution of up to 1 instruction to avoid branching to " 73 "splitted critical edge"), 74 cl::init(40), cl::Hidden); 75 76 STATISTIC(NumSunk, "Number of machine instructions sunk"); 77 STATISTIC(NumSplit, "Number of critical edges split"); 78 STATISTIC(NumCoalesces, "Number of copies coalesced"); 79 STATISTIC(NumPostRACopySink, "Number of copies sunk after RA"); 80 81 namespace { 82 83 class MachineSinking : public MachineFunctionPass { 84 const TargetInstrInfo *TII; 85 const TargetRegisterInfo *TRI; 86 MachineRegisterInfo *MRI; // Machine register information 87 MachineDominatorTree *DT; // Machine dominator tree 88 MachinePostDominatorTree *PDT; // Machine post dominator tree 89 MachineLoopInfo *LI; 90 const MachineBlockFrequencyInfo *MBFI; 91 const MachineBranchProbabilityInfo *MBPI; 92 AliasAnalysis *AA; 93 94 // Remember which edges have been considered for breaking. 95 SmallSet<std::pair<MachineBasicBlock*, MachineBasicBlock*>, 8> 96 CEBCandidates; 97 // Remember which edges we are about to split. 98 // This is different from CEBCandidates since those edges 99 // will be split. 100 SetVector<std::pair<MachineBasicBlock *, MachineBasicBlock *>> ToSplit; 101 102 SparseBitVector<> RegsToClearKillFlags; 103 104 using AllSuccsCache = 105 std::map<MachineBasicBlock *, SmallVector<MachineBasicBlock *, 4>>; 106 107 public: 108 static char ID; // Pass identification 109 110 MachineSinking() : MachineFunctionPass(ID) { 111 initializeMachineSinkingPass(*PassRegistry::getPassRegistry()); 112 } 113 114 bool runOnMachineFunction(MachineFunction &MF) override; 115 116 void getAnalysisUsage(AnalysisUsage &AU) const override { 117 AU.setPreservesCFG(); 118 MachineFunctionPass::getAnalysisUsage(AU); 119 AU.addRequired<AAResultsWrapperPass>(); 120 AU.addRequired<MachineDominatorTree>(); 121 AU.addRequired<MachinePostDominatorTree>(); 122 AU.addRequired<MachineLoopInfo>(); 123 AU.addRequired<MachineBranchProbabilityInfo>(); 124 AU.addPreserved<MachineDominatorTree>(); 125 AU.addPreserved<MachinePostDominatorTree>(); 126 AU.addPreserved<MachineLoopInfo>(); 127 if (UseBlockFreqInfo) 128 AU.addRequired<MachineBlockFrequencyInfo>(); 129 } 130 131 void releaseMemory() override { 132 CEBCandidates.clear(); 133 } 134 135 private: 136 bool ProcessBlock(MachineBasicBlock &MBB); 137 bool isWorthBreakingCriticalEdge(MachineInstr &MI, 138 MachineBasicBlock *From, 139 MachineBasicBlock *To); 140 141 /// Postpone the splitting of the given critical 142 /// edge (\p From, \p To). 143 /// 144 /// We do not split the edges on the fly. Indeed, this invalidates 145 /// the dominance information and thus triggers a lot of updates 146 /// of that information underneath. 147 /// Instead, we postpone all the splits after each iteration of 148 /// the main loop. That way, the information is at least valid 149 /// for the lifetime of an iteration. 150 /// 151 /// \return True if the edge is marked as toSplit, false otherwise. 152 /// False can be returned if, for instance, this is not profitable. 153 bool PostponeSplitCriticalEdge(MachineInstr &MI, 154 MachineBasicBlock *From, 155 MachineBasicBlock *To, 156 bool BreakPHIEdge); 157 bool SinkInstruction(MachineInstr &MI, bool &SawStore, 158 159 AllSuccsCache &AllSuccessors); 160 bool AllUsesDominatedByBlock(unsigned Reg, MachineBasicBlock *MBB, 161 MachineBasicBlock *DefMBB, 162 bool &BreakPHIEdge, bool &LocalUse) const; 163 MachineBasicBlock *FindSuccToSinkTo(MachineInstr &MI, MachineBasicBlock *MBB, 164 bool &BreakPHIEdge, AllSuccsCache &AllSuccessors); 165 bool isProfitableToSinkTo(unsigned Reg, MachineInstr &MI, 166 MachineBasicBlock *MBB, 167 MachineBasicBlock *SuccToSinkTo, 168 AllSuccsCache &AllSuccessors); 169 170 bool PerformTrivialForwardCoalescing(MachineInstr &MI, 171 MachineBasicBlock *MBB); 172 173 SmallVector<MachineBasicBlock *, 4> & 174 GetAllSortedSuccessors(MachineInstr &MI, MachineBasicBlock *MBB, 175 AllSuccsCache &AllSuccessors) const; 176 }; 177 178 } // end anonymous namespace 179 180 char MachineSinking::ID = 0; 181 182 char &llvm::MachineSinkingID = MachineSinking::ID; 183 184 INITIALIZE_PASS_BEGIN(MachineSinking, DEBUG_TYPE, 185 "Machine code sinking", false, false) 186 INITIALIZE_PASS_DEPENDENCY(MachineBranchProbabilityInfo) 187 INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree) 188 INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo) 189 INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass) 190 INITIALIZE_PASS_END(MachineSinking, DEBUG_TYPE, 191 "Machine code sinking", false, false) 192 193 bool MachineSinking::PerformTrivialForwardCoalescing(MachineInstr &MI, 194 MachineBasicBlock *MBB) { 195 if (!MI.isCopy()) 196 return false; 197 198 unsigned SrcReg = MI.getOperand(1).getReg(); 199 unsigned DstReg = MI.getOperand(0).getReg(); 200 if (!TargetRegisterInfo::isVirtualRegister(SrcReg) || 201 !TargetRegisterInfo::isVirtualRegister(DstReg) || 202 !MRI->hasOneNonDBGUse(SrcReg)) 203 return false; 204 205 const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg); 206 const TargetRegisterClass *DRC = MRI->getRegClass(DstReg); 207 if (SRC != DRC) 208 return false; 209 210 MachineInstr *DefMI = MRI->getVRegDef(SrcReg); 211 if (DefMI->isCopyLike()) 212 return false; 213 LLVM_DEBUG(dbgs() << "Coalescing: " << *DefMI); 214 LLVM_DEBUG(dbgs() << "*** to: " << MI); 215 MRI->replaceRegWith(DstReg, SrcReg); 216 MI.eraseFromParent(); 217 218 // Conservatively, clear any kill flags, since it's possible that they are no 219 // longer correct. 