llvm/lib/Transforms/Utils/CodeExtractor.cpp

   1 //===- CodeExtractor.cpp - Pull code region into a new function -----------===//
   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 file implements the interface to tear out a code region, such as an
  10 // individual loop or a parallel section, into a new function, replacing it with
  11 // a call to the new function.
  12 //
  13 //===----------------------------------------------------------------------===//
  14
  15 #include "llvm/Transforms/Utils/CodeExtractor.h"
  16 #include "llvm/ADT/ArrayRef.h"
  17 #include "llvm/ADT/DenseMap.h"
  18 #include "llvm/ADT/STLExtras.h"
  19 #include "llvm/ADT/SetVector.h"
  20 #include "llvm/ADT/SmallPtrSet.h"
  21 #include "llvm/ADT/SmallVector.h"
  22 #include "llvm/Analysis/AssumptionCache.h"
  23 #include "llvm/Analysis/BlockFrequencyInfo.h"
  24 #include "llvm/Analysis/BlockFrequencyInfoImpl.h"
  25 #include "llvm/Analysis/BranchProbabilityInfo.h"
  26 #include "llvm/Analysis/LoopInfo.h"
  27 #include "llvm/IR/Argument.h"
  28 #include "llvm/IR/Attributes.h"
  29 #include "llvm/IR/BasicBlock.h"
  30 #include "llvm/IR/CFG.h"
  31 #include "llvm/IR/Constant.h"
  32 #include "llvm/IR/Constants.h"
  33 #include "llvm/IR/DIBuilder.h"
  34 #include "llvm/IR/DataLayout.h"
  35 #include "llvm/IR/DebugInfo.h"
  36 #include "llvm/IR/DebugInfoMetadata.h"
  37 #include "llvm/IR/DerivedTypes.h"
  38 #include "llvm/IR/Dominators.h"
  39 #include "llvm/IR/Function.h"
  40 #include "llvm/IR/GlobalValue.h"
  41 #include "llvm/IR/InstIterator.h"
  42 #include "llvm/IR/InstrTypes.h"
  43 #include "llvm/IR/Instruction.h"
  44 #include "llvm/IR/Instructions.h"
  45 #include "llvm/IR/IntrinsicInst.h"
  46 #include "llvm/IR/Intrinsics.h"
  47 #include "llvm/IR/LLVMContext.h"
  48 #include "llvm/IR/MDBuilder.h"
  49 #include "llvm/IR/Module.h"
  50 #include "llvm/IR/PatternMatch.h"
  51 #include "llvm/IR/Type.h"
  52 #include "llvm/IR/User.h"
  53 #include "llvm/IR/Value.h"
  54 #include "llvm/IR/Verifier.h"
  55 #include "llvm/Support/BlockFrequency.h"
  56 #include "llvm/Support/BranchProbability.h"
  57 #include "llvm/Support/Casting.h"
  58 #include "llvm/Support/CommandLine.h"
  59 #include "llvm/Support/Debug.h"
  60 #include "llvm/Support/ErrorHandling.h"
  61 #include "llvm/Support/raw_ostream.h"
  62 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
  63 #include <cassert>
  64 #include <cstdint>
  65 #include <iterator>
  66 #include <map>
  67 #include <utility>
  68 #include <vector>
  69
  70 using namespace llvm;
  71 using namespace llvm::PatternMatch;
  72 using ProfileCount = Function::ProfileCount;
  73
  74 #define DEBUG_TYPE "code-extractor"
  75
  76 // Provide a command-line option to aggregate function arguments into a struct
  77 // for functions produced by the code extractor. This is useful when converting
  78 // extracted functions to pthread-based code, as only one argument (void*) can
  79 // be passed in to pthread_create().
  80 static cl::opt<bool>
  81 AggregateArgsOpt("aggregate-extracted-args", cl::Hidden,
  82                  cl::desc("Aggregate arguments to code-extracted functions"));
  83
  84 /// Test whether a block is valid for extraction.
  85 static bool isBlockValidForExtraction(const BasicBlock &BB,
  86                                       const SetVector<BasicBlock *> &Result,
  87                                       bool AllowVarArgs, bool AllowAlloca) {
  88   // taking the address of a basic block moved to another function is illegal
  89   if (BB.hasAddressTaken())
  90     return false;
  91
  92   // don't hoist code that uses another basicblock address, as it's likely to
  93   // lead to unexpected behavior, like cross-function jumps
  94   SmallPtrSet<User const *, 16> Visited;
  95   SmallVector<User const *, 16> ToVisit;
  96
  97   for (Instruction const &Inst : BB)
  98     ToVisit.push_back(&Inst);
  99
 100   while (!ToVisit.empty()) {
 101     User const *Curr = ToVisit.pop_back_val();
 102     if (!Visited.insert(Curr).second)
 103       continue;
 104     if (isa<BlockAddress const>(Curr))
 105       return false; // even a reference to self is likely to be not compatible
 106
 107     if (isa<Instruction>(Curr) && cast<Instruction>(Curr)->getParent() != &BB)
 108       continue;
 109
 110     for (auto const &U : Curr->operands()) {
 111       if (auto *UU = dyn_cast<User>(U))
 112         ToVisit.push_back(UU);
 113     }
 114   }
 115
 116   // If explicitly requested, allow vastart and alloca. For invoke instructions
 117   // verify that extraction is valid.
 118   for (BasicBlock::const_iterator I = BB.begin(), E = BB.end(); I != E; ++I) {
 119     if (isa<AllocaInst>(I)) {
 120        if (!AllowAlloca)
 121          return false;
 122        continue;
 123     }
 124
 125     if (const auto *II = dyn_cast<InvokeInst>(I)) {
 126       // Unwind destination (either a landingpad, catchswitch, or cleanuppad)
 127       // must be a part of the subgraph which is being extracted.
 128       if (auto *UBB = II->getUnwindDest())
 129         if (!Result.count(UBB))
 130           return false;
 131       continue;
 132     }
 133
 134     // All catch handlers of a catchswitch instruction as well as the unwind
 135     // destination must be in the subgraph.
 136     if (const auto *CSI = dyn_cast<CatchSwitchInst>(I)) {
 137       if (auto *UBB = CSI->getUnwindDest())
 138         if (!Result.count(UBB))
 139           return false;
 140       for (const auto *HBB : CSI->handlers())
 141         if (!Result.count(const_cast<BasicBlock*>(HBB)))
 142           return false;
 143       continue;
 144     }
 145
 146     // Make sure that entire catch handler is within subgraph. It is sufficient
 147     // to check that catch return's block is in the list.
 148     if (const auto *CPI = dyn_cast<CatchPadInst>(I)) {
 149       for (const auto *U : CPI->users())
 150         if (const auto *CRI = dyn_cast<CatchReturnInst>(U))
 151           if (!Result.count(const_cast<BasicBlock*>(CRI->getParent())))
 152             return false;
 153       continue;
 154     }
 155
 156     // And do similar checks for cleanup handler - the entire handler must be
 157     // in subgraph which is going to be extracted. For cleanup return should
 158     // additionally check that the unwind destination is also in the subgraph.
 159     if (const auto *CPI = dyn_cast<CleanupPadInst>(I)) {
 160       for (const auto *U : CPI->users())
 161         if (const auto *CRI = dyn_cast<CleanupReturnInst>(U))
 162           if (!Result.count(const_cast<BasicBlock*>(CRI->getParent())))
 163             return false;
 164       continue;
 165     }
 166     if (const auto *CRI = dyn_cast<CleanupReturnInst>(I)) {
 167       if (auto *UBB = CRI->getUnwindDest())
 168         if (!Result.count(UBB))
 169           return false;
 170       continue;
 171     }
 172
 173     if (const CallInst *CI = dyn_cast<CallInst>(I)) {
 174       if (const Function *F = CI->getCalledFunction()) {
 175         auto IID = F->getIntrinsicID();
 176         if (IID == Intrinsic::vastart) {
 177           if (AllowVarArgs)
 178             continue;
 179           else
 180             return false;
 181         }
 182
 183         // Currently, we miscompile outlined copies of eh_typid_for. There are
 184         // proposals for fixing this in llvm.org/PR39545.
 185         if (IID == Intrinsic::eh_typeid_for)
 186           return false;
 187       }
 188     }
 189   }
 190
 191   return true;
 192 }
 193
 194 /// Build a set of blocks to extract if the input blocks are viable.
 195 static SetVector<BasicBlock *>
 196 buildExtractionBlockSet(ArrayRef<BasicBlock *> BBs, DominatorTree *DT,
 197                         bool AllowVarArgs, bool AllowAlloca) {
 198   assert(!BBs.empty() && "The set of blocks to extract must be non-empty");
 199   SetVector<BasicBlock *> Result;
 200
 201   // Loop over the blocks, adding them to our set-vector, and aborting with an
 202   // empty set if we encounter invalid blocks.
 203   for (BasicBlock *BB : BBs) {
 204     // If this block is dead, don't process it.
 205     if (DT && !DT->isReachableFromEntry(BB))
 206       continue;
 207
 208     if (!Result.insert(BB))
 209       llvm_unreachable("Repeated basic blocks in extraction input");
 210   }
 211
 212   LLVM_DEBUG(dbgs() << "Region front block: " << Result.front()->getName()
 213                     << '\n');
 214
 215   for (auto *BB : Result) {
 216     if (!isBlockValidForExtraction(*BB, Result, AllowVarArgs, AllowAlloca))
 217       return {};
 218
 219     // Make sure that the first block is not a landing pad.
 220     if (BB == Result.front()) {
 221       if (BB->isEHPad()) {
 222         LLVM_DEBUG(dbgs() << "The first block cannot be an unwind block\n");
 223         return {};
 224       }
 225       continue;
 226     }
 227
 228     // All blocks other than the first must not have predecessors outside of
 229     // the subgraph which is being extracted.
