C++ SmallSetVector::size方法代码示例

MDNode *MDNode::getMostGenericTBAA(MDNode *A, MDNode *B) {
  if (!A || !B)
    return nullptr;

  if (A == B)
    return A;

  // For struct-path aware TBAA, we use the access type of the tag.
  bool StructPath = isStructPathTBAA(A) && isStructPathTBAA(B);
  if (StructPath) {
    A = cast_or_null<MDNode>(A->getOperand(1));
    if (!A) return nullptr;
    B = cast_or_null<MDNode>(B->getOperand(1));
    if (!B) return nullptr;
  }

  SmallSetVector<MDNode *, 4> PathA;
  MDNode *T = A;
  while (T) {
    if (PathA.count(T))
      report_fatal_error("Cycle found in TBAA metadata.");
    PathA.insert(T);
    T = T->getNumOperands() >= 2 ? cast_or_null<MDNode>(T->getOperand(1))
                                 : nullptr;
  }

  SmallSetVector<MDNode *, 4> PathB;
  T = B;
  while (T) {
    if (PathB.count(T))
      report_fatal_error("Cycle found in TBAA metadata.");
    PathB.insert(T);
    T = T->getNumOperands() >= 2 ? cast_or_null<MDNode>(T->getOperand(1))
                                 : nullptr;
  }

  int IA = PathA.size() - 1;
  int IB = PathB.size() - 1;

  MDNode *Ret = nullptr;
  while (IA >= 0 && IB >=0) {
    if (PathA[IA] == PathB[IB])
      Ret = PathA[IA];
    else
      break;
    --IA;
    --IB;
  }
  if (!StructPath)
    return Ret;

  if (!Ret)
    return nullptr;
  // We need to convert from a type node to a tag node.
  Type *Int64 = IntegerType::get(A->getContext(), 64);
  Metadata *Ops[3] = {Ret, Ret,
                      ConstantAsMetadata::get(ConstantInt::get(Int64, 0))};
  return MDNode::get(A->getContext(), Ops);
}

MDNode *MDNode::getMostGenericTBAA(MDNode *A, MDNode *B) {
  if (!A || !B)
    return nullptr;

  if (A == B)
    return A;

  // For struct-path aware TBAA, we use the access type of the tag.
  assert(isStructPathTBAA(A) && isStructPathTBAA(B) &&
         "Auto upgrade should have taken care of this!");
  A = cast_or_null<MDNode>(MutableTBAAStructTagNode(A).getAccessType());
  if (!A)
    return nullptr;
  B = cast_or_null<MDNode>(MutableTBAAStructTagNode(B).getAccessType());
  if (!B)
    return nullptr;

  SmallSetVector<MDNode *, 4> PathA;
  MutableTBAANode TA(A);
  while (TA.getNode()) {
    if (PathA.count(TA.getNode()))
      report_fatal_error("Cycle found in TBAA metadata.");
    PathA.insert(TA.getNode());
    TA = TA.getParent();
  }

  SmallSetVector<MDNode *, 4> PathB;
  MutableTBAANode TB(B);
  while (TB.getNode()) {
    if (PathB.count(TB.getNode()))
      report_fatal_error("Cycle found in TBAA metadata.");
    PathB.insert(TB.getNode());
    TB = TB.getParent();
  }

  int IA = PathA.size() - 1;
  int IB = PathB.size() - 1;

  MDNode *Ret = nullptr;
  while (IA >= 0 && IB >= 0) {
    if (PathA[IA] == PathB[IB])
      Ret = PathA[IA];
    else
      break;
    --IA;
    --IB;
  }

  if (!Ret)
    return nullptr;
  // We need to convert from a type node to a tag node.
  Type *Int64 = IntegerType::get(A->getContext(), 64);
  Metadata *Ops[3] = {Ret, Ret,
                      ConstantAsMetadata::get(ConstantInt::get(Int64, 0))};
  return MDNode::get(A->getContext(), Ops);
}