220 MRI->clearKillFlags(SrcReg); 221 222 ++NumCoalesces; 223 return true; 224 } 225 226 /// AllUsesDominatedByBlock - Return true if all uses of the specified register 227 /// occur in blocks dominated by the specified block. If any use is in the 228 /// definition block, then return false since it is never legal to move def 229 /// after uses. 230 bool 231 MachineSinking::AllUsesDominatedByBlock(unsigned Reg, 232 MachineBasicBlock *MBB, 233 MachineBasicBlock *DefMBB, 234 bool &BreakPHIEdge, 235 bool &LocalUse) const { 236 assert(TargetRegisterInfo::isVirtualRegister(Reg) && 237 "Only makes sense for vregs"); 238 239 // Ignore debug uses because debug info doesn't affect the code. 240 if (MRI->use_nodbg_empty(Reg)) 241 return true; 242 243 // BreakPHIEdge is true if all the uses are in the successor MBB being sunken 244 // into and they are all PHI nodes. In this case, machine-sink must break 245 // the critical edge first. e.g. 246 // 247 // %bb.1: derived from LLVM BB %bb4.preheader 248 // Predecessors according to CFG: %bb.0 249 // ... 250 // %reg16385 = DEC64_32r %reg16437, implicit-def dead %eflags 251 // ... 252 // JE_4 <%bb.37>, implicit %eflags 253 // Successors according to CFG: %bb.37 %bb.2 254 // 255 // %bb.2: derived from LLVM BB %bb.nph 256 // Predecessors according to CFG: %bb.0 %bb.1 257 // %reg16386 = PHI %reg16434, %bb.0, %reg16385, %bb.1 258 BreakPHIEdge = true; 259 for (MachineOperand &MO : MRI->use_nodbg_operands(Reg)) { 260 MachineInstr *UseInst = MO.getParent(); 261 unsigned OpNo = &MO - &UseInst->getOperand(0); 262 MachineBasicBlock *UseBlock = UseInst->getParent(); 263 if (!(UseBlock == MBB && UseInst->isPHI() && 264 UseInst->getOperand(OpNo+1).getMBB() == DefMBB)) { 265 BreakPHIEdge = false; 266 break; 267 } 268 } 269 if (BreakPHIEdge) 270 return true; 271 272 for (MachineOperand &MO : MRI->use_nodbg_operands(Reg)) { 273 // Determine the block of the use. 274 MachineInstr *UseInst = MO.getParent(); 275 unsigned OpNo = &MO - &UseInst->getOperand(0); 276 MachineBasicBlock *UseBlock = UseInst->getParent(); 277 if (UseInst->isPHI()) { 278 // PHI nodes use the operand in the predecessor block, not the block with 279 // the PHI. 280 UseBlock = UseInst->getOperand(OpNo+1).getMBB(); 281 } else if (UseBlock == DefMBB) { 282 LocalUse = true; 283 return false; 284 } 285 286 // Check that it dominates. 287 if (!DT->dominates(MBB, UseBlock)) 288 return false; 289 } 290 291 return true; 292 } 293 294 bool MachineSinking::runOnMachineFunction(MachineFunction &MF) { 295 if (skipFunction(MF.getFunction())) 296 return false; 297 298 LLVM_DEBUG(dbgs() << "******** Machine Sinking ********\n"); 299 300 TII = MF.getSubtarget().getInstrInfo(); 301 TRI = MF.getSubtarget().getRegisterInfo(); 302 MRI = &MF.getRegInfo(); 303 DT = &getAnalysis<MachineDominatorTree>(); 304 PDT = &getAnalysis<MachinePostDominatorTree>(); 305 LI = &getAnalysis<MachineLoopInfo>(); 306 MBFI = UseBlockFreqInfo ? &getAnalysis<MachineBlockFrequencyInfo>() : nullptr; 307 MBPI = &getAnalysis<MachineBranchProbabilityInfo>(); 308 AA = &getAnalysis<AAResultsWrapperPass>().getAAResults(); 309 310 bool EverMadeChange = false; 311 312 while (true) { 313 bool MadeChange = false; 314 315 // Process all basic blocks. 316 CEBCandidates.clear(); 317 ToSplit.clear(); 318 for (auto &MBB: MF) 319 MadeChange |= ProcessBlock(MBB); 320 321 // If we have anything we marked as toSplit, split it now. 322 for (auto &Pair : ToSplit) { 323 auto NewSucc = Pair.first->SplitCriticalEdge(Pair.second, *this); 324 if (NewSucc != nullptr) { 325 LLVM_DEBUG(dbgs() << " *** Splitting critical edge: " 326 << printMBBReference(*Pair.first) << " -- " 327 << printMBBReference(*NewSucc) << " -- " 328 << printMBBReference(*Pair.second) << '\n'); 329 MadeChange = true; 330 ++NumSplit; 331 } else 332 LLVM_DEBUG(dbgs() << " *** Not legal to break critical edge\n"); 333 } 334 // If this iteration over the code changed anything, keep iterating. 335 if (!MadeChange) break; 336 EverMadeChange = true; 337 } 338 339 // Now clear any kill flags for recorded registers. 340 for (auto I : RegsToClearKillFlags) 341 MRI->clearKillFlags(I); 342 RegsToClearKillFlags.clear(); 343 344 return EverMadeChange; 345 } 346 347 bool MachineSinking::ProcessBlock(MachineBasicBlock &MBB) { 348 // Can't sink anything out of a block that has less than two successors. 349 if (MBB.succ_size() <= 1 || MBB.empty()) return false; 350 351 // Don't bother sinking code out of unreachable blocks. In addition to being 352 // unprofitable, it can also lead to infinite looping, because in an 353 // unreachable loop there may be nowhere to stop. 354 if (!DT->isReachableFromEntry(&MBB)) return false; 355 356 bool MadeChange = false; 357 358 // Cache all successors, sorted by frequency info and loop depth. 359 AllSuccsCache AllSuccessors; 360 361 // Walk the basic block bottom-up. Remember if we saw a store. 