 230     for (auto *PBB : predecessors(BB))
 231       if (!Result.count(PBB)) {
 232         LLVM_DEBUG(dbgs() << "No blocks in this region may have entries from "
 233                              "outside the region except for the first block!\n"
 234                           << "Problematic source BB: " << BB->getName() << "\n"
 235                           << "Problematic destination BB: " << PBB->getName()
 236                           << "\n");
 237         return {};
 238       }
 239   }
 240
 241   return Result;
 242 }
 243
 244 CodeExtractor::CodeExtractor(ArrayRef<BasicBlock *> BBs, DominatorTree *DT,
 245                              bool AggregateArgs, BlockFrequencyInfo *BFI,
 246                              BranchProbabilityInfo *BPI, AssumptionCache *AC,
 247                              bool AllowVarArgs, bool AllowAlloca,
 248                              BasicBlock *AllocationBlock, std::string Suffix)
 249     : DT(DT), AggregateArgs(AggregateArgs || AggregateArgsOpt), BFI(BFI),
 250       BPI(BPI), AC(AC), AllocationBlock(AllocationBlock),
 251       AllowVarArgs(AllowVarArgs),
 252       Blocks(buildExtractionBlockSet(BBs, DT, AllowVarArgs, AllowAlloca)),
 253       Suffix(Suffix) {}
 254
 255 CodeExtractor::CodeExtractor(DominatorTree &DT, Loop &L, bool AggregateArgs,
 256                              BlockFrequencyInfo *BFI,
 257                              BranchProbabilityInfo *BPI, AssumptionCache *AC,
 258                              std::string Suffix)
 259     : DT(&DT), AggregateArgs(AggregateArgs || AggregateArgsOpt), BFI(BFI),
 260       BPI(BPI), AC(AC), AllocationBlock(nullptr), AllowVarArgs(false),
 261       Blocks(buildExtractionBlockSet(L.getBlocks(), &DT,
 262                                      /* AllowVarArgs */ false,
 263                                      /* AllowAlloca */ false)),
 264       Suffix(Suffix) {}
 265
 266 /// definedInRegion - Return true if the specified value is defined in the
 267 /// extracted region.
 268 static bool definedInRegion(const SetVector<BasicBlock *> &Blocks, Value *V) {
 269   if (Instruction *I = dyn_cast<Instruction>(V))
 270     if (Blocks.count(I->getParent()))
 271       return true;
 272   return false;
 273 }
 274
 275 /// definedInCaller - Return true if the specified value is defined in the
 276 /// function being code extracted, but not in the region being extracted.
 277 /// These values must be passed in as live-ins to the function.
 278 static bool definedInCaller(const SetVector<BasicBlock *> &Blocks, Value *V) {
 279   if (isa<Argument>(V)) return true;
 280   if (Instruction *I = dyn_cast<Instruction>(V))
 281     if (!Blocks.count(I->getParent()))
 282       return true;
 283   return false;
 284 }
 285
 286 static BasicBlock *getCommonExitBlock(const SetVector<BasicBlock *> &Blocks) {
 287   BasicBlock *CommonExitBlock = nullptr;
 288   auto hasNonCommonExitSucc = [&](BasicBlock *Block) {
 289     for (auto *Succ : successors(Block)) {
 290       // Internal edges, ok.
 291       if (Blocks.count(Succ))
 292         continue;
 293       if (!CommonExitBlock) {
 294         CommonExitBlock = Succ;
 295         continue;
 296       }
 297       if (CommonExitBlock != Succ)
 298         return true;
 299     }
 300     return false;
 301   };
 302
 303   if (any_of(Blocks, hasNonCommonExitSucc))
 304     return nullptr;
 305
 306   return CommonExitBlock;
 307 }
 308
 309 CodeExtractorAnalysisCache::CodeExtractorAnalysisCache(Function &F) {
 310   for (BasicBlock &BB : F) {
 311     for (Instruction &II : BB.instructionsWithoutDebug())
 312       if (auto *AI = dyn_cast<AllocaInst>(&II))
 313         Allocas.push_back(AI);
 314
 315     findSideEffectInfoForBlock(BB);
 316   }
 317 }
 318
 319 void CodeExtractorAnalysisCache::findSideEffectInfoForBlock(BasicBlock &BB) {
 320   for (Instruction &II : BB.instructionsWithoutDebug()) {
 321     unsigned Opcode = II.getOpcode();
 322     Value *MemAddr = nullptr;
 323     switch (Opcode) {
 324     case Instruction::Store:
 325     case Instruction::Load: {
 326       if (Opcode == Instruction::Store) {
 327         StoreInst *SI = cast<StoreInst>(&II);
 328         MemAddr = SI->getPointerOperand();
 329       } else {
 330         LoadInst *LI = cast<LoadInst>(&II);
 331         MemAddr = LI->getPointerOperand();
 332       }
 333       // Global variable can not be aliased with locals.
 334       if (isa<Constant>(MemAddr))
 335         break;
 336       Value *Base = MemAddr->stripInBoundsConstantOffsets();
 337       if (!isa<AllocaInst>(Base)) {
 338         SideEffectingBlocks.insert(&BB);
 339         return;
 340       }
 341       BaseMemAddrs[&BB].insert(Base);
 342       break;
 343     }
 344     default: {
 345       IntrinsicInst *IntrInst = dyn_cast<IntrinsicInst>(&II);
 346       if (IntrInst) {
 347         if (IntrInst->isLifetimeStartOrEnd())
 348           break;
 349         SideEffectingBlocks.insert(&BB);
 350         return;
 351       }
 352       // Treat all the other cases conservatively if it has side effects.
 353       if (II.mayHaveSideEffects()) {
 354         SideEffectingBlocks.insert(&BB);
 355         return;
 356       }
 357     }
 358     }
 359   }
 360 }
 361
 362 bool CodeExtractorAnalysisCache::doesBlockContainClobberOfAddr(
 363     BasicBlock &BB, AllocaInst *Addr) const {
 364   if (SideEffectingBlocks.count(&BB))
 365     return true;
 366   auto It = BaseMemAddrs.find(&BB);
 367   if (It != BaseMemAddrs.end())
 368     return It->second.count(Addr);
 369   return false;
 370 }
 371
 372 bool CodeExtractor::isLegalToShrinkwrapLifetimeMarkers(
 373     const CodeExtractorAnalysisCache &CEAC, Instruction *Addr) const {
 374   AllocaInst *AI = cast<AllocaInst>(Addr->stripInBoundsConstantOffsets());
 375   Function *Func = (*Blocks.begin())->getParent();
 376   for (BasicBlock &BB : *Func) {
 377     if (Blocks.count(&BB))
 378       continue;
 379     if (CEAC.doesBlockContainClobberOfAddr(BB, AI))
 380       return false;
 381   }
 382   return true;
 383 }
 384
 385 BasicBlock *
 386 CodeExtractor::findOrCreateBlockForHoisting(BasicBlock *CommonExitBlock) {
 387   BasicBlock *SinglePredFromOutlineRegion = nullptr;
 388   assert(!Blocks.count(CommonExitBlock) &&
 389          "Expect a block outside the region!");
 390   for (auto *Pred : predecessors(CommonExitBlock)) {
 391     if (!Blocks.count(Pred))
 392       continue;
 393     if (!SinglePredFromOutlineRegion) {
 394       SinglePredFromOutlineRegion = Pred;
 395     } else if (SinglePredFromOutlineRegion != Pred) {
 396       SinglePredFromOutlineRegion = nullptr;
 397       break;
 398     }
 399   }
 400
 401   if (SinglePredFromOutlineRegion)
 402     return SinglePredFromOutlineRegion;
 403
 404 #ifndef NDEBUG
 405   auto getFirstPHI = [](BasicBlock *BB) {
 406     BasicBlock::iterator I = BB->begin();
 407     PHINode *FirstPhi = nullptr;
 408     while (I != BB->end()) {
 409       PHINode *Phi = dyn_cast<PHINode>(I);
 410       if (!Phi)
 411         break;
 412       if (!FirstPhi) {
 413         FirstPhi = Phi;
 414         break;
 415       }
 416     }
 417     return FirstPhi;
 418   };
 419   // If there are any phi nodes, the single pred either exists or has already
 420   // be created before code extraction.
 421   assert(!getFirstPHI(CommonExitBlock) && "Phi not expected");
 422 #endif
 423
 424   BasicBlock *NewExitBlock = CommonExitBlock->splitBasicBlock(
 425       CommonExitBlock->getFirstNonPHI()->getIterator());
 426
 427   for (BasicBlock *Pred :
 428        llvm::make_early_inc_range(predecessors(CommonExitBlock))) {
 429     if (Blocks.count(Pred))
 430       continue;
 431     Pred->getTerminator()->replaceUsesOfWith(CommonExitBlock, NewExitBlock);
 432   }
 433   // Now add the old exit block to the outline region.
 434   Blocks.insert(CommonExitBlock);
 435   OldTargets.push_back(NewExitBlock);
 436   return CommonExitBlock;
 437 }
 438
 439 // Find the pair of life time markers for address 'Addr' that are either
 440 // defined inside the outline region or can legally be shrinkwrapped into the
 441 // outline region. If there are not other untracked uses of the address, return
 442 // the pair of markers if found; otherwise return a pair of nullptr.
 443 CodeExtractor::LifetimeMarkerInfo
 444 CodeExtractor::getLifetimeMarkers(const CodeExtractorAnalysisCache &CEAC,
 445                                   Instruction *Addr,
 446                                   BasicBlock *ExitBlock) const {
 447   LifetimeMarkerInfo Info;
 448
 449   for (User *U : Addr->users()) {
 450     IntrinsicInst *IntrInst = dyn_cast<IntrinsicInst>(U);
 451     if (IntrInst) {
 452       // We don't model addresses with multiple start/end markers, but the
 453       // markers do not need to be in the region.
 454       if (IntrInst->getIntrinsicID() == Intrinsic::lifetime_start) {
 455         if (Info.LifeStart)
 456           return {};
 457         Info.LifeStart = IntrInst;
 458         continue;
 459       }
 460       if (IntrInst->getIntrinsicID() == Intrinsic::lifetime_end) {
 461         if (Info.LifeEnd)
 462           return {};
 463         Info.LifeEnd = IntrInst;
 464         continue;
 465       }
 466       // At this point, permit debug uses outside of the region.
 467       // This is fixed in a later call to fixupDebugInfoPostExtraction().
 468       if (isa<DbgInfoIntrinsic>(IntrInst))
 469         continue;
 470     }
 471     // Find untracked uses of the address, bail.
 472     if (!definedInRegion(Blocks, U))
 473       return {};
 474   }
 475
 476   if (!Info.LifeStart || !Info.LifeEnd)
 477     return {};
 478
 479   Info.SinkLifeStart = !definedInRegion(Blocks, Info.LifeStart);
 480   Info.HoistLifeEnd = !definedInRegion(Blocks, Info.LifeEnd);
 481   // Do legality check.
 482   if ((Info.SinkLifeStart || Info.HoistLifeEnd) &&
 483       !isLegalToShrinkwrapLifetimeMarkers(CEAC, Addr))
 484     return {};
 485
 486   // Check to see if we have a place to do hoisting, if not, bail.