/// Tests whether this function is known to not return null.
///
/// Requires that the function returns a pointer.
///
/// Returns true if it believes the function will not return a null, and sets
/// \p Speculative based on whether the returned conclusion is a speculative
/// conclusion due to SCC calls.
static bool isReturnNonNull(Function *F, const SCCNodeSet &SCCNodes,
                            bool &Speculative) {
  assert(F->getReturnType()->isPointerTy() &&
         "nonnull only meaningful on pointer types");
  Speculative = false;

  SmallSetVector<Value *, 8> FlowsToReturn;
  for (BasicBlock &BB : *F)
    if (auto *Ret = dyn_cast<ReturnInst>(BB.getTerminator()))
      FlowsToReturn.insert(Ret->getReturnValue());

  for (unsigned i = 0; i != FlowsToReturn.size(); ++i) {
    Value *RetVal = FlowsToReturn[i];

    // If this value is locally known to be non-null, we're good
    if (isKnownNonNull(RetVal))
      continue;

    // Otherwise, we need to look upwards since we can't make any local
    // conclusions.
    Instruction *RVI = dyn_cast<Instruction>(RetVal);
    if (!RVI)
      return false;
    switch (RVI->getOpcode()) {
    // Extend the analysis by looking upwards.
    case Instruction::BitCast:
    case Instruction::GetElementPtr:
    case Instruction::AddrSpaceCast:
      FlowsToReturn.insert(RVI->getOperand(0));
      continue;
    case Instruction::Select: {
      SelectInst *SI = cast<SelectInst>(RVI);
      FlowsToReturn.insert(SI->getTrueValue());
      FlowsToReturn.insert(SI->getFalseValue());
      continue;
    }
    case Instruction::PHI: {
      PHINode *PN = cast<PHINode>(RVI);
      for (int i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
        FlowsToReturn.insert(PN->getIncomingValue(i));
      continue;
    }
    case Instruction::Call:
    case Instruction::Invoke: {
      CallSite CS(RVI);
      Function *Callee = CS.getCalledFunction();
      // A call to a node within the SCC is assumed to return null until
      // proven otherwise
      if (Callee && SCCNodes.count(Callee)) {
        Speculative = true;
        continue;
      }
      return false;
    }
    default:
      return false; // Unknown source, may be null
    };
    llvm_unreachable("should have either continued or returned");
  }

  return true;
}

/// Tests whether a function is "malloc-like".
///
/// A function is "malloc-like" if it returns either null or a pointer that
/// doesn't alias any other pointer visible to the caller.
static bool isFunctionMallocLike(Function *F, const SCCNodeSet &SCCNodes) {
  SmallSetVector<Value *, 8> FlowsToReturn;
  for (BasicBlock &BB : *F)
    if (ReturnInst *Ret = dyn_cast<ReturnInst>(BB.getTerminator()))
      FlowsToReturn.insert(Ret->getReturnValue());

  for (unsigned i = 0; i != FlowsToReturn.size(); ++i) {
    Value *RetVal = FlowsToReturn[i];

    if (Constant *C = dyn_cast<Constant>(RetVal)) {
      if (!C->isNullValue() && !isa<UndefValue>(C))
        return false;

      continue;
    }

    if (isa<Argument>(RetVal))
      return false;

    if (Instruction *RVI = dyn_cast<Instruction>(RetVal))
      switch (RVI->getOpcode()) {
      // Extend the analysis by looking upwards.
      case Instruction::BitCast:
      case Instruction::GetElementPtr:
      case Instruction::AddrSpaceCast:
        FlowsToReturn.insert(RVI->getOperand(0));
        continue;
      case Instruction::Select: {
        SelectInst *SI = cast<SelectInst>(RVI);
        FlowsToReturn.insert(SI->getTrueValue());
        FlowsToReturn.insert(SI->getFalseValue());
        continue;
      }
      case Instruction::PHI: {
        PHINode *PN = cast<PHINode>(RVI);
        for (Value *IncValue : PN->incoming_values())
          FlowsToReturn.insert(IncValue);
        continue;
      }