362 MachineBasicBlock::iterator I = MBB.end(); 363 --I; 364 bool ProcessedBegin, SawStore = false; 365 do { 366 MachineInstr &MI = *I; // The instruction to sink. 367 368 // Predecrement I (if it's not begin) so that it isn't invalidated by 369 // sinking. 370 ProcessedBegin = I == MBB.begin(); 371 if (!ProcessedBegin) 372 --I; 373 374 if (MI.isDebugInstr()) 375 continue; 376 377 bool Joined = PerformTrivialForwardCoalescing(MI, &MBB); 378 if (Joined) { 379 MadeChange = true; 380 continue; 381 } 382 383 if (SinkInstruction(MI, SawStore, AllSuccessors)) { 384 ++NumSunk; 385 MadeChange = true; 386 } 387 388 // If we just processed the first instruction in the block, we're done. 389 } while (!ProcessedBegin); 390 391 return MadeChange; 392 } 393 394 bool MachineSinking::isWorthBreakingCriticalEdge(MachineInstr &MI, 395 MachineBasicBlock *From, 396 MachineBasicBlock *To) { 397 // FIXME: Need much better heuristics. 398 399 // If the pass has already considered breaking this edge (during this pass 400 // through the function), then let's go ahead and break it. This means 401 // sinking multiple "cheap" instructions into the same block. 402 if (!CEBCandidates.insert(std::make_pair(From, To)).second) 403 return true; 404 405 if (!MI.isCopy() && !TII->isAsCheapAsAMove(MI)) 406 return true; 407 408 if (From->isSuccessor(To) && MBPI->getEdgeProbability(From, To) <= 409 BranchProbability(SplitEdgeProbabilityThreshold, 100)) 410 return true; 411 412 // MI is cheap, we probably don't want to break the critical edge for it. 413 // However, if this would allow some definitions of its source operands 414 // to be sunk then it's probably worth it. 415 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) { 416 const MachineOperand &MO = MI.getOperand(i); 417 if (!MO.isReg() || !MO.isUse()) 418 continue; 419 unsigned Reg = MO.getReg(); 420 if (Reg == 0) 421 continue; 422 423 // We don't move live definitions of physical registers, 424 // so sinking their uses won't enable any opportunities. 425 if (TargetRegisterInfo::isPhysicalRegister(Reg)) 426 continue; 427 428 // If this instruction is the only user of a virtual register, 429 // check if breaking the edge will enable sinking 430 // both this instruction and the defining instruction. 431 if (MRI->hasOneNonDBGUse(Reg)) { 432 // If the definition resides in same MBB, 433 // claim it's likely we can sink these together. 434 // If definition resides elsewhere, we aren't 435 // blocking it from being sunk so don't break the edge. 436 MachineInstr *DefMI = MRI->getVRegDef(Reg); 437 if (DefMI->getParent() == MI.getParent()) 438 return true; 439 } 440 } 441 442 return false; 443 } 444 445 bool MachineSinking::PostponeSplitCriticalEdge(MachineInstr &MI, 446 MachineBasicBlock *FromBB, 447 MachineBasicBlock *ToBB, 448 bool BreakPHIEdge) { 449 if (!isWorthBreakingCriticalEdge(MI, FromBB, ToBB)) 450 return false; 451 452 // Avoid breaking back edge. From == To means backedge for single BB loop. 453 if (!SplitEdges || FromBB == ToBB) 454 return false; 455 456 // Check for backedges of more "complex" loops. 457 if (LI->getLoopFor(FromBB) == LI->getLoopFor(ToBB) && 458 LI->isLoopHeader(ToBB)) 459 return false; 460 461 // It's not always legal to break critical edges and sink the computation 462 // to the edge. 463 // 464 // %bb.1: 465 // v1024 466 // Beq %bb.3 467 // <fallthrough> 468 // %bb.2: 469 // ... no uses of v1024 470 // <fallthrough> 471 // %bb.3: 472 // ... 473 // = v1024 474 // 475 // If %bb.1 -> %bb.3 edge is broken and computation of v1024 is inserted: 476 // 477 // %bb.1: 478 // ... 479 // Bne %bb.2 480 // %bb.4: 481 // v1024 = 482 // B %bb.3 483 // %bb.2: 484 // ... no uses of v1024 485 // <fallthrough> 486 // %bb.3: 487 // ... 488 // = v1024 489 // 490 // This is incorrect since v1024 is not computed along the %bb.1->%bb.2->%bb.3 491 // flow. We need to ensure the new basic block where the computation is 492 // sunk to dominates all the uses. 493 // It's only legal to break critical edge and sink the computation to the 494 // new block if all the predecessors of "To", except for "From", are 495 // not dominated by "From". Given SSA property, this means these 496 // predecessors are dominated by "To". 497 // 498 // There is no need to do this check if all the uses are PHI nodes. PHI 499 // sources are only defined on the specific predecessor edges. 500 if (!BreakPHIEdge) { 501 for (MachineBasicBlock::pred_iterator PI = ToBB->pred_begin(), 502 E = ToBB->pred_end(); PI != E; ++PI) { 503 if (*PI == FromBB) 504 continue; 505 if (!DT->dominates(ToBB, *PI)) 506 return false; 507 } 508 } 509 510 ToSplit.insert(std::make_pair(FromBB, ToBB)); 511 512 return true; 513 } 514 515 /// isProfitableToSinkTo - Return true if it is profitable to sink MI. 516 bool MachineSinking::isProfitableToSinkTo(unsigned Reg, MachineInstr &MI, 517 MachineBasicBlock *MBB, 518 MachineBasicBlock *SuccToSinkTo, 519 AllSuccsCache &AllSuccessors) { 520 assert (SuccToSinkTo && "Invalid SinkTo Candidate BB"); 521 522 if (MBB == SuccToSinkTo) 523 return false; 524 525 // It is profitable if SuccToSinkTo does not post dominate current block. 