 487   if (Info.HoistLifeEnd && !ExitBlock)
 488     return {};
 489
 490   return Info;
 491 }
 492
 493 void CodeExtractor::findAllocas(const CodeExtractorAnalysisCache &CEAC,
 494                                 ValueSet &SinkCands, ValueSet &HoistCands,
 495                                 BasicBlock *&ExitBlock) const {
 496   Function *Func = (*Blocks.begin())->getParent();
 497   ExitBlock = getCommonExitBlock(Blocks);
 498
 499   auto moveOrIgnoreLifetimeMarkers =
 500       [&](const LifetimeMarkerInfo &LMI) -> bool {
 501     if (!LMI.LifeStart)
 502       return false;
 503     if (LMI.SinkLifeStart) {
 504       LLVM_DEBUG(dbgs() << "Sinking lifetime.start: " << *LMI.LifeStart
 505                         << "\n");
 506       SinkCands.insert(LMI.LifeStart);
 507     }
 508     if (LMI.HoistLifeEnd) {
 509       LLVM_DEBUG(dbgs() << "Hoisting lifetime.end: " << *LMI.LifeEnd << "\n");
 510       HoistCands.insert(LMI.LifeEnd);
 511     }
 512     return true;
 513   };
 514
 515   // Look up allocas in the original function in CodeExtractorAnalysisCache, as
 516   // this is much faster than walking all the instructions.
 517   for (AllocaInst *AI : CEAC.getAllocas()) {
 518     BasicBlock *BB = AI->getParent();
 519     if (Blocks.count(BB))
 520       continue;
 521
 522     // As a prior call to extractCodeRegion() may have shrinkwrapped the alloca,
 523     // check whether it is actually still in the original function.
 524     Function *AIFunc = BB->getParent();
 525     if (AIFunc != Func)
 526       continue;
 527
 528     LifetimeMarkerInfo MarkerInfo = getLifetimeMarkers(CEAC, AI, ExitBlock);
 529     bool Moved = moveOrIgnoreLifetimeMarkers(MarkerInfo);
 530     if (Moved) {
 531       LLVM_DEBUG(dbgs() << "Sinking alloca: " << *AI << "\n");
 532       SinkCands.insert(AI);
 533       continue;
 534     }
 535
 536     // Find bitcasts in the outlined region that have lifetime marker users
 537     // outside that region. Replace the lifetime marker use with an
 538     // outside region bitcast to avoid unnecessary alloca/reload instructions
 539     // and extra lifetime markers.
 540     SmallVector<Instruction *, 2> LifetimeBitcastUsers;
 541     for (User *U : AI->users()) {
 542       if (!definedInRegion(Blocks, U))
 543         continue;
 544
 545       if (U->stripInBoundsConstantOffsets() != AI)
 546         continue;
 547
 548       Instruction *Bitcast = cast<Instruction>(U);
 549       for (User *BU : Bitcast->users()) {
 550         IntrinsicInst *IntrInst = dyn_cast<IntrinsicInst>(BU);
 551         if (!IntrInst)
 552           continue;
 553
 554         if (!IntrInst->isLifetimeStartOrEnd())
 555           continue;
 556
 557         if (definedInRegion(Blocks, IntrInst))
 558           continue;
 559
 560         LLVM_DEBUG(dbgs() << "Replace use of extracted region bitcast"
 561                           << *Bitcast << " in out-of-region lifetime marker "
 562                           << *IntrInst << "\n");
 563         LifetimeBitcastUsers.push_back(IntrInst);
 564       }
 565     }
 566
 567     for (Instruction *I : LifetimeBitcastUsers) {
 568       Module *M = AIFunc->getParent();
 569       LLVMContext &Ctx = M->getContext();
 570       auto *Int8PtrTy = Type::getInt8PtrTy(Ctx);
 571       CastInst *CastI =
 572           CastInst::CreatePointerCast(AI, Int8PtrTy, "lt.cast", I);
 573       I->replaceUsesOfWith(I->getOperand(1), CastI);
 574     }
 575
 576     // Follow any bitcasts.
 577     SmallVector<Instruction *, 2> Bitcasts;
 578     SmallVector<LifetimeMarkerInfo, 2> BitcastLifetimeInfo;
 579     for (User *U : AI->users()) {
 580       if (U->stripInBoundsConstantOffsets() == AI) {
 581         Instruction *Bitcast = cast<Instruction>(U);
 582         LifetimeMarkerInfo LMI = getLifetimeMarkers(CEAC, Bitcast, ExitBlock);
 583         if (LMI.LifeStart) {
 584           Bitcasts.push_back(Bitcast);
 585           BitcastLifetimeInfo.push_back(LMI);
 586           continue;
 587         }
 588       }
 589
 590       // Found unknown use of AI.
 591       if (!definedInRegion(Blocks, U)) {
 592         Bitcasts.clear();
 593         break;
 594       }
 595     }
 596
 597     // Either no bitcasts reference the alloca or there are unknown uses.
 598     if (Bitcasts.empty())
 599       continue;
 600
 601     LLVM_DEBUG(dbgs() << "Sinking alloca (via bitcast): " << *AI << "\n");
 602     SinkCands.insert(AI);
 603     for (unsigned I = 0, E = Bitcasts.size(); I != E; ++I) {
 604       Instruction *BitcastAddr = Bitcasts[I];
 605       const LifetimeMarkerInfo &LMI = BitcastLifetimeInfo[I];
 606       assert(LMI.LifeStart &&
 607              "Unsafe to sink bitcast without lifetime markers");
 608       moveOrIgnoreLifetimeMarkers(LMI);
 609       if (!definedInRegion(Blocks, BitcastAddr)) {
 610         LLVM_DEBUG(dbgs() << "Sinking bitcast-of-alloca: " << *BitcastAddr
 611                           << "\n");
 612         SinkCands.insert(BitcastAddr);
 613       }
 614     }
 615   }
 616 }
 617
 618 bool CodeExtractor::isEligible() const {
 619   if (Blocks.empty())
 620     return false;
 621   BasicBlock *Header = *Blocks.begin();
 622   Function *F = Header->getParent();
 623
 624   // For functions with varargs, check that varargs handling is only done in the
 625   // outlined function, i.e vastart and vaend are only used in outlined blocks.
 626   if (AllowVarArgs && F->getFunctionType()->isVarArg()) {
 627     auto containsVarArgIntrinsic = [](const Instruction &I) {
 628       if (const CallInst *CI = dyn_cast<CallInst>(&I))
 629         if (const Function *Callee = CI->getCalledFunction())
 630           return Callee->getIntrinsicID() == Intrinsic::vastart ||
 631                  Callee->getIntrinsicID() == Intrinsic::vaend;
 632       return false;
 633     };
 634
 635     for (auto &BB : *F) {
 636       if (Blocks.count(&BB))
 637         continue;
 638       if (llvm::any_of(BB, containsVarArgIntrinsic))
 639         return false;
 640     }
 641   }
 642   return true;
 643 }
 644
 645 void CodeExtractor::findInputsOutputs(ValueSet &Inputs, ValueSet &Outputs,
 646                                       const ValueSet &SinkCands) const {
 647   for (BasicBlock *BB : Blocks) {
 648     // If a used value is defined outside the region, it's an input.  If an
 649     // instruction is used outside the region, it's an output.
 650     for (Instruction &II : *BB) {
 651       for (auto &OI : II.operands()) {
 652         Value *V = OI;
 653         if (!SinkCands.count(V) && definedInCaller(Blocks, V))
 654           Inputs.insert(V);
 655       }
 656
 657       for (User *U : II.users())
 658         if (!definedInRegion(Blocks, U)) {
 659           Outputs.insert(&II);
 660           break;
 661         }
 662     }
 663   }
 664 }
 665
 666 /// severSplitPHINodesOfEntry - If a PHI node has multiple inputs from outside
 667 /// of the region, we need to split the entry block of the region so that the
 668 /// PHI node is easier to deal with.
 669 void CodeExtractor::severSplitPHINodesOfEntry(BasicBlock *&Header) {
 670   unsigned NumPredsFromRegion = 0;
 671   unsigned NumPredsOutsideRegion = 0;
 672
 673   if (Header != &Header->getParent()->getEntryBlock()) {
 674     PHINode *PN = dyn_cast<PHINode>(Header->begin());
 675     if (!PN) return;  // No PHI nodes.
 676
 677     // If the header node contains any PHI nodes, check to see if there is more
 678     // than one entry from outside the region.  If so, we need to sever the
 679     // header block into two.
 680     for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
 681       if (Blocks.count(PN->getIncomingBlock(i)))
 682         ++NumPredsFromRegion;
 683       else
 684         ++NumPredsOutsideRegion;
 685
 686     // If there is one (or fewer) predecessor from outside the region, we don't
 687     // need to do anything special.
 688     if (NumPredsOutsideRegion <= 1) return;
 689   }
 690
 691   // Otherwise, we need to split the header block into two pieces: one
 692   // containing PHI nodes merging values from outside of the region, and a
 693   // second that contains all of the code for the block and merges back any
 694   // incoming values from inside of the region.
 695   BasicBlock *NewBB = SplitBlock(Header, Header->getFirstNonPHI(), DT);
 696
 697   // We only want to code extract the second block now, and it becomes the new
 698   // header of the region.
 699   BasicBlock *OldPred = Header;
 700   Blocks.remove(OldPred);
 701   Blocks.insert(NewBB);
 702   Header = NewBB;
 703
 704   // Okay, now we need to adjust the PHI nodes and any branches from within the
 705   // region to go to the new header block instead of the old header block.
 706   if (NumPredsFromRegion) {
 707     PHINode *PN = cast<PHINode>(OldPred->begin());
 708     // Loop over all of the predecessors of OldPred that are in the region,
 709     // changing them to branch to NewBB instead.
 710     for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
 711       if (Blocks.count(PN->getIncomingBlock(i))) {
 712         Instruction *TI = PN->getIncomingBlock(i)->getTerminator();
 713         TI->replaceUsesOfWith(OldPred, NewBB);
 714       }
 715
 716     // Okay, everything within the region is now branching to the right block, we
 717     // just have to update the PHI nodes now, inserting PHI nodes into NewBB.
 718     BasicBlock::iterator AfterPHIs;
 719     for (AfterPHIs = OldPred->begin(); isa<PHINode>(AfterPHIs); ++AfterPHIs) {
 720       PHINode *PN = cast<PHINode>(AfterPHIs);
 721       // Create a new PHI node in the new region, which has an incoming value
 722       // from OldPred of PN.
 723       PHINode *NewPN = PHINode::Create(PN->getType(), 1 + NumPredsFromRegion,
 724                                        PN->getName() + ".ce");
 725       NewPN->insertBefore(NewBB->begin());
 726       PN->replaceAllUsesWith(NewPN);
 727       NewPN->addIncoming(PN, OldPred);
 728
 729       // Loop over all of the incoming value in PN, moving them to NewPN if they
 730       // are from the extracted region.
 731       for (unsigned i = 0; i != PN->getNumIncomingValues(); ++i) {
 732         if (Blocks.count(PN->getIncomingBlock(i))) {
 733           NewPN->addIncoming(PN->getIncomingValue(i), PN->getIncomingBlock(i));
 734           PN->removeIncomingValue(i);
 735           --i;
 736         }
 737       }
 738     }
 739   }
 740 }
 741
 742 /// severSplitPHINodesOfExits - if PHI nodes in exit blocks have inputs from
 743 /// outlined region, we split these PHIs on two: one with inputs from region
 744 /// and other with remaining incoming blocks; then first PHIs are placed in
 745 /// outlined region.