      // Check whether the pointer came from an allocation.
      case Instruction::Alloca:
        break;
      case Instruction::Call:
      case Instruction::Invoke: {
        CallSite CS(RVI);
        if (CS.hasRetAttr(Attribute::NoAlias))
          break;
        if (CS.getCalledFunction() && SCCNodes.count(CS.getCalledFunction()))
          break;
        LLVM_FALLTHROUGH;
      }
      default:
        return false; // Did not come from an allocation.
      }

    if (PointerMayBeCaptured(RetVal, false, /*StoreCaptures=*/false))
      return false;
  }

  return true;
}

/// Sort the blocks, taking special care to make sure that loops are not
/// interrupted by blocks not dominated by their header.
/// TODO: There are many opportunities for improving the heuristics here.
/// Explore them.
static void SortBlocks(MachineFunction &MF, const MachineLoopInfo &MLI,
                       const MachineDominatorTree &MDT) {
  // Prepare for a topological sort: Record the number of predecessors each
  // block has, ignoring loop backedges.
  MF.RenumberBlocks();
  SmallVector<unsigned, 16> NumPredsLeft(MF.getNumBlockIDs(), 0);
  for (MachineBasicBlock &MBB : MF) {
    unsigned N = MBB.pred_size();
    if (MachineLoop *L = MLI.getLoopFor(&MBB))
      if (L->getHeader() == &MBB)
        for (const MachineBasicBlock *Pred : MBB.predecessors())
          if (L->contains(Pred))
            --N;
    NumPredsLeft[MBB.getNumber()] = N;
  }

  // Topological sort the CFG, with additional constraints:
  //  - Between a loop header and the last block in the loop, there can be
  //    no blocks not dominated by the loop header.
  //  - It's desirable to preserve the original block order when possible.
  // We use two ready lists; Preferred and Ready. Preferred has recently
  // processed sucessors, to help preserve block sequences from the original
  // order. Ready has the remaining ready blocks.
  PriorityQueue<MachineBasicBlock *, std::vector<MachineBasicBlock *>,
                CompareBlockNumbers>
      Preferred;
  PriorityQueue<MachineBasicBlock *, std::vector<MachineBasicBlock *>,
                CompareBlockNumbersBackwards>
      Ready;
  SmallVector<Entry, 4> Loops;
  for (MachineBasicBlock *MBB = &MF.front();;) {
    const MachineLoop *L = MLI.getLoopFor(MBB);
    if (L) {
      // If MBB is a loop header, add it to the active loop list. We can't put
      // any blocks that it doesn't dominate until we see the end of the loop.
      if (L->getHeader() == MBB)
        Loops.push_back(Entry(L));
      // For each active loop the block is in, decrement the count. If MBB is
      // the last block in an active loop, take it off the list and pick up any
      // blocks deferred because the header didn't dominate them.
      for (Entry &E : Loops)
        if (E.Loop->contains(MBB) && --E.NumBlocksLeft == 0)
          for (auto DeferredBlock : E.Deferred)
            Ready.push(DeferredBlock);
      while (!Loops.empty() && Loops.back().NumBlocksLeft == 0)
        Loops.pop_back();
    }
    // The main topological sort logic.
    for (MachineBasicBlock *Succ : MBB->successors()) {
      // Ignore backedges.
      if (MachineLoop *SuccL = MLI.getLoopFor(Succ))
        if (SuccL->getHeader() == Succ && SuccL->contains(MBB))
          continue;
      // Decrement the predecessor count. If it's now zero, it's ready.
      if (--NumPredsLeft[Succ->getNumber()] == 0)
        Preferred.push(Succ);
    }
    // Determine the block to follow MBB. First try to find a preferred block,
    // to preserve the original block order when possible.
    MachineBasicBlock *Next = nullptr;
    while (!Preferred.empty()) {
      Next = Preferred.top();
      Preferred.pop();
      // If X isn't dominated by the top active loop header, defer it until that
      // loop is done.
      if (!Loops.empty() &&
          !MDT.dominates(Loops.back().Loop->getHeader(), Next)) {
        Loops.back().Deferred.push_back(Next);
        Next = nullptr;
        continue;
      }
      // If Next was originally ordered before MBB, and it isn't because it was
      // loop-rotated above the header, it's not preferred.
      if (Next->getNumber() < MBB->getNumber() &&
          (!L || !L->contains(Next) ||
           L->getHeader()->getNumber() < Next->getNumber())) {
        Ready.push(Next);
        Next = nullptr;
        continue;
      }
      break;
    }
    // If we didn't find a suitable block in the Preferred list, check the
    // general Ready list.
    if (!Next) {
      // If there are no more blocks to process, we're done.
      if (Ready.empty()) {
        MaybeUpdateTerminator(MBB);
        break;
      }
      for (;;) {
        Next = Ready.top();
        Ready.pop();
        // If Next isn't dominated by the top active loop header, defer it until
        // that loop is done.
        if (!Loops.empty() &&
//.........这里部分代码省略.........