526 if (!PDT->dominates(SuccToSinkTo, MBB)) 527 return true; 528 529 // It is profitable to sink an instruction from a deeper loop to a shallower 530 // loop, even if the latter post-dominates the former (PR21115). 531 if (LI->getLoopDepth(MBB) > LI->getLoopDepth(SuccToSinkTo)) 532 return true; 533 534 // Check if only use in post dominated block is PHI instruction. 535 bool NonPHIUse = false; 536 for (MachineInstr &UseInst : MRI->use_nodbg_instructions(Reg)) { 537 MachineBasicBlock *UseBlock = UseInst.getParent(); 538 if (UseBlock == SuccToSinkTo && !UseInst.isPHI()) 539 NonPHIUse = true; 540 } 541 if (!NonPHIUse) 542 return true; 543 544 // If SuccToSinkTo post dominates then also it may be profitable if MI 545 // can further profitably sinked into another block in next round. 546 bool BreakPHIEdge = false; 547 // FIXME - If finding successor is compile time expensive then cache results. 548 if (MachineBasicBlock *MBB2 = 549 FindSuccToSinkTo(MI, SuccToSinkTo, BreakPHIEdge, AllSuccessors)) 550 return isProfitableToSinkTo(Reg, MI, SuccToSinkTo, MBB2, AllSuccessors); 551 552 // If SuccToSinkTo is final destination and it is a post dominator of current 553 // block then it is not profitable to sink MI into SuccToSinkTo block. 554 return false; 555 } 556 557 /// Get the sorted sequence of successors for this MachineBasicBlock, possibly 558 /// computing it if it was not already cached. 559 SmallVector<MachineBasicBlock *, 4> & 560 MachineSinking::GetAllSortedSuccessors(MachineInstr &MI, MachineBasicBlock *MBB, 561 AllSuccsCache &AllSuccessors) const { 562 // Do we have the sorted successors in cache ? 563 auto Succs = AllSuccessors.find(MBB); 564 if (Succs != AllSuccessors.end()) 565 return Succs->second; 566 567 SmallVector<MachineBasicBlock *, 4> AllSuccs(MBB->succ_begin(), 568 MBB->succ_end()); 569 570 // Handle cases where sinking can happen but where the sink point isn't a 571 // successor. For example: 572 // 573 // x = computation 574 // if () {} else {} 575 // use x 576 // 577 const std::vector<MachineDomTreeNode *> &Children = 578 DT->getNode(MBB)->getChildren(); 579 for (const auto &DTChild : Children) 580 // DomTree children of MBB that have MBB as immediate dominator are added. 581 if (DTChild->getIDom()->getBlock() == MI.getParent() && 582 // Skip MBBs already added to the AllSuccs vector above. 583 !MBB->isSuccessor(DTChild->getBlock())) 584 AllSuccs.push_back(DTChild->getBlock()); 585 586 // Sort Successors according to their loop depth or block frequency info. 587 llvm::stable_sort( 588 AllSuccs, [this](const MachineBasicBlock *L, const MachineBasicBlock *R) { 589 uint64_t LHSFreq = MBFI ? MBFI->getBlockFreq(L).getFrequency() : 0; 590 uint64_t RHSFreq = MBFI ? MBFI->getBlockFreq(R).getFrequency() : 0; 591 bool HasBlockFreq = LHSFreq != 0 && RHSFreq != 0; 592 return HasBlockFreq ? LHSFreq < RHSFreq 593 : LI->getLoopDepth(L) < LI->getLoopDepth(R); 594 }); 595 596 auto it = AllSuccessors.insert(std::make_pair(MBB, AllSuccs)); 597 598 return it.first->second; 599 } 600 601 /// FindSuccToSinkTo - Find a successor to sink this instruction to. 602 MachineBasicBlock * 603 MachineSinking::FindSuccToSinkTo(MachineInstr &MI, MachineBasicBlock *MBB, 604 bool &BreakPHIEdge, 605 AllSuccsCache &AllSuccessors) { 606 assert (MBB && "Invalid MachineBasicBlock!"); 607 608 // Loop over all the operands of the specified instruction. If there is 609 // anything we can't handle, bail out. 610 611 // SuccToSinkTo - This is the successor to sink this instruction to, once we 612 // decide. 613 MachineBasicBlock *SuccToSinkTo = nullptr; 614 for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) { 615 const MachineOperand &MO = MI.getOperand(i); 616 if (!MO.isReg()) continue; // Ignore non-register operands. 617 618 unsigned Reg = MO.getReg(); 619 if (Reg == 0) continue; 620 621 if (TargetRegisterInfo::isPhysicalRegister(Reg)) { 622 if (MO.isUse()) { 623 // If the physreg has no defs anywhere, it's just an ambient register 624 // and we can freely move its uses. Alternatively, if it's allocatable, 625 // it could get allocated to something with a def during allocation. 626 if (!MRI->isConstantPhysReg(Reg)) 627 return nullptr; 628 } else if (!MO.isDead()) { 629 // A def that isn't dead. We can't move it. 630 return nullptr; 631 } 632 } else { 633 // Virtual register uses are always safe to sink. 634 if (MO.isUse()) continue; 635 636 // If it's not safe to move defs of the register class, then abort. 637 if (!TII->isSafeToMoveRegClassDefs(MRI->getRegClass(Reg))) 638 return nullptr; 639 640 // Virtual register defs can only be sunk if all their uses are in blocks 641 // dominated by one of the successors. 642 if (SuccToSinkTo) { 643 // If a previous operand picked a block to sink to, then this operand 644 // must be sinkable to the same block. 645 bool LocalUse = false; 646 if (!AllUsesDominatedByBlock(Reg, SuccToSinkTo, MBB, 647 BreakPHIEdge, LocalUse)) 648 return nullptr; 649 650 continue; 651 } 652 653 // Otherwise, we should look at all the successors and decide which one 654 // we should sink to. If we have reliable block frequency information 655 // (frequency != 0) available, give successors with smaller frequencies 656 // higher priority, otherwise prioritize smaller loop depths. 