 746 void CodeExtractor::severSplitPHINodesOfExits(
 747     const SmallPtrSetImpl<BasicBlock *> &Exits) {
 748   for (BasicBlock *ExitBB : Exits) {
 749     BasicBlock *NewBB = nullptr;
 750
 751     for (PHINode &PN : ExitBB->phis()) {
 752       // Find all incoming values from the outlining region.
 753       SmallVector<unsigned, 2> IncomingVals;
 754       for (unsigned i = 0; i < PN.getNumIncomingValues(); ++i)
 755         if (Blocks.count(PN.getIncomingBlock(i)))
 756           IncomingVals.push_back(i);
 757
 758       // Do not process PHI if there is one (or fewer) predecessor from region.
 759       // If PHI has exactly one predecessor from region, only this one incoming
 760       // will be replaced on codeRepl block, so it should be safe to skip PHI.
 761       if (IncomingVals.size() <= 1)
 762         continue;
 763
 764       // Create block for new PHIs and add it to the list of outlined if it
 765       // wasn't done before.
 766       if (!NewBB) {
 767         NewBB = BasicBlock::Create(ExitBB->getContext(),
 768                                    ExitBB->getName() + ".split",
 769                                    ExitBB->getParent(), ExitBB);
 770         SmallVector<BasicBlock *, 4> Preds(predecessors(ExitBB));
 771         for (BasicBlock *PredBB : Preds)
 772           if (Blocks.count(PredBB))
 773             PredBB->getTerminator()->replaceUsesOfWith(ExitBB, NewBB);
 774         BranchInst::Create(ExitBB, NewBB);
 775         Blocks.insert(NewBB);
 776       }
 777
 778       // Split this PHI.
 779       PHINode *NewPN = PHINode::Create(PN.getType(), IncomingVals.size(),
 780                                        PN.getName() + ".ce");
 781       NewPN->insertBefore(NewBB->getFirstNonPHIIt());
 782       for (unsigned i : IncomingVals)
 783         NewPN->addIncoming(PN.getIncomingValue(i), PN.getIncomingBlock(i));
 784       for (unsigned i : reverse(IncomingVals))
 785         PN.removeIncomingValue(i, false);
 786       PN.addIncoming(NewPN, NewBB);
 787     }
 788   }
 789 }
 790
 791 void CodeExtractor::splitReturnBlocks() {
 792   for (BasicBlock *Block : Blocks)
 793     if (ReturnInst *RI = dyn_cast<ReturnInst>(Block->getTerminator())) {
 794       BasicBlock *New =
 795           Block->splitBasicBlock(RI->getIterator(), Block->getName() + ".ret");
 796       if (DT) {
 797         // Old dominates New. New node dominates all other nodes dominated
 798         // by Old.
 799         DomTreeNode *OldNode = DT->getNode(Block);
 800         SmallVector<DomTreeNode *, 8> Children(OldNode->begin(),
 801                                                OldNode->end());
 802
 803         DomTreeNode *NewNode = DT->addNewBlock(New, Block);
 804
 805         for (DomTreeNode *I : Children)
 806           DT->changeImmediateDominator(I, NewNode);
 807       }
 808     }
 809 }
 810
 811 /// constructFunction - make a function based on inputs and outputs, as follows:
 812 /// f(in0, ..., inN, out0, ..., outN)
 813 Function *CodeExtractor::constructFunction(const ValueSet &inputs,
 814                                            const ValueSet &outputs,
 815                                            BasicBlock *header,
 816                                            BasicBlock *newRootNode,
 817                                            BasicBlock *newHeader,
 818                                            Function *oldFunction,
 819                                            Module *M) {
 820   LLVM_DEBUG(dbgs() << "inputs: " << inputs.size() << "\n");
 821   LLVM_DEBUG(dbgs() << "outputs: " << outputs.size() << "\n");
 822
 823   // This function returns unsigned, outputs will go back by reference.
 824   switch (NumExitBlocks) {
 825   case 0:
 826   case 1: RetTy = Type::getVoidTy(header->getContext()); break;
 827   case 2: RetTy = Type::getInt1Ty(header->getContext()); break;
 828   default: RetTy = Type::getInt16Ty(header->getContext()); break;
 829   }
 830
 831   std::vector<Type *> ParamTy;
 832   std::vector<Type *> AggParamTy;
 833   ValueSet StructValues;
 834   const DataLayout &DL = M->getDataLayout();
 835
 836   // Add the types of the input values to the function's argument list
 837   for (Value *value : inputs) {
 838     LLVM_DEBUG(dbgs() << "value used in func: " << *value << "\n");
 839     if (AggregateArgs && !ExcludeArgsFromAggregate.contains(value)) {
 840       AggParamTy.push_back(value->getType());
 841       StructValues.insert(value);
 842     } else
 843       ParamTy.push_back(value->getType());
 844   }
 845
 846   // Add the types of the output values to the function's argument list.
 847   for (Value *output : outputs) {
 848     LLVM_DEBUG(dbgs() << "instr used in func: " << *output << "\n");
 849     if (AggregateArgs && !ExcludeArgsFromAggregate.contains(output)) {
 850       AggParamTy.push_back(output->getType());
 851       StructValues.insert(output);
 852     } else
 853       ParamTy.push_back(
 854           PointerType::get(output->getType(), DL.getAllocaAddrSpace()));
 855   }
 856
 857   assert(
 858       (ParamTy.size() + AggParamTy.size()) ==
 859           (inputs.size() + outputs.size()) &&
 860       "Number of scalar and aggregate params does not match inputs, outputs");
 861   assert((StructValues.empty() || AggregateArgs) &&
 862          "Expeced StructValues only with AggregateArgs set");
 863
 864   // Concatenate scalar and aggregate params in ParamTy.
 865   size_t NumScalarParams = ParamTy.size();
 866   StructType *StructTy = nullptr;
 867   if (AggregateArgs && !AggParamTy.empty()) {
 868     StructTy = StructType::get(M->getContext(), AggParamTy);
 869     ParamTy.push_back(PointerType::get(StructTy, DL.getAllocaAddrSpace()));
 870   }
 871
 872   LLVM_DEBUG({
 873     dbgs() << "Function type: " << *RetTy << " f(";
 874     for (Type *i : ParamTy)
 875       dbgs() << *i << ", ";
 876     dbgs() << ")\n";
 877   });
 878
 879   FunctionType *funcType = FunctionType::get(
 880       RetTy, ParamTy, AllowVarArgs && oldFunction->isVarArg());
 881
 882   std::string SuffixToUse =
 883       Suffix.empty()
 884           ? (header->getName().empty() ? "extracted" : header->getName().str())
 885           : Suffix;
 886   // Create the new function
 887   Function *newFunction = Function::Create(
 888       funcType, GlobalValue::InternalLinkage, oldFunction->getAddressSpace(),
 889       oldFunction->getName() + "." + SuffixToUse, M);
 890
 891   // Inherit all of the target dependent attributes and white-listed
 892   // target independent attributes.
 893   //  (e.g. If the extracted region contains a call to an x86.sse
 894   //  instruction we need to make sure that the extracted region has the
 895   //  "target-features" attribute allowing it to be lowered.
 896   // FIXME: This should be changed to check to see if a specific
 897   //           attribute can not be inherited.
 898   for (const auto &Attr : oldFunction->getAttributes().getFnAttrs()) {
 899     if (Attr.isStringAttribute()) {
 900       if (Attr.getKindAsString() == "thunk")
 901         continue;
 902     } else
 903       switch (Attr.getKindAsEnum()) {
 904       // Those attributes cannot be propagated safely. Explicitly list them
 905       // here so we get a warning if new attributes are added.
 906       case Attribute::AllocSize:
 907       case Attribute::Builtin:
 908       case Attribute::Convergent:
 909       case Attribute::JumpTable:
 910       case Attribute::Naked:
 911       case Attribute::NoBuiltin:
 912       case Attribute::NoMerge:
 913       case Attribute::NoReturn:
 914       case Attribute::NoSync:
 915       case Attribute::ReturnsTwice:
 916       case Attribute::Speculatable:
 917       case Attribute::StackAlignment:
 918       case Attribute::WillReturn:
 919       case Attribute::AllocKind:
 920       case Attribute::PresplitCoroutine:
 921       case Attribute::Memory:
 922       case Attribute::NoFPClass:
 923         continue;
 924       // Those attributes should be safe to propagate to the extracted function.
 925       case Attribute::AlwaysInline:
 926       case Attribute::Cold:
 927       case Attribute::DisableSanitizerInstrumentation:
 928       case Attribute::FnRetThunkExtern:
 929       case Attribute::Hot:
 930       case Attribute::NoRecurse:
 931       case Attribute::InlineHint:
 932       case Attribute::MinSize:
 933       case Attribute::NoCallback:
 934       case Attribute::NoDuplicate:
 935       case Attribute::NoFree:
 936       case Attribute::NoImplicitFloat:
 937       case Attribute::NoInline:
 938       case Attribute::NonLazyBind:
 939       case Attribute::NoRedZone:
 940       case Attribute::NoUnwind:
 941       case Attribute::NoSanitizeBounds:
 942       case Attribute::NoSanitizeCoverage:
 943       case Attribute::NullPointerIsValid:
 944       case Attribute::OptForFuzzing:
 945       case Attribute::OptimizeNone:
 946       case Attribute::OptimizeForSize:
 947       case Attribute::SafeStack:
 948       case Attribute::ShadowCallStack:
 949       case Attribute::SanitizeAddress:
 950       case Attribute::SanitizeMemory:
 951       case Attribute::SanitizeThread:
 952       case Attribute::SanitizeHWAddress:
 953       case Attribute::SanitizeMemTag:
 954       case Attribute::SpeculativeLoadHardening:
 955       case Attribute::StackProtect:
 956       case Attribute::StackProtectReq:
 957       case Attribute::StackProtectStrong:
 958       case Attribute::StrictFP:
 959       case Attribute::UWTable:
 960       case Attribute::VScaleRange:
 961       case Attribute::NoCfCheck:
 962       case Attribute::MustProgress:
 963       case Attribute::NoProfile:
 964       case Attribute::SkipProfile:
 965         break;
 966       // These attributes cannot be applied to functions.