/// \brief Figure out if the loop is worth full unrolling.
///
/// Complete loop unrolling can make some loads constant, and we need to know
/// if that would expose any further optimization opportunities.  This routine
/// estimates this optimization.  It computes cost of unrolled loop
/// (UnrolledCost) and dynamic cost of the original loop (RolledDynamicCost). By
/// dynamic cost we mean that we won't count costs of blocks that are known not
/// to be executed (i.e. if we have a branch in the loop and we know that at the
/// given iteration its condition would be resolved to true, we won't add up the
/// cost of the 'false'-block).
/// \returns Optional value, holding the RolledDynamicCost and UnrolledCost. If
/// the analysis failed (no benefits expected from the unrolling, or the loop is
/// too big to analyze), the returned value is None.
Optional<EstimatedUnrollCost>
analyzeLoopUnrollCost(const Loop *L, unsigned TripCount, ScalarEvolution &SE,
                      const TargetTransformInfo &TTI,
                      unsigned MaxUnrolledLoopSize) {
  // We want to be able to scale offsets by the trip count and add more offsets
  // to them without checking for overflows, and we already don't want to
  // analyze *massive* trip counts, so we force the max to be reasonably small.
  assert(UnrollMaxIterationsCountToAnalyze < (INT_MAX / 2) &&
         "The unroll iterations max is too large!");

  // Don't simulate loops with a big or unknown tripcount
  if (!UnrollMaxIterationsCountToAnalyze || !TripCount ||
      TripCount > UnrollMaxIterationsCountToAnalyze)
    return None;

  SmallSetVector<BasicBlock *, 16> BBWorklist;
  DenseMap<Value *, Constant *> SimplifiedValues;

  // The estimated cost of the unrolled form of the loop. We try to estimate
  // this by simplifying as much as we can while computing the estimate.
  unsigned UnrolledCost = 0;
  // We also track the estimated dynamic (that is, actually executed) cost in
  // the rolled form. This helps identify cases when the savings from unrolling
  // aren't just exposing dead control flows, but actual reduced dynamic
  // instructions due to the simplifications which we expect to occur after
  // unrolling.
  unsigned RolledDynamicCost = 0;

  // Simulate execution of each iteration of the loop counting instructions,
  // which would be simplified.
  // Since the same load will take different values on different iterations,
  // we literally have to go through all loop's iterations.
  for (unsigned Iteration = 0; Iteration < TripCount; ++Iteration) {
    SimplifiedValues.clear();
    UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, L, SE);

    BBWorklist.clear();
    BBWorklist.insert(L->getHeader());
    // Note that we *must not* cache the size, this loop grows the worklist.
    for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
      BasicBlock *BB = BBWorklist[Idx];

      // Visit all instructions in the given basic block and try to simplify
      // it.  We don't change the actual IR, just count optimization
      // opportunities.
      for (Instruction &I : *BB) {
        unsigned InstCost = TTI.getUserCost(&I);

        // Visit the instruction to analyze its loop cost after unrolling,
        // and if the visitor returns false, include this instruction in the
        // unrolled cost.
        if (!Analyzer.visit(I))
          UnrolledCost += InstCost;

        // Also track this instructions expected cost when executing the rolled
        // loop form.
        RolledDynamicCost += InstCost;

        // If unrolled body turns out to be too big, bail out.
        if (UnrolledCost > MaxUnrolledLoopSize)
          return None;
      }

      // Add BB's successors to the worklist.
      for (BasicBlock *Succ : successors(BB))
        if (L->contains(Succ))
          BBWorklist.insert(Succ);
    }

    // If we found no optimization opportunities on the first iteration, we
    // won't find them on later ones too.
    if (UnrolledCost == RolledDynamicCost)
      return None;
  }
  return {{UnrolledCost, RolledDynamicCost}};
}

//.........这里部分代码省略.........