657 for (MachineBasicBlock *SuccBlock : 658 GetAllSortedSuccessors(MI, MBB, AllSuccessors)) { 659 bool LocalUse = false; 660 if (AllUsesDominatedByBlock(Reg, SuccBlock, MBB, 661 BreakPHIEdge, LocalUse)) { 662 SuccToSinkTo = SuccBlock; 663 break; 664 } 665 if (LocalUse) 666 // Def is used locally, it's never safe to move this def. 667 return nullptr; 668 } 669 670 // If we couldn't find a block to sink to, ignore this instruction. 671 if (!SuccToSinkTo) 672 return nullptr; 673 if (!isProfitableToSinkTo(Reg, MI, MBB, SuccToSinkTo, AllSuccessors)) 674 return nullptr; 675 } 676 } 677 678 // It is not possible to sink an instruction into its own block. This can 679 // happen with loops. 680 if (MBB == SuccToSinkTo) 681 return nullptr; 682 683 // It's not safe to sink instructions to EH landing pad. Control flow into 684 // landing pad is implicitly defined. 685 if (SuccToSinkTo && SuccToSinkTo->isEHPad()) 686 return nullptr; 687 688 return SuccToSinkTo; 689 } 690 691 /// Return true if MI is likely to be usable as a memory operation by the 692 /// implicit null check optimization. 693 /// 694 /// This is a "best effort" heuristic, and should not be relied upon for 695 /// correctness. This returning true does not guarantee that the implicit null 696 /// check optimization is legal over MI, and this returning false does not 697 /// guarantee MI cannot possibly be used to do a null check. 698 static bool SinkingPreventsImplicitNullCheck(MachineInstr &MI, 699 const TargetInstrInfo *TII, 700 const TargetRegisterInfo *TRI) { 701 using MachineBranchPredicate = TargetInstrInfo::MachineBranchPredicate; 702 703 auto *MBB = MI.getParent(); 704 if (MBB->pred_size() != 1) 705 return false; 706 707 auto *PredMBB = *MBB->pred_begin(); 708 auto *PredBB = PredMBB->getBasicBlock(); 709 710 // Frontends that don't use implicit null checks have no reason to emit 711 // branches with make.implicit metadata, and this function should always 712 // return false for them. 713 if (!PredBB || 714 !PredBB->getTerminator()->getMetadata(LLVMContext::MD_make_implicit)) 715 return false; 716 717 const MachineOperand *BaseOp; 718 int64_t Offset; 719 if (!TII->getMemOperandWithOffset(MI, BaseOp, Offset, TRI)) 720 return false; 721 722 if (!BaseOp->isReg()) 723 return false; 724 725 if (!(MI.mayLoad() && !MI.isPredicable())) 726 return false; 727 728 MachineBranchPredicate MBP; 729 if (TII->analyzeBranchPredicate(*PredMBB, MBP, false)) 730 return false; 731 732 return MBP.LHS.isReg() && MBP.RHS.isImm() && MBP.RHS.getImm() == 0 && 733 (MBP.Predicate == MachineBranchPredicate::PRED_NE || 734 MBP.Predicate == MachineBranchPredicate::PRED_EQ) && 735 MBP.LHS.getReg() == BaseOp->getReg(); 736 } 737 738 /// Sink an instruction and its associated debug instructions. If the debug 739 /// instructions to be sunk are already known, they can be provided in DbgVals. 740 static void performSink(MachineInstr &MI, MachineBasicBlock &SuccToSinkTo, 741 MachineBasicBlock::iterator InsertPos, 742 SmallVectorImpl<MachineInstr *> *DbgVals = nullptr) { 743 // If debug values are provided use those, otherwise call collectDebugValues. 744 SmallVector<MachineInstr *, 2> DbgValuesToSink; 745 if (DbgVals) 746 DbgValuesToSink.insert(DbgValuesToSink.begin(), 747 DbgVals->begin(), DbgVals->end()); 748 else 749 MI.collectDebugValues(DbgValuesToSink); 750 751 // If we cannot find a location to use (merge with), then we erase the debug 752 // location to prevent debug-info driven tools from potentially reporting 753 // wrong location information. 754 if (!SuccToSinkTo.empty() && InsertPos != SuccToSinkTo.end()) 755 MI.setDebugLoc(DILocation::getMergedLocation(MI.getDebugLoc(), 756 InsertPos->getDebugLoc())); 757 else 758 MI.setDebugLoc(DebugLoc()); 759 760 // Move the instruction. 761 MachineBasicBlock *ParentBlock = MI.getParent(); 762 SuccToSinkTo.splice(InsertPos, ParentBlock, MI, 763 ++MachineBasicBlock::iterator(MI)); 764 765 // Move previously adjacent debug value instructions to the insert position. 766 for (SmallVectorImpl<MachineInstr *>::iterator DBI = DbgValuesToSink.begin(), 767 DBE = DbgValuesToSink.end(); 768 DBI != DBE; ++DBI) { 769 MachineInstr *DbgMI = *DBI; 770 SuccToSinkTo.splice(InsertPos, ParentBlock, DbgMI, 771 ++MachineBasicBlock::iterator(DbgMI)); 772 } 773 } 774 775 /// SinkInstruction - Determine whether it is safe to sink the specified machine 776 /// instruction out of its current block into a successor. 777 bool MachineSinking::SinkInstruction(MachineInstr &MI, bool &SawStore, 778 AllSuccsCache &AllSuccessors) { 779 // Don't sink instructions that the target prefers not to sink. 780 if (!TII->shouldSink(MI)) 781 return false; 782 783 // Check if it's safe to move the instruction. 784 if (!MI.isSafeToMove(AA, SawStore)) 785 return false; 786 787 // Convergent operations may not be made control-dependent on additional 788 // values. 789 if (MI.isConvergent()) 790 return false; 791 792 // Don't break implicit null checks. This is a performance heuristic, and not 793 // required for correctness. 