 967       case Attribute::Alignment:
 968       case Attribute::AllocatedPointer:
 969       case Attribute::AllocAlign:
 970       case Attribute::ByVal:
 971       case Attribute::Dereferenceable:
 972       case Attribute::DereferenceableOrNull:
 973       case Attribute::ElementType:
 974       case Attribute::InAlloca:
 975       case Attribute::InReg:
 976       case Attribute::Nest:
 977       case Attribute::NoAlias:
 978       case Attribute::NoCapture:
 979       case Attribute::NoUndef:
 980       case Attribute::NonNull:
 981       case Attribute::Preallocated:
 982       case Attribute::ReadNone:
 983       case Attribute::ReadOnly:
 984       case Attribute::Returned:
 985       case Attribute::SExt:
 986       case Attribute::StructRet:
 987       case Attribute::SwiftError:
 988       case Attribute::SwiftSelf:
 989       case Attribute::SwiftAsync:
 990       case Attribute::ZExt:
 991       case Attribute::ImmArg:
 992       case Attribute::ByRef:
 993       case Attribute::WriteOnly:
 994       //  These are not really attributes.
 995       case Attribute::None:
 996       case Attribute::EndAttrKinds:
 997       case Attribute::EmptyKey:
 998       case Attribute::TombstoneKey:
 999         llvm_unreachable("Not a function attribute");
1000       }
1001
1002     newFunction->addFnAttr(Attr);
1003   }
1004   newFunction->insert(newFunction->end(), newRootNode);
1005
1006   // Create scalar and aggregate iterators to name all of the arguments we
1007   // inserted.
1008   Function::arg_iterator ScalarAI = newFunction->arg_begin();
1009   Function::arg_iterator AggAI = std::next(ScalarAI, NumScalarParams);
1010
1011   // Rewrite all users of the inputs in the extracted region to use the
1012   // arguments (or appropriate addressing into struct) instead.
1013   for (unsigned i = 0, e = inputs.size(), aggIdx = 0; i != e; ++i) {
1014     Value *RewriteVal;
1015     if (AggregateArgs && StructValues.contains(inputs[i])) {
1016       Value *Idx[2];
1017       Idx[0] = Constant::getNullValue(Type::getInt32Ty(header->getContext()));
1018       Idx[1] = ConstantInt::get(Type::getInt32Ty(header->getContext()), aggIdx);
1019       Instruction *TI = newFunction->begin()->getTerminator();
1020       GetElementPtrInst *GEP = GetElementPtrInst::Create(
1021           StructTy, &*AggAI, Idx, "gep_" + inputs[i]->getName(), TI);
1022       RewriteVal = new LoadInst(StructTy->getElementType(aggIdx), GEP,
1023                                 "loadgep_" + inputs[i]->getName(), TI);
1024       ++aggIdx;
1025     } else
1026       RewriteVal = &*ScalarAI++;
1027
1028     std::vector<User *> Users(inputs[i]->user_begin(), inputs[i]->user_end());
1029     for (User *use : Users)
1030       if (Instruction *inst = dyn_cast<Instruction>(use))
1031         if (Blocks.count(inst->getParent()))
1032           inst->replaceUsesOfWith(inputs[i], RewriteVal);
1033   }
1034
1035   // Set names for input and output arguments.
1036   if (NumScalarParams) {
1037     ScalarAI = newFunction->arg_begin();
1038     for (unsigned i = 0, e = inputs.size(); i != e; ++i, ++ScalarAI)
1039       if (!StructValues.contains(inputs[i]))
1040         ScalarAI->setName(inputs[i]->getName());
1041     for (unsigned i = 0, e = outputs.size(); i != e; ++i, ++ScalarAI)
1042       if (!StructValues.contains(outputs[i]))
1043         ScalarAI->setName(outputs[i]->getName() + ".out");
1044   }
1045
1046   // Rewrite branches to basic blocks outside of the loop to new dummy blocks
1047   // within the new function. This must be done before we lose track of which
1048   // blocks were originally in the code region.
1049   std::vector<User *> Users(header->user_begin(), header->user_end());
1050   for (auto &U : Users)
1051     // The BasicBlock which contains the branch is not in the region
1052     // modify the branch target to a new block
1053     if (Instruction *I = dyn_cast<Instruction>(U))
1054       if (I->isTerminator() && I->getFunction() == oldFunction &&
1055           !Blocks.count(I->getParent()))
1056         I->replaceUsesOfWith(header, newHeader);
1057
1058   return newFunction;
1059 }
1060
1061 /// Erase lifetime.start markers which reference inputs to the extraction
1062 /// region, and insert the referenced memory into \p LifetimesStart.
1063 ///
1064 /// The extraction region is defined by a set of blocks (\p Blocks), and a set
1065 /// of allocas which will be moved from the caller function into the extracted
1066 /// function (\p SunkAllocas).
1067 static void eraseLifetimeMarkersOnInputs(const SetVector<BasicBlock *> &Blocks,
1068                                          const SetVector<Value *> &SunkAllocas,
1069                                          SetVector<Value *> &LifetimesStart) {
1070   for (BasicBlock *BB : Blocks) {
1071     for (Instruction &I : llvm::make_early_inc_range(*BB)) {
1072       auto *II = dyn_cast<IntrinsicInst>(&I);
1073       if (!II || !II->isLifetimeStartOrEnd())
1074         continue;
1075
1076       // Get the memory operand of the lifetime marker. If the underlying
1077       // object is a sunk alloca, or is otherwise defined in the extraction
1078       // region, the lifetime marker must not be erased.
1079       Value *Mem = II->getOperand(1)->stripInBoundsOffsets();
1080       if (SunkAllocas.count(Mem) || definedInRegion(Blocks, Mem))
1081         continue;
1082
1083       if (II->getIntrinsicID() == Intrinsic::lifetime_start)
1084         LifetimesStart.insert(Mem);
1085       II->eraseFromParent();
1086     }
1087   }
1088 }
1089
1090 /// Insert lifetime start/end markers surrounding the call to the new function
1091 /// for objects defined in the caller.
1092 static void insertLifetimeMarkersSurroundingCall(
1093     Module *M, ArrayRef<Value *> LifetimesStart, ArrayRef<Value *> LifetimesEnd,
1094     CallInst *TheCall) {
1095   LLVMContext &Ctx = M->getContext();
1096   auto NegativeOne = ConstantInt::getSigned(Type::getInt64Ty(Ctx), -1);
1097   Instruction *Term = TheCall->getParent()->getTerminator();
1098
1099   // Emit lifetime markers for the pointers given in \p Objects. Insert the
1100   // markers before the call if \p InsertBefore, and after the call otherwise.
1101   auto insertMarkers = [&](Intrinsic::ID MarkerFunc, ArrayRef<Value *> Objects,
1102                            bool InsertBefore) {
1103     for (Value *Mem : Objects) {
1104       assert((!isa<Instruction>(Mem) || cast<Instruction>(Mem)->getFunction() ==
1105                                             TheCall->getFunction()) &&
1106              "Input memory not defined in original function");
1107
1108       Function *Func = Intrinsic::getDeclaration(M, MarkerFunc, Mem->getType());
1109       auto Marker = CallInst::Create(Func, {NegativeOne, Mem});
1110       if (InsertBefore)
1111         Marker->insertBefore(TheCall);
1112       else
1113         Marker->insertBefore(Term);
1114     }
1115   };
1116
1117   if (!LifetimesStart.empty()) {
1118     insertMarkers(Intrinsic::lifetime_start, LifetimesStart,
1119                   /*InsertBefore=*/true);
1120   }
1121
1122   if (!LifetimesEnd.empty()) {
1123     insertMarkers(Intrinsic::lifetime_end, LifetimesEnd,
1124                   /*InsertBefore=*/false);
1125   }
1126 }
1127
1128 /// emitCallAndSwitchStatement - This method sets up the caller side by adding
1129 /// the call instruction, splitting any PHI nodes in the header block as
1130 /// necessary.
1131 CallInst *CodeExtractor::emitCallAndSwitchStatement(Function *newFunction,
1132                                                     BasicBlock *codeReplacer,
1133                                                     ValueSet &inputs,
1134                                                     ValueSet &outputs) {
1135   // Emit a call to the new function, passing in: *pointer to struct (if
1136   // aggregating parameters), or plan inputs and allocated memory for outputs
1137   std::vector<Value *> params, ReloadOutputs, Reloads;
1138   ValueSet StructValues;
1139
1140   Module *M = newFunction->getParent();
1141   LLVMContext &Context = M->getContext();
1142   const DataLayout &DL = M->getDataLayout();
1143   CallInst *call = nullptr;
1144
1145   // Add inputs as params, or to be filled into the struct
1146   unsigned ScalarInputArgNo = 0;
1147   SmallVector<unsigned, 1> SwiftErrorArgs;
1148   for (Value *input : inputs) {
1149     if (AggregateArgs && !ExcludeArgsFromAggregate.contains(input))
1150       StructValues.insert(input);
1151     else {
1152       params.push_back(input);
1153       if (input->isSwiftError())
1154         SwiftErrorArgs.push_back(ScalarInputArgNo);
1155     }
1156     ++ScalarInputArgNo;
1157   }
1158
1159   // Create allocas for the outputs
1160   unsigned ScalarOutputArgNo = 0;
1161   for (Value *output : outputs) {
1162     if (AggregateArgs && !ExcludeArgsFromAggregate.contains(output)) {
1163       StructValues.insert(output);
1164     } else {
1165       AllocaInst *alloca =
1166         new AllocaInst(output->getType(), DL.getAllocaAddrSpace(),
1167                        nullptr, output->getName() + ".loc",
1168                        &codeReplacer->getParent()->front().front());
1169       ReloadOutputs.push_back(alloca);
1170       params.push_back(alloca);
1171       ++ScalarOutputArgNo;
1172     }
1173   }
1174
1175   StructType *StructArgTy = nullptr;
1176   AllocaInst *Struct = nullptr;
1177   unsigned NumAggregatedInputs = 0;
1178   if (AggregateArgs && !StructValues.empty()) {
1179     std::vector<Type *> ArgTypes;
1180     for (Value *V : StructValues)
1181       ArgTypes.push_back(V->getType());
1182
1183     // Allocate a struct at the beginning of this function
1184     StructArgTy = StructType::get(newFunction->getContext(), ArgTypes);
1185     Struct = new AllocaInst(
1186         StructArgTy, DL.getAllocaAddrSpace(), nullptr, "structArg",
1187         AllocationBlock ? &*AllocationBlock->getFirstInsertionPt()
1188                         : &codeReplacer->getParent()->front().front());
1189     params.push_back(Struct);
1190
1191     // Store aggregated inputs in the struct.