    // Prepare for the iteration by collecting any simplified entry or backedge
    // inputs.
    for (Instruction &I : *L->getHeader()) {
      auto *PHI = dyn_cast<PHINode>(&I);
      if (!PHI)
        break;

      // The loop header PHI nodes must have exactly two input: one from the
      // loop preheader and one from the loop latch.
      assert(
          PHI->getNumIncomingValues() == 2 &&
          "Must have an incoming value only for the preheader and the latch.");

      Value *V = PHI->getIncomingValueForBlock(
          Iteration == 0 ? L->getLoopPreheader() : L->getLoopLatch());
      Constant *C = dyn_cast<Constant>(V);
      if (Iteration != 0 && !C)
        C = SimplifiedValues.lookup(V);
      if (C)
        SimplifiedInputValues.push_back({PHI, C});
    }

    // Now clear and re-populate the map for the next iteration.
    SimplifiedValues.clear();
    while (!SimplifiedInputValues.empty())
      SimplifiedValues.insert(SimplifiedInputValues.pop_back_val());

    UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, SE, L);

    BBWorklist.clear();
    BBWorklist.insert(L->getHeader());
    // Note that we *must not* cache the size, this loop grows the worklist.
    for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
      BasicBlock *BB = BBWorklist[Idx];

      // Visit all instructions in the given basic block and try to simplify
      // it.  We don't change the actual IR, just count optimization
      // opportunities.
      for (Instruction &I : *BB) {
        // Track this instruction's expected baseline cost when executing the
        // rolled loop form.
        RolledDynamicCost += TTI.getUserCost(&I);

        // Visit the instruction to analyze its loop cost after unrolling,
        // and if the visitor returns true, mark the instruction as free after
        // unrolling and continue.
        bool IsFree = Analyzer.visit(I);
        bool Inserted = InstCostMap.insert({&I, (int)Iteration,
                                           (unsigned)IsFree,
                                           /*IsCounted*/ false}).second;
        (void)Inserted;
        assert(Inserted && "Cannot have a state for an unvisited instruction!");

        if (IsFree)
          continue;

        // If the instruction might have a side-effect recursively account for
        // the cost of it and all the instructions leading up to it.
        if (I.mayHaveSideEffects())
          AddCostRecursively(I, Iteration);

        // Can't properly model a cost of a call.
        // FIXME: With a proper cost model we should be able to do it.
        if(isa<CallInst>(&I))
          return None;

/// This works like CloneAndPruneFunctionInto, except that it does not clone the
/// entire function. Instead it starts at an instruction provided by the caller
/// and copies (and prunes) only the code reachable from that instruction.
void llvm::CloneAndPruneIntoFromInst(Function *NewFunc, const Function *OldFunc,
                                     const Instruction *StartingInst,
                                     ValueToValueMapTy &VMap,
                                     bool ModuleLevelChanges,
                                     SmallVectorImpl<ReturnInst *> &Returns,
                                     const char *NameSuffix,
                                     ClonedCodeInfo *CodeInfo) {
  assert(NameSuffix && "NameSuffix cannot be null!");

  ValueMapTypeRemapper *TypeMapper = nullptr;
  ValueMaterializer *Materializer = nullptr;

#ifndef NDEBUG
  // If the cloning starts at the beginning of the function, verify that
  // the function arguments are mapped.
  if (!StartingInst)
    for (const Argument &II : OldFunc->args())
      assert(VMap.count(&II) && "No mapping from source argument specified!");
#endif

  PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, ModuleLevelChanges,
                            NameSuffix, CodeInfo);
  const BasicBlock *StartingBB;
  if (StartingInst)
    StartingBB = StartingInst->getParent();
  else {
    StartingBB = &OldFunc->getEntryBlock();
    StartingInst = &StartingBB->front();
  }