794 if (SinkingPreventsImplicitNullCheck(MI, TII, TRI)) 795 return false; 796 797 // FIXME: This should include support for sinking instructions within the 798 // block they are currently in to shorten the live ranges. We often get 799 // instructions sunk into the top of a large block, but it would be better to 800 // also sink them down before their first use in the block. This xform has to 801 // be careful not to *increase* register pressure though, e.g. sinking 802 // "x = y + z" down if it kills y and z would increase the live ranges of y 803 // and z and only shrink the live range of x. 804 805 bool BreakPHIEdge = false; 806 MachineBasicBlock *ParentBlock = MI.getParent(); 807 MachineBasicBlock *SuccToSinkTo = 808 FindSuccToSinkTo(MI, ParentBlock, BreakPHIEdge, AllSuccessors); 809 810 // If there are no outputs, it must have side-effects. 811 if (!SuccToSinkTo) 812 return false; 813 814 // If the instruction to move defines a dead physical register which is live 815 // when leaving the basic block, don't move it because it could turn into a 816 // "zombie" define of that preg. E.g., EFLAGS. (<rdar://problem/8030636>) 817 for (unsigned I = 0, E = MI.getNumOperands(); I != E; ++I) { 818 const MachineOperand &MO = MI.getOperand(I); 819 if (!MO.isReg()) continue; 820 unsigned Reg = MO.getReg(); 821 if (Reg == 0 || !TargetRegisterInfo::isPhysicalRegister(Reg)) continue; 822 if (SuccToSinkTo->isLiveIn(Reg)) 823 return false; 824 } 825 826 LLVM_DEBUG(dbgs() << "Sink instr " << MI << "\tinto block " << *SuccToSinkTo); 827 828 // If the block has multiple predecessors, this is a critical edge. 829 // Decide if we can sink along it or need to break the edge. 830 if (SuccToSinkTo->pred_size() > 1) { 831 // We cannot sink a load across a critical edge - there may be stores in 832 // other code paths. 833 bool TryBreak = false; 834 bool store = true; 835 if (!MI.isSafeToMove(AA, store)) { 836 LLVM_DEBUG(dbgs() << " *** NOTE: Won't sink load along critical edge.\n"); 837 TryBreak = true; 838 } 839 840 // We don't want to sink across a critical edge if we don't dominate the 841 // successor. We could be introducing calculations to new code paths. 842 if (!TryBreak && !DT->dominates(ParentBlock, SuccToSinkTo)) { 843 LLVM_DEBUG(dbgs() << " *** NOTE: Critical edge found\n"); 844 TryBreak = true; 845 } 846 847 // Don't sink instructions into a loop. 848 if (!TryBreak && LI->isLoopHeader(SuccToSinkTo)) { 849 LLVM_DEBUG(dbgs() << " *** NOTE: Loop header found\n"); 850 TryBreak = true; 851 } 852 853 // Otherwise we are OK with sinking along a critical edge. 854 if (!TryBreak) 855 LLVM_DEBUG(dbgs() << "Sinking along critical edge.\n"); 856 else { 857 // Mark this edge as to be split. 858 // If the edge can actually be split, the next iteration of the main loop 859 // will sink MI in the newly created block. 860 bool Status = 861 PostponeSplitCriticalEdge(MI, ParentBlock, SuccToSinkTo, BreakPHIEdge); 862 if (!Status) 863 LLVM_DEBUG(dbgs() << " *** PUNTING: Not legal or profitable to " 864 "break critical edge\n"); 865 // The instruction will not be sunk this time. 866 return false; 867 } 868 } 869 870 if (BreakPHIEdge) { 871 // BreakPHIEdge is true if all the uses are in the successor MBB being 872 // sunken into and they are all PHI nodes. In this case, machine-sink must 873 // break the critical edge first. 874 bool Status = PostponeSplitCriticalEdge(MI, ParentBlock, 875 SuccToSinkTo, BreakPHIEdge); 876 if (!Status) 877 LLVM_DEBUG(dbgs() << " *** PUNTING: Not legal or profitable to " 878 "break critical edge\n"); 879 // The instruction will not be sunk this time. 880 return false; 881 } 882 883 // Determine where to insert into. Skip phi nodes. 884 MachineBasicBlock::iterator InsertPos = SuccToSinkTo->begin(); 885 while (InsertPos != SuccToSinkTo->end() && InsertPos->isPHI()) 886 ++InsertPos; 887 888 performSink(MI, *SuccToSinkTo, InsertPos); 889 890 // Conservatively, clear any kill flags, since it's possible that they are no 891 // longer correct. 892 // Note that we have to clear the kill flags for any register this instruction 893 // uses as we may sink over another instruction which currently kills the 894 // used registers. 895 for (MachineOperand &MO : MI.operands()) { 896 if (MO.isReg() && MO.isUse()) 897 RegsToClearKillFlags.set(MO.getReg()); // Remember to clear kill flags. 898 } 899 900 return true; 901 } 902 903 //===----------------------------------------------------------------------===// 904 // This pass is not intended to be a replacement or a complete alternative 905 // for the pre-ra machine sink pass. It is only designed to sink COPY 906 // instructions which should be handled after RA. 907 // 908 // This pass sinks COPY instructions into a successor block, if the COPY is not 909 // used in the current block and the COPY is live-in to a single successor 910 // (i.e., doesn't require the COPY to be duplicated). This avoids executing the 911 // copy on paths where their results aren't needed. This also exposes 912 // additional opportunites for dead copy elimination and shrink wrapping. 913 // 914 // These copies were either not handled by or are inserted after the MachineSink 915 // pass. As an example of the former case, the MachineSink pass cannot sink 916 // COPY instructions with allocatable source registers; for AArch64 these type 917 // of copy instructions are frequently used to move function parameters (PhyReg) 918 // into virtual registers in the entry block. 