1192     for (unsigned i = 0, e = StructValues.size(); i != e; ++i) {
1193       if (inputs.contains(StructValues[i])) {
1194         Value *Idx[2];
1195         Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context));
1196         Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), i);
1197         GetElementPtrInst *GEP = GetElementPtrInst::Create(
1198             StructArgTy, Struct, Idx, "gep_" + StructValues[i]->getName());
1199         GEP->insertInto(codeReplacer, codeReplacer->end());
1200         new StoreInst(StructValues[i], GEP, codeReplacer);
1201         NumAggregatedInputs++;
1202       }
1203     }
1204   }
1205
1206   // Emit the call to the function
1207   call = CallInst::Create(newFunction, params,
1208                           NumExitBlocks > 1 ? "targetBlock" : "");
1209   // Add debug location to the new call, if the original function has debug
1210   // info. In that case, the terminator of the entry block of the extracted
1211   // function contains the first debug location of the extracted function,
1212   // set in extractCodeRegion.
1213   if (codeReplacer->getParent()->getSubprogram()) {
1214     if (auto DL = newFunction->getEntryBlock().getTerminator()->getDebugLoc())
1215       call->setDebugLoc(DL);
1216   }
1217   call->insertInto(codeReplacer, codeReplacer->end());
1218
1219   // Set swifterror parameter attributes.
1220   for (unsigned SwiftErrArgNo : SwiftErrorArgs) {
1221     call->addParamAttr(SwiftErrArgNo, Attribute::SwiftError);
1222     newFunction->addParamAttr(SwiftErrArgNo, Attribute::SwiftError);
1223   }
1224
1225   // Reload the outputs passed in by reference, use the struct if output is in
1226   // the aggregate or reload from the scalar argument.
1227   for (unsigned i = 0, e = outputs.size(), scalarIdx = 0,
1228                 aggIdx = NumAggregatedInputs;
1229        i != e; ++i) {
1230     Value *Output = nullptr;
1231     if (AggregateArgs && StructValues.contains(outputs[i])) {
1232       Value *Idx[2];
1233       Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context));
1234       Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), aggIdx);
1235       GetElementPtrInst *GEP = GetElementPtrInst::Create(
1236           StructArgTy, Struct, Idx, "gep_reload_" + outputs[i]->getName());
1237       GEP->insertInto(codeReplacer, codeReplacer->end());
1238       Output = GEP;
1239       ++aggIdx;
1240     } else {
1241       Output = ReloadOutputs[scalarIdx];
1242       ++scalarIdx;
1243     }
1244     LoadInst *load = new LoadInst(outputs[i]->getType(), Output,
1245                                   outputs[i]->getName() + ".reload",
1246                                   codeReplacer);
1247     Reloads.push_back(load);
1248     std::vector<User *> Users(outputs[i]->user_begin(), outputs[i]->user_end());
1249     for (User *U : Users) {
1250       Instruction *inst = cast<Instruction>(U);
1251       if (!Blocks.count(inst->getParent()))
1252         inst->replaceUsesOfWith(outputs[i], load);
1253     }
1254   }
1255
1256   // Now we can emit a switch statement using the call as a value.
1257   SwitchInst *TheSwitch =
1258       SwitchInst::Create(Constant::getNullValue(Type::getInt16Ty(Context)),
1259                          codeReplacer, 0, codeReplacer);
1260
1261   // Since there may be multiple exits from the original region, make the new
1262   // function return an unsigned, switch on that number.  This loop iterates
1263   // over all of the blocks in the extracted region, updating any terminator
1264   // instructions in the to-be-extracted region that branch to blocks that are
1265   // not in the region to be extracted.
1266   std::map<BasicBlock *, BasicBlock *> ExitBlockMap;
1267
1268   // Iterate over the previously collected targets, and create new blocks inside
1269   // the function to branch to.
1270   unsigned switchVal = 0;
1271   for (BasicBlock *OldTarget : OldTargets) {
1272     if (Blocks.count(OldTarget))
1273       continue;
1274     BasicBlock *&NewTarget = ExitBlockMap[OldTarget];
1275     if (NewTarget)
1276       continue;
1277
1278     // If we don't already have an exit stub for this non-extracted
1279     // destination, create one now!
1280     NewTarget = BasicBlock::Create(Context,
1281                                     OldTarget->getName() + ".exitStub",
1282                                     newFunction);
1283     unsigned SuccNum = switchVal++;
1284
1285     Value *brVal = nullptr;
1286     assert(NumExitBlocks < 0xffff && "too many exit blocks for switch");
1287     switch (NumExitBlocks) {
1288     case 0:
1289     case 1: break;  // No value needed.
1290     case 2:         // Conditional branch, return a bool
1291       brVal = ConstantInt::get(Type::getInt1Ty(Context), !SuccNum);
1292       break;
1293     default:
1294       brVal = ConstantInt::get(Type::getInt16Ty(Context), SuccNum);
1295       break;
1296     }
1297
1298     ReturnInst::Create(Context, brVal, NewTarget);
1299
1300     // Update the switch instruction.
1301     TheSwitch->addCase(ConstantInt::get(Type::getInt16Ty(Context),
1302                                         SuccNum),
1303                         OldTarget);
1304   }
1305
1306   for (BasicBlock *Block : Blocks) {
1307     Instruction *TI = Block->getTerminator();
1308     for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) {
1309       if (Blocks.count(TI->getSuccessor(i)))
1310         continue;
1311       BasicBlock *OldTarget = TI->getSuccessor(i);
1312       // add a new basic block which returns the appropriate value
1313       BasicBlock *NewTarget = ExitBlockMap[OldTarget];
1314       assert(NewTarget && "Unknown target block!");
1315
1316       // rewrite the original branch instruction with this new target
1317       TI->setSuccessor(i, NewTarget);
1318    }
1319   }
1320
1321   // Store the arguments right after the definition of output value.
1322   // This should be proceeded after creating exit stubs to be ensure that invoke
1323   // result restore will be placed in the outlined function.
1324   Function::arg_iterator ScalarOutputArgBegin = newFunction->arg_begin();
1325   std::advance(ScalarOutputArgBegin, ScalarInputArgNo);
1326   Function::arg_iterator AggOutputArgBegin = newFunction->arg_begin();
1327   std::advance(AggOutputArgBegin, ScalarInputArgNo + ScalarOutputArgNo);
1328
1329   for (unsigned i = 0, e = outputs.size(), aggIdx = NumAggregatedInputs; i != e;
1330        ++i) {
1331     auto *OutI = dyn_cast<Instruction>(outputs[i]);
1332     if (!OutI)
1333       continue;
1334
1335     // Find proper insertion point.
1336     BasicBlock::iterator InsertPt;
1337     // In case OutI is an invoke, we insert the store at the beginning in the
1338     // 'normal destination' BB. Otherwise we insert the store right after OutI.
1339     if (auto *InvokeI = dyn_cast<InvokeInst>(OutI))
1340       InsertPt = InvokeI->getNormalDest()->getFirstInsertionPt();
1341     else if (auto *Phi = dyn_cast<PHINode>(OutI))
1342       InsertPt = Phi->getParent()->getFirstInsertionPt();
1343     else
1344       InsertPt = std::next(OutI->getIterator());
1345
1346     Instruction *InsertBefore = &*InsertPt;
1347     assert((InsertBefore->getFunction() == newFunction ||
1348             Blocks.count(InsertBefore->getParent())) &&
1349            "InsertPt should be in new function");
1350     if (AggregateArgs && StructValues.contains(outputs[i])) {
1351       assert(AggOutputArgBegin != newFunction->arg_end() &&
1352              "Number of aggregate output arguments should match "
1353              "the number of defined values");
1354       Value *Idx[2];
1355       Idx[0] = Constant::getNullValue(Type::getInt32Ty(Context));
1356       Idx[1] = ConstantInt::get(Type::getInt32Ty(Context), aggIdx);
1357       GetElementPtrInst *GEP = GetElementPtrInst::Create(
1358           StructArgTy, &*AggOutputArgBegin, Idx, "gep_" + outputs[i]->getName(),
1359           InsertBefore);
1360       new StoreInst(outputs[i], GEP, InsertBefore);
1361       ++aggIdx;
1362       // Since there should be only one struct argument aggregating
1363       // all the output values, we shouldn't increment AggOutputArgBegin, which
1364       // always points to the struct argument, in this case.
1365     } else {
1366       assert(ScalarOutputArgBegin != newFunction->arg_end() &&
1367              "Number of scalar output arguments should match "
1368              "the number of defined values");
1369       new StoreInst(outputs[i], &*ScalarOutputArgBegin, InsertBefore);
1370       ++ScalarOutputArgBegin;
1371     }
1372   }
1373
1374   // Now that we've done the deed, simplify the switch instruction.
1375   Type *OldFnRetTy = TheSwitch->getParent()->getParent()->getReturnType();
1376   switch (NumExitBlocks) {
1377   case 0:
1378     // There are no successors (the block containing the switch itself), which
1379     // means that previously this was the last part of the function, and hence
1380     // this should be rewritten as a `ret'
1381
1382     // Check if the function should return a value
1383     if (OldFnRetTy->isVoidTy()) {
1384       ReturnInst::Create(Context, nullptr, TheSwitch);  // Return void
1385     } else if (OldFnRetTy == TheSwitch->getCondition()->getType()) {
1386       // return what we have
1387       ReturnInst::Create(Context, TheSwitch->getCondition(), TheSwitch);
1388     } else {
1389       // Otherwise we must have code extracted an unwind or something, just
1390       // return whatever we want.
1391       ReturnInst::Create(Context,
1392                          Constant::getNullValue(OldFnRetTy), TheSwitch);
1393     }
1394
1395     TheSwitch->eraseFromParent();
1396     break;
1397   case 1:
1398     // Only a single destination, change the switch into an unconditional
1399     // branch.
1400     BranchInst::Create(TheSwitch->getSuccessor(1), TheSwitch);
1401     TheSwitch->eraseFromParent();
1402     break;
1403   case 2:
1404     BranchInst::Create(TheSwitch->getSuccessor(1), TheSwitch->getSuccessor(2),
1405                        call, TheSwitch);
1406     TheSwitch->eraseFromParent();
1407     break;
1408   default:
1409     // Otherwise, make the default destination of the switch instruction be one
1410     // of the other successors.
1411     TheSwitch->setCondition(call);
1412     TheSwitch->setDefaultDest(TheSwitch->getSuccessor(NumExitBlocks));
1413     // Remove redundant case
1414     TheSwitch->removeCase(SwitchInst::CaseIt(TheSwitch, NumExitBlocks-1));
1415     break;
1416   }
1417
1418   // Insert lifetime markers around the reloads of any output values. The
1419   // allocas output values are stored in are only in-use in the codeRepl block.
1420   insertLifetimeMarkersSurroundingCall(M, ReloadOutputs, ReloadOutputs, call);
1421
1422   return call;
1423 }
1424
1425 void CodeExtractor::moveCodeToFunction(Function *newFunction) {
1426   auto newFuncIt = newFunction->front().getIterator();
1427   for (BasicBlock *Block : Blocks) {
1428     // Delete the basic block from the old function, and the list of blocks
1429     Block->removeFromParent();
1430
1431     // Insert this basic block into the new function
1432     // Insert the original blocks after the entry block created
1433     // for the new function. The entry block may be followed
1434     // by a set of exit blocks at this point, but these exit
1435     // blocks better be placed at the end of the new function.