  // Clone the entry block, and anything recursively reachable from it.
  std::vector<const BasicBlock*> CloneWorklist;
  PFC.CloneBlock(StartingBB, StartingInst->getIterator(), CloneWorklist);
  while (!CloneWorklist.empty()) {
    const BasicBlock *BB = CloneWorklist.back();
    CloneWorklist.pop_back();
    PFC.CloneBlock(BB, BB->begin(), CloneWorklist);
  }
  
  // Loop over all of the basic blocks in the old function.  If the block was
  // reachable, we have cloned it and the old block is now in the value map:
  // insert it into the new function in the right order.  If not, ignore it.
  //
  // Defer PHI resolution until rest of function is resolved.
  SmallVector<const PHINode*, 16> PHIToResolve;
  for (const BasicBlock &BI : *OldFunc) {
    Value *V = VMap.lookup(&BI);
    BasicBlock *NewBB = cast_or_null<BasicBlock>(V);
    if (!NewBB) continue;  // Dead block.

    // Add the new block to the new function.
    NewFunc->getBasicBlockList().push_back(NewBB);

    // Handle PHI nodes specially, as we have to remove references to dead
    // blocks.
    for (BasicBlock::const_iterator I = BI.begin(), E = BI.end(); I != E; ++I) {
      // PHI nodes may have been remapped to non-PHI nodes by the caller or
      // during the cloning process.
      if (const PHINode *PN = dyn_cast<PHINode>(I)) {
        if (isa<PHINode>(VMap[PN]))
          PHIToResolve.push_back(PN);
        else
          break;
      } else {
        break;
      }
    }

    // Finally, remap the terminator instructions, as those can't be remapped
    // until all BBs are mapped.
    RemapInstruction(NewBB->getTerminator(), VMap,
                     ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges,
                     TypeMapper, Materializer);
  }
  
  // Defer PHI resolution until rest of function is resolved, PHI resolution
  // requires the CFG to be up-to-date.
  for (unsigned phino = 0, e = PHIToResolve.size(); phino != e; ) {
    const PHINode *OPN = PHIToResolve[phino];
    unsigned NumPreds = OPN->getNumIncomingValues();
    const BasicBlock *OldBB = OPN->getParent();
    BasicBlock *NewBB = cast<BasicBlock>(VMap[OldBB]);

    // Map operands for blocks that are live and remove operands for blocks
    // that are dead.
    for (; phino != PHIToResolve.size() &&
         PHIToResolve[phino]->getParent() == OldBB; ++phino) {
      OPN = PHIToResolve[phino];
      PHINode *PN = cast<PHINode>(VMap[OPN]);
      for (unsigned pred = 0, e = NumPreds; pred != e; ++pred) {
        Value *V = VMap.lookup(PN->getIncomingBlock(pred));
        if (BasicBlock *MappedBlock = cast_or_null<BasicBlock>(V)) {
          Value *InVal = MapValue(PN->getIncomingValue(pred),
                                  VMap, 
                        ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges);
          assert(InVal && "Unknown input value?");
          PN->setIncomingValue(pred, InVal);
//.........这里部分代码省略.........

void AAEvaluator::runInternal(Function &F, AAResults &AA) {
  const DataLayout &DL = F.getParent()->getDataLayout();

  ++FunctionCount;

  SetVector<Value *> Pointers;
  SmallSetVector<CallBase *, 16> Calls;
  SetVector<Value *> Loads;
  SetVector<Value *> Stores;

  for (auto &I : F.args())
    if (I.getType()->isPointerTy())    // Add all pointer arguments.
      Pointers.insert(&I);

  for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
    if (I->getType()->isPointerTy()) // Add all pointer instructions.
      Pointers.insert(&*I);
    if (EvalAAMD && isa<LoadInst>(&*I))
      Loads.insert(&*I);
    if (EvalAAMD && isa<StoreInst>(&*I))
      Stores.insert(&*I);
    Instruction &Inst = *I;
    if (auto *Call = dyn_cast<CallBase>(&Inst)) {
      Value *Callee = Call->getCalledValue();
      // Skip actual functions for direct function calls.
      if (!isa<Function>(Callee) && isInterestingPointer(Callee))
        Pointers.insert(Callee);
      // Consider formals.
      for (Use &DataOp : Call->data_ops())
        if (isInterestingPointer(DataOp))
          Pointers.insert(DataOp);
      Calls.insert(Call);
    } else {
      // Consider all operands.
      for (Instruction::op_iterator OI = Inst.op_begin(), OE = Inst.op_end();
           OI != OE; ++OI)
        if (isInterestingPointer(*OI))
          Pointers.insert(*OI);
    }
  }