919 // 920 // For the machine IR below, this pass will sink %w19 in the entry into its 921 // successor (%bb.1) because %w19 is only live-in in %bb.1. 922 // %bb.0: 923 // %wzr = SUBSWri %w1, 1 924 // %w19 = COPY %w0 925 // Bcc 11, %bb.2 926 // %bb.1: 927 // Live Ins: %w19 928 // BL @fun 929 // %w0 = ADDWrr %w0, %w19 930 // RET %w0 931 // %bb.2: 932 // %w0 = COPY %wzr 933 // RET %w0 934 // As we sink %w19 (CSR in AArch64) into %bb.1, the shrink-wrapping pass will be 935 // able to see %bb.0 as a candidate. 936 //===----------------------------------------------------------------------===// 937 namespace { 938 939 class PostRAMachineSinking : public MachineFunctionPass { 940 public: 941 bool runOnMachineFunction(MachineFunction &MF) override; 942 943 static char ID; 944 PostRAMachineSinking() : MachineFunctionPass(ID) {} 945 StringRef getPassName() const override { return "PostRA Machine Sink"; } 946 947 void getAnalysisUsage(AnalysisUsage &AU) const override { 948 AU.setPreservesCFG(); 949 MachineFunctionPass::getAnalysisUsage(AU); 950 } 951 952 MachineFunctionProperties getRequiredProperties() const override { 953 return MachineFunctionProperties().set( 954 MachineFunctionProperties::Property::NoVRegs); 955 } 956 957 private: 958 /// Track which register units have been modified and used. 959 LiveRegUnits ModifiedRegUnits, UsedRegUnits; 960 961 /// Track DBG_VALUEs of (unmodified) register units. 962 DenseMap<unsigned, TinyPtrVector<MachineInstr*>> SeenDbgInstrs; 963 964 /// Sink Copy instructions unused in the same block close to their uses in 965 /// successors. 966 bool tryToSinkCopy(MachineBasicBlock &BB, MachineFunction &MF, 967 const TargetRegisterInfo *TRI, const TargetInstrInfo *TII); 968 }; 969 } // namespace 970 971 char PostRAMachineSinking::ID = 0; 972 char &llvm::PostRAMachineSinkingID = PostRAMachineSinking::ID; 973 974 INITIALIZE_PASS(PostRAMachineSinking, "postra-machine-sink", 975 "PostRA Machine Sink", false, false) 976 977 static bool aliasWithRegsInLiveIn(MachineBasicBlock &MBB, unsigned Reg, 978 const TargetRegisterInfo *TRI) { 979 LiveRegUnits LiveInRegUnits(*TRI); 980 LiveInRegUnits.addLiveIns(MBB); 981 return !LiveInRegUnits.available(Reg); 982 } 983 984 static MachineBasicBlock * 985 getSingleLiveInSuccBB(MachineBasicBlock &CurBB, 986 const SmallPtrSetImpl<MachineBasicBlock *> &SinkableBBs, 987 unsigned Reg, const TargetRegisterInfo *TRI) { 988 // Try to find a single sinkable successor in which Reg is live-in. 989 MachineBasicBlock *BB = nullptr; 990 for (auto *SI : SinkableBBs) { 991 if (aliasWithRegsInLiveIn(*SI, Reg, TRI)) { 992 // If BB is set here, Reg is live-in to at least two sinkable successors, 993 // so quit. 994 if (BB) 995 return nullptr; 996 BB = SI; 997 } 998 } 999 // Reg is not live-in to any sinkable successors. 1000 if (!BB) 1001 return nullptr; 1002 1003 // Check if any register aliased with Reg is live-in in other successors. 1004 for (auto *SI : CurBB.successors()) { 1005 if (!SinkableBBs.count(SI) && aliasWithRegsInLiveIn(*SI, Reg, TRI)) 1006 return nullptr; 1007 } 1008 return BB; 1009 } 1010 1011 static MachineBasicBlock * 1012 getSingleLiveInSuccBB(MachineBasicBlock &CurBB, 1013 const SmallPtrSetImpl<MachineBasicBlock *> &SinkableBBs, 1014 ArrayRef<unsigned> DefedRegsInCopy, 1015 const TargetRegisterInfo *TRI) { 1016 MachineBasicBlock *SingleBB = nullptr; 1017 for (auto DefReg : DefedRegsInCopy) { 1018 MachineBasicBlock *BB = 1019 getSingleLiveInSuccBB(CurBB, SinkableBBs, DefReg, TRI); 1020 if (!BB || (SingleBB && SingleBB != BB)) 1021 return nullptr; 1022 SingleBB = BB; 1023 } 1024 return SingleBB; 1025 } 1026 1027 static void clearKillFlags(MachineInstr *MI, MachineBasicBlock &CurBB, 1028 SmallVectorImpl<unsigned> &UsedOpsInCopy, 1029 LiveRegUnits &UsedRegUnits, 1030 const TargetRegisterInfo *TRI) { 1031 for (auto U : UsedOpsInCopy) { 1032 MachineOperand &MO = MI->getOperand(U); 1033 unsigned SrcReg = MO.getReg(); 1034 if (!UsedRegUnits.available(SrcReg)) { 1035 MachineBasicBlock::iterator NI = std::next(MI->getIterator()); 1036 for (MachineInstr &UI : make_range(NI, CurBB.end())) { 1037 if (UI.killsRegister(SrcReg, TRI)) { 1038 UI.clearRegisterKills(SrcReg, TRI); 1039 MO.setIsKill(true); 1040 break; 1041 } 1042 } 1043 } 1044 } 1045 } 1046 1047 static void updateLiveIn(MachineInstr *MI, MachineBasicBlock *SuccBB, 1048 SmallVectorImpl<unsigned> &UsedOpsInCopy, 1049 SmallVectorImpl<unsigned> &DefedRegsInCopy) { 1050 MachineFunction &MF = *SuccBB->getParent(); 1051 const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo(); 1052 for (unsigned DefReg : DefedRegsInCopy) 1053 for (MCSubRegIterator S(DefReg, TRI, true); S.isValid(); ++S) 1054 SuccBB->removeLiveIn(*S); 1055 for (auto U : UsedOpsInCopy) { 1056 unsigned Reg = MI->getOperand(U).getReg(); 1057 if (!SuccBB->isLiveIn(Reg)) 1058 SuccBB->addLiveIn(Reg); 1059 } 1060 } 1061 1062 static bool hasRegisterDependency(MachineInstr *MI, 1063 SmallVectorImpl<unsigned> &UsedOpsInCopy, 1064 SmallVectorImpl<unsigned> &DefedRegsInCopy, 1065 LiveRegUnits &ModifiedRegUnits, 1066 LiveRegUnits &UsedRegUnits) { 1067 bool HasRegDependency = false; 1068 for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) { 1069 MachineOperand &MO = MI->getOperand(i); 1070 if (!MO.isReg()) 1071 continue; 1072 unsigned Reg = MO.getReg(); 1073 if (!Reg) 1074 continue; 1075 if (MO.isDef()) { 1076 if (!ModifiedRegUnits.available(Reg) || !UsedRegUnits.available(Reg)) { 1077 HasRegDependency = true; 1078 break; 1079 } 1080 DefedRegsInCopy.push_back(Reg); 1081 1082 // FIXME: instead of isUse(), readsReg() would be a better fix here, 1083 // For example, we can ignore modifications in reg with undef. However, 1084 // it's not perfectly clear if skipping the internal read is safe in all 1085 // other targets. 