1436     newFuncIt = newFunction->insert(std::next(newFuncIt), Block);
1437   }
1438 }
1439
1440 void CodeExtractor::calculateNewCallTerminatorWeights(
1441     BasicBlock *CodeReplacer,
1442     DenseMap<BasicBlock *, BlockFrequency> &ExitWeights,
1443     BranchProbabilityInfo *BPI) {
1444   using Distribution = BlockFrequencyInfoImplBase::Distribution;
1445   using BlockNode = BlockFrequencyInfoImplBase::BlockNode;
1446
1447   // Update the branch weights for the exit block.
1448   Instruction *TI = CodeReplacer->getTerminator();
1449   SmallVector<unsigned, 8> BranchWeights(TI->getNumSuccessors(), 0);
1450
1451   // Block Frequency distribution with dummy node.
1452   Distribution BranchDist;
1453
1454   SmallVector<BranchProbability, 4> EdgeProbabilities(
1455       TI->getNumSuccessors(), BranchProbability::getUnknown());
1456
1457   // Add each of the frequencies of the successors.
1458   for (unsigned i = 0, e = TI->getNumSuccessors(); i < e; ++i) {
1459     BlockNode ExitNode(i);
1460     uint64_t ExitFreq = ExitWeights[TI->getSuccessor(i)].getFrequency();
1461     if (ExitFreq != 0)
1462       BranchDist.addExit(ExitNode, ExitFreq);
1463     else
1464       EdgeProbabilities[i] = BranchProbability::getZero();
1465   }
1466
1467   // Check for no total weight.
1468   if (BranchDist.Total == 0) {
1469     BPI->setEdgeProbability(CodeReplacer, EdgeProbabilities);
1470     return;
1471   }
1472
1473   // Normalize the distribution so that they can fit in unsigned.
1474   BranchDist.normalize();
1475
1476   // Create normalized branch weights and set the metadata.
1477   for (unsigned I = 0, E = BranchDist.Weights.size(); I < E; ++I) {
1478     const auto &Weight = BranchDist.Weights[I];
1479
1480     // Get the weight and update the current BFI.
1481     BranchWeights[Weight.TargetNode.Index] = Weight.Amount;
1482     BranchProbability BP(Weight.Amount, BranchDist.Total);
1483     EdgeProbabilities[Weight.TargetNode.Index] = BP;
1484   }
1485   BPI->setEdgeProbability(CodeReplacer, EdgeProbabilities);
1486   TI->setMetadata(
1487       LLVMContext::MD_prof,
1488       MDBuilder(TI->getContext()).createBranchWeights(BranchWeights));
1489 }
1490
1491 /// Erase debug info intrinsics which refer to values in \p F but aren't in
1492 /// \p F.
1493 static void eraseDebugIntrinsicsWithNonLocalRefs(Function &F) {
1494   for (Instruction &I : instructions(F)) {
1495     SmallVector<DbgVariableIntrinsic *, 4> DbgUsers;
1496     findDbgUsers(DbgUsers, &I);
1497     for (DbgVariableIntrinsic *DVI : DbgUsers)
1498       if (DVI->getFunction() != &F)
1499         DVI->eraseFromParent();
1500   }
1501 }
1502
1503 /// Fix up the debug info in the old and new functions by pointing line
1504 /// locations and debug intrinsics to the new subprogram scope, and by deleting
1505 /// intrinsics which point to values outside of the new function.
1506 static void fixupDebugInfoPostExtraction(Function &OldFunc, Function &NewFunc,
1507                                          CallInst &TheCall) {
1508   DISubprogram *OldSP = OldFunc.getSubprogram();
1509   LLVMContext &Ctx = OldFunc.getContext();
1510
1511   if (!OldSP) {
1512     // Erase any debug info the new function contains.
1513     stripDebugInfo(NewFunc);
1514     // Make sure the old function doesn't contain any non-local metadata refs.
1515     eraseDebugIntrinsicsWithNonLocalRefs(NewFunc);
1516     return;
1517   }
1518
1519   // Create a subprogram for the new function. Leave out a description of the
1520   // function arguments, as the parameters don't correspond to anything at the
1521   // source level.
1522   assert(OldSP->getUnit() && "Missing compile unit for subprogram");
1523   DIBuilder DIB(*OldFunc.getParent(), /*AllowUnresolved=*/false,
1524                 OldSP->getUnit());
1525   auto SPType =
1526       DIB.createSubroutineType(DIB.getOrCreateTypeArray(std::nullopt));
1527   DISubprogram::DISPFlags SPFlags = DISubprogram::SPFlagDefinition |
1528                                     DISubprogram::SPFlagOptimized |
1529                                     DISubprogram::SPFlagLocalToUnit;
1530   auto NewSP = DIB.createFunction(
1531       OldSP->getUnit(), NewFunc.getName(), NewFunc.getName(), OldSP->getFile(),
1532       /*LineNo=*/0, SPType, /*ScopeLine=*/0, DINode::FlagZero, SPFlags);
1533   NewFunc.setSubprogram(NewSP);
1534
1535   // Debug intrinsics in the new function need to be updated in one of two
1536   // ways:
1537   //  1) They need to be deleted, because they describe a value in the old
1538   //     function.
1539   //  2) They need to point to fresh metadata, e.g. because they currently
1540   //     point to a variable in the wrong scope.
1541   SmallDenseMap<DINode *, DINode *> RemappedMetadata;
1542   SmallVector<Instruction *, 4> DebugIntrinsicsToDelete;
1543   DenseMap<const MDNode *, MDNode *> Cache;
1544   for (Instruction &I : instructions(NewFunc)) {
1545     auto *DII = dyn_cast<DbgInfoIntrinsic>(&I);
1546     if (!DII)
1547       continue;
1548
1549     // Point the intrinsic to a fresh label within the new function if the
1550     // intrinsic was not inlined from some other function.
1551     if (auto *DLI = dyn_cast<DbgLabelInst>(&I)) {
1552       if (DLI->getDebugLoc().getInlinedAt())
1553         continue;
1554       DILabel *OldLabel = DLI->getLabel();
1555       DINode *&NewLabel = RemappedMetadata[OldLabel];
1556       if (!NewLabel) {
1557         DILocalScope *NewScope = DILocalScope::cloneScopeForSubprogram(
1558             *OldLabel->getScope(), *NewSP, Ctx, Cache);
1559         NewLabel = DILabel::get(Ctx, NewScope, OldLabel->getName(),
1560                                 OldLabel->getFile(), OldLabel->getLine());
1561       }
1562       DLI->setArgOperand(0, MetadataAsValue::get(Ctx, NewLabel));
1563       continue;
1564     }
1565
1566     auto IsInvalidLocation = [&NewFunc](Value *Location) {
1567       // Location is invalid if it isn't a constant or an instruction, or is an
1568       // instruction but isn't in the new function.
1569       if (!Location ||
1570           (!isa<Constant>(Location) && !isa<Instruction>(Location)))
1571         return true;
1572       Instruction *LocationInst = dyn_cast<Instruction>(Location);
1573       return LocationInst && LocationInst->getFunction() != &NewFunc;
1574     };
1575
1576     auto *DVI = cast<DbgVariableIntrinsic>(DII);
1577     // If any of the used locations are invalid, delete the intrinsic.
1578     if (any_of(DVI->location_ops(), IsInvalidLocation)) {
1579       DebugIntrinsicsToDelete.push_back(DVI);
1580       continue;
1581     }
1582     // If the variable was in the scope of the old function, i.e. it was not
1583     // inlined, point the intrinsic to a fresh variable within the new function.
1584     if (!DVI->getDebugLoc().getInlinedAt()) {
1585       DILocalVariable *OldVar = DVI->getVariable();
1586       DINode *&NewVar = RemappedMetadata[OldVar];
1587       if (!NewVar) {
1588         DILocalScope *NewScope = DILocalScope::cloneScopeForSubprogram(
1589             *OldVar->getScope(), *NewSP, Ctx, Cache);
1590         NewVar = DIB.createAutoVariable(
1591             NewScope, OldVar->getName(), OldVar->getFile(), OldVar->getLine(),
1592             OldVar->getType(), /*AlwaysPreserve=*/false, DINode::FlagZero,
1593             OldVar->getAlignInBits());
1594       }
1595       DVI->setVariable(cast<DILocalVariable>(NewVar));
1596     }
1597   }
1598
1599   for (auto *DII : DebugIntrinsicsToDelete)
1600     DII->eraseFromParent();
1601   DIB.finalizeSubprogram(NewSP);
1602
1603   // Fix up the scope information attached to the line locations in the new
1604   // function.
1605   for (Instruction &I : instructions(NewFunc)) {
1606     if (const DebugLoc &DL = I.getDebugLoc())
1607       I.setDebugLoc(
1608           DebugLoc::replaceInlinedAtSubprogram(DL, *NewSP, Ctx, Cache));
1609
1610     // Loop info metadata may contain line locations. Fix them up.
1611     auto updateLoopInfoLoc = [&Ctx, &Cache, NewSP](Metadata *MD) -> Metadata * {
1612       if (auto *Loc = dyn_cast_or_null<DILocation>(MD))
1613         return DebugLoc::replaceInlinedAtSubprogram(Loc, *NewSP, Ctx, Cache);
1614       return MD;
1615     };
1616     updateLoopMetadataDebugLocations(I, updateLoopInfoLoc);
1617   }
1618   if (!TheCall.getDebugLoc())
1619     TheCall.setDebugLoc(DILocation::get(Ctx, 0, 0, OldSP));
1620
1621   eraseDebugIntrinsicsWithNonLocalRefs(NewFunc);
1622 }
1623
1624 Function *
1625 CodeExtractor::extractCodeRegion(const CodeExtractorAnalysisCache &CEAC) {
1626   ValueSet Inputs, Outputs;
1627   return extractCodeRegion(CEAC, Inputs, Outputs);
1628 }
1629
1630 Function *
1631 CodeExtractor::extractCodeRegion(const CodeExtractorAnalysisCache &CEAC,
1632                                  ValueSet &inputs, ValueSet &outputs) {
1633   if (!isEligible())
1634     return nullptr;
1635
1636   // Assumption: this is a single-entry code region, and the header is the first
1637   // block in the region.
1638   BasicBlock *header = *Blocks.begin();
1639   Function *oldFunction = header->getParent();
1640
1641   // Calculate the entry frequency of the new function before we change the root
1642   //   block.