  if (PrintAll || PrintNoAlias || PrintMayAlias || PrintPartialAlias ||
      PrintMustAlias || PrintNoModRef || PrintMod || PrintRef || PrintModRef)
    errs() << "Function: " << F.getName() << ": " << Pointers.size()
           << " pointers, " << Calls.size() << " call sites\n";

  // iterate over the worklist, and run the full (n^2)/2 disambiguations
  for (SetVector<Value *>::iterator I1 = Pointers.begin(), E = Pointers.end();
       I1 != E; ++I1) {
    auto I1Size = LocationSize::unknown();
    Type *I1ElTy = cast<PointerType>((*I1)->getType())->getElementType();
    if (I1ElTy->isSized())
      I1Size = LocationSize::precise(DL.getTypeStoreSize(I1ElTy));

    for (SetVector<Value *>::iterator I2 = Pointers.begin(); I2 != I1; ++I2) {
      auto I2Size = LocationSize::unknown();
      Type *I2ElTy = cast<PointerType>((*I2)->getType())->getElementType();
      if (I2ElTy->isSized())
        I2Size = LocationSize::precise(DL.getTypeStoreSize(I2ElTy));

      AliasResult AR = AA.alias(*I1, I1Size, *I2, I2Size);
      switch (AR) {
      case NoAlias:
        PrintResults(AR, PrintNoAlias, *I1, *I2, F.getParent());
        ++NoAliasCount;
        break;
      case MayAlias:
        PrintResults(AR, PrintMayAlias, *I1, *I2, F.getParent());
        ++MayAliasCount;
        break;
      case PartialAlias:
        PrintResults(AR, PrintPartialAlias, *I1, *I2, F.getParent());
        ++PartialAliasCount;
        break;
      case MustAlias:
        PrintResults(AR, PrintMustAlias, *I1, *I2, F.getParent());
        ++MustAliasCount;
        break;
      }
    }
  }

  if (EvalAAMD) {
    // iterate over all pairs of load, store
    for (Value *Load : Loads) {
      for (Value *Store : Stores) {
        AliasResult AR = AA.alias(MemoryLocation::get(cast<LoadInst>(Load)),
                                  MemoryLocation::get(cast<StoreInst>(Store)));
        switch (AR) {
        case NoAlias:
          PrintLoadStoreResults(AR, PrintNoAlias, Load, Store, F.getParent());
          ++NoAliasCount;
          break;
        case MayAlias:
          PrintLoadStoreResults(AR, PrintMayAlias, Load, Store, F.getParent());
          ++MayAliasCount;
          break;
        case PartialAlias:
          PrintLoadStoreResults(AR, PrintPartialAlias, Load, Store, F.getParent());
          ++PartialAliasCount;
//.........这里部分代码省略.........

void MCObjectDisassembler::buildCFG(MCModule *Module) {
  typedef std::map<uint64_t, BBInfo> BBInfoByAddrTy;
  BBInfoByAddrTy BBInfos;
  AddressSetTy Splits;
  AddressSetTy Calls;

  error_code ec;
  for (symbol_iterator SI = Obj.begin_symbols(), SE = Obj.end_symbols();
       SI != SE; SI.increment(ec)) {
    if (ec)
      break;
    SymbolRef::Type SymType;
    SI->getType(SymType);
    if (SymType == SymbolRef::ST_Function) {
      uint64_t SymAddr;
      SI->getAddress(SymAddr);
      SymAddr = getEffectiveLoadAddr(SymAddr);
      Calls.push_back(SymAddr);
      Splits.push_back(SymAddr);
    }
  }

  assert(Module->func_begin() == Module->func_end()
         && "Module already has a CFG!");