1086 } else if (MO.isUse()) { 1087 if (!ModifiedRegUnits.available(Reg)) { 1088 HasRegDependency = true; 1089 break; 1090 } 1091 UsedOpsInCopy.push_back(i); 1092 } 1093 } 1094 return HasRegDependency; 1095 } 1096 1097 bool PostRAMachineSinking::tryToSinkCopy(MachineBasicBlock &CurBB, 1098 MachineFunction &MF, 1099 const TargetRegisterInfo *TRI, 1100 const TargetInstrInfo *TII) { 1101 SmallPtrSet<MachineBasicBlock *, 2> SinkableBBs; 1102 // FIXME: For now, we sink only to a successor which has a single predecessor 1103 // so that we can directly sink COPY instructions to the successor without 1104 // adding any new block or branch instruction. 1105 for (MachineBasicBlock *SI : CurBB.successors()) 1106 if (!SI->livein_empty() && SI->pred_size() == 1) 1107 SinkableBBs.insert(SI); 1108 1109 if (SinkableBBs.empty()) 1110 return false; 1111 1112 bool Changed = false; 1113 1114 // Track which registers have been modified and used between the end of the 1115 // block and the current instruction. 1116 ModifiedRegUnits.clear(); 1117 UsedRegUnits.clear(); 1118 SeenDbgInstrs.clear(); 1119 1120 for (auto I = CurBB.rbegin(), E = CurBB.rend(); I != E;) { 1121 MachineInstr *MI = &*I; 1122 ++I; 1123 1124 // Track the operand index for use in Copy. 1125 SmallVector<unsigned, 2> UsedOpsInCopy; 1126 // Track the register number defed in Copy. 1127 SmallVector<unsigned, 2> DefedRegsInCopy; 1128 1129 // We must sink this DBG_VALUE if its operand is sunk. To avoid searching 1130 // for DBG_VALUEs later, record them when they're encountered. 1131 if (MI->isDebugValue()) { 1132 auto &MO = MI->getOperand(0); 1133 if (MO.isReg() && TRI->isPhysicalRegister(MO.getReg())) { 1134 // Bail if we can already tell the sink would be rejected, rather 1135 // than needlessly accumulating lots of DBG_VALUEs. 1136 if (hasRegisterDependency(MI, UsedOpsInCopy, DefedRegsInCopy, 1137 ModifiedRegUnits, UsedRegUnits)) 1138 continue; 1139 1140 // Record debug use of this register. 1141 SeenDbgInstrs[MO.getReg()].push_back(MI); 1142 } 1143 continue; 1144 } 1145 1146 if (MI->isDebugInstr()) 1147 continue; 1148 1149 // Do not move any instruction across function call. 1150 if (MI->isCall()) 1151 return false; 1152 1153 if (!MI->isCopy() || !MI->getOperand(0).isRenamable()) { 1154 LiveRegUnits::accumulateUsedDefed(*MI, ModifiedRegUnits, UsedRegUnits, 1155 TRI); 1156 continue; 1157 } 1158 1159 // Don't sink the COPY if it would violate a register dependency. 1160 if (hasRegisterDependency(MI, UsedOpsInCopy, DefedRegsInCopy, 1161 ModifiedRegUnits, UsedRegUnits)) { 1162 LiveRegUnits::accumulateUsedDefed(*MI, ModifiedRegUnits, UsedRegUnits, 1163 TRI); 1164 continue; 1165 } 1166 assert((!UsedOpsInCopy.empty() && !DefedRegsInCopy.empty()) && 1167 "Unexpect SrcReg or DefReg"); 1168 MachineBasicBlock *SuccBB = 1169 getSingleLiveInSuccBB(CurBB, SinkableBBs, DefedRegsInCopy, TRI); 1170 // Don't sink if we cannot find a single sinkable successor in which Reg 1171 // is live-in. 1172 if (!SuccBB) { 1173 LiveRegUnits::accumulateUsedDefed(*MI, ModifiedRegUnits, UsedRegUnits, 1174 TRI); 1175 continue; 1176 } 1177 assert((SuccBB->pred_size() == 1 && *SuccBB->pred_begin() == &CurBB) && 1178 "Unexpected predecessor"); 1179 1180 // Collect DBG_VALUEs that must sink with this copy. 1181 SmallVector<MachineInstr *, 4> DbgValsToSink; 1182 for (auto &MO : MI->operands()) { 1183 if (!MO.isReg() || !MO.isDef()) 1184 continue; 1185 unsigned reg = MO.getReg(); 1186 for (auto *MI : SeenDbgInstrs.lookup(reg)) 1187 DbgValsToSink.push_back(MI); 1188 } 1189 1190 // Clear the kill flag if SrcReg is killed between MI and the end of the 1191 // block. 1192 clearKillFlags(MI, CurBB, UsedOpsInCopy, UsedRegUnits, TRI); 1193 MachineBasicBlock::iterator InsertPos = SuccBB->getFirstNonPHI(); 1194 performSink(*MI, *SuccBB, InsertPos, &DbgValsToSink); 1195 updateLiveIn(MI, SuccBB, UsedOpsInCopy, DefedRegsInCopy); 1196 1197 Changed = true; 1198 ++NumPostRACopySink; 1199 } 1200 return Changed; 1201 } 1202 1203 bool PostRAMachineSinking::runOnMachineFunction(MachineFunction &MF) { 1204 if (skipFunction(MF.getFunction())) 1205 return false; 1206 1207 bool Changed = false; 1208 const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo(); 1209 const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo(); 1210 1211 ModifiedRegUnits.init(*TRI); 1212 UsedRegUnits.init(*TRI); 1213 for (auto &BB : MF) 1214 Changed |= tryToSinkCopy(BB, MF, TRI, TII); 1215 1216 return Changed; 1217 } 1218