1643   BlockFrequency EntryFreq;
1644   if (BFI) {
1645     assert(BPI && "Both BPI and BFI are required to preserve profile info");
1646     for (BasicBlock *Pred : predecessors(header)) {
1647       if (Blocks.count(Pred))
1648         continue;
1649       EntryFreq +=
1650           BFI->getBlockFreq(Pred) * BPI->getEdgeProbability(Pred, header);
1651     }
1652   }
1653
1654   // Remove @llvm.assume calls that will be moved to the new function from the
1655   // old function's assumption cache.
1656   for (BasicBlock *Block : Blocks) {
1657     for (Instruction &I : llvm::make_early_inc_range(*Block)) {
1658       if (auto *AI = dyn_cast<AssumeInst>(&I)) {
1659         if (AC)
1660           AC->unregisterAssumption(AI);
1661         AI->eraseFromParent();
1662       }
1663     }
1664   }
1665
1666   // If we have any return instructions in the region, split those blocks so
1667   // that the return is not in the region.
1668   splitReturnBlocks();
1669
1670   // Calculate the exit blocks for the extracted region and the total exit
1671   // weights for each of those blocks.
1672   DenseMap<BasicBlock *, BlockFrequency> ExitWeights;
1673   SmallPtrSet<BasicBlock *, 1> ExitBlocks;
1674   for (BasicBlock *Block : Blocks) {
1675     for (BasicBlock *Succ : successors(Block)) {
1676       if (!Blocks.count(Succ)) {
1677         // Update the branch weight for this successor.
1678         if (BFI) {
1679           BlockFrequency &BF = ExitWeights[Succ];
1680           BF += BFI->getBlockFreq(Block) * BPI->getEdgeProbability(Block, Succ);
1681         }
1682         ExitBlocks.insert(Succ);
1683       }
1684     }
1685   }
1686   NumExitBlocks = ExitBlocks.size();
1687
1688   for (BasicBlock *Block : Blocks) {
1689     Instruction *TI = Block->getTerminator();
1690     for (unsigned i = 0, e = TI->getNumSuccessors(); i != e; ++i) {
1691       if (Blocks.count(TI->getSuccessor(i)))
1692         continue;
1693       BasicBlock *OldTarget = TI->getSuccessor(i);
1694       OldTargets.push_back(OldTarget);
1695     }
1696   }
1697
1698   // If we have to split PHI nodes of the entry or exit blocks, do so now.
1699   severSplitPHINodesOfEntry(header);
1700   severSplitPHINodesOfExits(ExitBlocks);
1701
1702   // This takes place of the original loop
1703   BasicBlock *codeReplacer = BasicBlock::Create(header->getContext(),
1704                                                 "codeRepl", oldFunction,
1705                                                 header);
1706
1707   // The new function needs a root node because other nodes can branch to the
1708   // head of the region, but the entry node of a function cannot have preds.
1709   BasicBlock *newFuncRoot = BasicBlock::Create(header->getContext(),
1710                                                "newFuncRoot");
1711   auto *BranchI = BranchInst::Create(header);
1712   // If the original function has debug info, we have to add a debug location
1713   // to the new branch instruction from the artificial entry block.
1714   // We use the debug location of the first instruction in the extracted
1715   // blocks, as there is no other equivalent line in the source code.
1716   if (oldFunction->getSubprogram()) {
1717     any_of(Blocks, [&BranchI](const BasicBlock *BB) {
1718       return any_of(*BB, [&BranchI](const Instruction &I) {
1719         if (!I.getDebugLoc())
1720           return false;
1721         BranchI->setDebugLoc(I.getDebugLoc());
1722         return true;
1723       });
1724     });
1725   }
1726   BranchI->insertInto(newFuncRoot, newFuncRoot->end());
1727
1728   ValueSet SinkingCands, HoistingCands;
1729   BasicBlock *CommonExit = nullptr;
1730   findAllocas(CEAC, SinkingCands, HoistingCands, CommonExit);
1731   assert(HoistingCands.empty() || CommonExit);
1732
1733   // Find inputs to, outputs from the code region.
1734   findInputsOutputs(inputs, outputs, SinkingCands);
1735
1736   // Now sink all instructions which only have non-phi uses inside the region.
1737   // Group the allocas at the start of the block, so that any bitcast uses of
1738   // the allocas are well-defined.
1739   AllocaInst *FirstSunkAlloca = nullptr;
1740   for (auto *II : SinkingCands) {
1741     if (auto *AI = dyn_cast<AllocaInst>(II)) {
1742       AI->moveBefore(*newFuncRoot, newFuncRoot->getFirstInsertionPt());
1743       if (!FirstSunkAlloca)
1744         FirstSunkAlloca = AI;
1745     }
1746   }
1747   assert((SinkingCands.empty() || FirstSunkAlloca) &&
1748          "Did not expect a sink candidate without any allocas");
1749   for (auto *II : SinkingCands) {
1750     if (!isa<AllocaInst>(II)) {
1751       cast<Instruction>(II)->moveAfter(FirstSunkAlloca);
1752     }
1753   }
1754
1755   if (!HoistingCands.empty()) {
1756     auto *HoistToBlock = findOrCreateBlockForHoisting(CommonExit);
1757     Instruction *TI = HoistToBlock->getTerminator();
1758     for (auto *II : HoistingCands)
1759       cast<Instruction>(II)->moveBefore(TI);
1760   }
1761
1762   // Collect objects which are inputs to the extraction region and also
1763   // referenced by lifetime start markers within it. The effects of these
1764   // markers must be replicated in the calling function to prevent the stack
1765   // coloring pass from merging slots which store input objects.
1766   ValueSet LifetimesStart;
1767   eraseLifetimeMarkersOnInputs(Blocks, SinkingCands, LifetimesStart);
1768
1769   // Construct new function based on inputs/outputs & add allocas for all defs.
1770   Function *newFunction =
1771       constructFunction(inputs, outputs, header, newFuncRoot, codeReplacer,
1772                         oldFunction, oldFunction->getParent());
1773
1774   // Update the entry count of the function.
1775   if (BFI) {
1776     auto Count = BFI->getProfileCountFromFreq(EntryFreq.getFrequency());
1777     if (Count)
1778       newFunction->setEntryCount(
1779           ProfileCount(*Count, Function::PCT_Real)); // FIXME
1780     BFI->setBlockFreq(codeReplacer, EntryFreq.getFrequency());
1781   }
1782
1783   CallInst *TheCall =
1784       emitCallAndSwitchStatement(newFunction, codeReplacer, inputs, outputs);
1785
1786   moveCodeToFunction(newFunction);
1787
1788   // Replicate the effects of any lifetime start/end markers which referenced
1789   // input objects in the extraction region by placing markers around the call.
1790   insertLifetimeMarkersSurroundingCall(
1791       oldFunction->getParent(), LifetimesStart.getArrayRef(), {}, TheCall);
1792
1793   // Propagate personality info to the new function if there is one.
1794   if (oldFunction->hasPersonalityFn())
1795     newFunction->setPersonalityFn(oldFunction->getPersonalityFn());
1796
1797   // Update the branch weights for the exit block.
1798   if (BFI && NumExitBlocks > 1)
1799     calculateNewCallTerminatorWeights(codeReplacer, ExitWeights, BPI);
1800
1801   // Loop over all of the PHI nodes in the header and exit blocks, and change
1802   // any references to the old incoming edge to be the new incoming edge.
1803   for (BasicBlock::iterator I = header->begin(); isa<PHINode>(I); ++I) {
1804     PHINode *PN = cast<PHINode>(I);
1805     for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
1806       if (!Blocks.count(PN->getIncomingBlock(i)))
1807         PN->setIncomingBlock(i, newFuncRoot);
1808   }
1809
1810   for (BasicBlock *ExitBB : ExitBlocks)
1811     for (PHINode &PN : ExitBB->phis()) {
1812       Value *IncomingCodeReplacerVal = nullptr;
1813       for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i) {
1814         // Ignore incoming values from outside of the extracted region.
1815         if (!Blocks.count(PN.getIncomingBlock(i)))
1816           continue;
1817
1818         // Ensure that there is only one incoming value from codeReplacer.
1819         if (!IncomingCodeReplacerVal) {
1820           PN.setIncomingBlock(i, codeReplacer);
1821           IncomingCodeReplacerVal = PN.getIncomingValue(i);
1822         } else
1823           assert(IncomingCodeReplacerVal == PN.getIncomingValue(i) &&
1824                  "PHI has two incompatbile incoming values from codeRepl");
1825       }
1826     }
1827
1828   fixupDebugInfoPostExtraction(*oldFunction, *newFunction, *TheCall);
1829
1830   // Mark the new function `noreturn` if applicable. Terminators which resume
1831   // exception propagation are treated as returning instructions. This is to
1832   // avoid inserting traps after calls to outlined functions which unwind.
1833   bool doesNotReturn = none_of(*newFunction, [](const BasicBlock &BB) {
1834     const Instruction *Term = BB.getTerminator();
1835     return isa<ReturnInst>(Term) || isa<ResumeInst>(Term);
1836   });
1837   if (doesNotReturn)
1838     newFunction->setDoesNotReturn();
1839
1840   LLVM_DEBUG(if (verifyFunction(*newFunction, &errs())) {
1841     newFunction->dump();
1842     report_fatal_error("verification of newFunction failed!");
1843   });
1844   LLVM_DEBUG(if (verifyFunction(*oldFunction))
1845              report_fatal_error("verification of oldFunction failed!"));
1846   LLVM_DEBUG(if (AC && verifyAssumptionCache(*oldFunction, *newFunction, AC))
1847                  report_fatal_error("Stale Asumption cache for old Function!"));
1848   return newFunction;
1849 }
1850
1851 bool CodeExtractor::verifyAssumptionCache(const Function &OldFunc,
1852                                           const Function &NewFunc,
1853                                           AssumptionCache *AC) {
1854   for (auto AssumeVH : AC->assumptions()) {
1855     auto *I = dyn_cast_or_null<CallInst>(AssumeVH);
1856     if (!I)
1857       continue;
1858
1859     // There shouldn't be any llvm.assume intrinsics in the new function.
1860     if (I->getFunction() != &OldFunc)
1861       return true;
1862
1863     // There shouldn't be any stale affected values in the assumption cache
1864     // that were previously in the old function, but that have now been moved
1865     // to the new function.
1866     for (auto AffectedValVH : AC->assumptionsFor(I->getOperand(0))) {
1867       auto *AffectedCI = dyn_cast_or_null<CallInst>(AffectedValVH);
1868       if (!AffectedCI)
1869         continue;
1870       if (AffectedCI->getFunction() != &OldFunc)
1871         return true;
1872       auto *AssumedInst = cast<Instruction>(AffectedCI->getOperand(0));
1873       if (AssumedInst->getFunction() != &OldFunc)
1874         return true;
1875     }
1876   }
1877   return false;
1878 }
1879
1880 void CodeExtractor::excludeArgFromAggregate(Value *Arg) {
1881   ExcludeArgsFromAggregate.insert(Arg);
1882 }