  // First, determine the basic block boundaries and call targets.
  for (MCModule::atom_iterator AI = Module->atom_begin(),
                               AE = Module->atom_end();
       AI != AE; ++AI) {
    MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
    if (!TA) continue;
    Calls.push_back(TA->getBeginAddr());
    BBInfos[TA->getBeginAddr()].Atom = TA;
    for (MCTextAtom::const_iterator II = TA->begin(), IE = TA->end();
         II != IE; ++II) {
      if (MIA.isTerminator(II->Inst))
        Splits.push_back(II->Address + II->Size);
      uint64_t Target;
      if (MIA.evaluateBranch(II->Inst, II->Address, II->Size, Target)) {
        if (MIA.isCall(II->Inst))
          Calls.push_back(Target);
        Splits.push_back(Target);
      }
    }
  }

  RemoveDupsFromAddressVector(Splits);
  RemoveDupsFromAddressVector(Calls);

  // Split text atoms into basic block atoms.
  for (AddressSetTy::const_iterator SI = Splits.begin(), SE = Splits.end();
       SI != SE; ++SI) {
    MCAtom *A = Module->findAtomContaining(*SI);
    if (!A) continue;
    MCTextAtom *TA = cast<MCTextAtom>(A);
    if (TA->getBeginAddr() == *SI)
      continue;
    MCTextAtom *NewAtom = TA->split(*SI);
    BBInfos[NewAtom->getBeginAddr()].Atom = NewAtom;
    StringRef BBName = TA->getName();
    BBName = BBName.substr(0, BBName.find_last_of(':'));
    NewAtom->setName((BBName + ":" + utohexstr(*SI)).str());
  }

  // Compute succs/preds.
  for (MCModule::atom_iterator AI = Module->atom_begin(),
                               AE = Module->atom_end();
                               AI != AE; ++AI) {
    MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
    if (!TA) continue;
    BBInfo &CurBB = BBInfos[TA->getBeginAddr()];
    const MCDecodedInst &LI = TA->back();
    if (MIA.isBranch(LI.Inst)) {
      uint64_t Target;
      if (MIA.evaluateBranch(LI.Inst, LI.Address, LI.Size, Target))
        CurBB.addSucc(BBInfos[Target]);
      if (MIA.isConditionalBranch(LI.Inst))
        CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
    } else if (!MIA.isTerminator(LI.Inst))
      CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
  }


  // Create functions and basic blocks.
  for (AddressSetTy::const_iterator CI = Calls.begin(), CE = Calls.end();
       CI != CE; ++CI) {
    BBInfo &BBI = BBInfos[*CI];
    if (!BBI.Atom) continue;

    MCFunction &MCFN = *Module->createFunction(BBI.Atom->getName());

    // Create MCBBs.
    SmallSetVector<BBInfo*, 16> Worklist;
    Worklist.insert(&BBI);
    for (size_t wi = 0; wi < Worklist.size(); ++wi) {
      BBInfo *BBI = Worklist[wi];
      if (!BBI->Atom)
        continue;
      BBI->BB = &MCFN.createBlock(*BBI->Atom);
      // Add all predecessors and successors to the worklist.
//.........这里部分代码省略.........

//.........这里部分代码省略.........
      continue;
    }

    // If there is no existing flag with this ID, just add it.
    if (!DstOp) {
      Flags[ID] = std::make_pair(SrcOp, DstModFlags->getNumOperands());
      DstModFlags->addOperand(SrcOp);
      continue;
    }

    // Otherwise, perform a merge.
    ConstantInt *DstBehavior =
        mdconst::extract<ConstantInt>(DstOp->getOperand(0));
    unsigned DstBehaviorValue = DstBehavior->getZExtValue();

    // If either flag has override behavior, handle it first.
    if (DstBehaviorValue == Module::Override) {
      // Diagnose inconsistent flags which both have override behavior.
      if (SrcBehaviorValue == Module::Override &&
          SrcOp->getOperand(2) != DstOp->getOperand(2))
        return stringErr("linking module flags '" + ID->getString() +
                         "': IDs have conflicting override values");
      continue;
    } else if (SrcBehaviorValue == Module::Override) {
      // Update the destination flag to that of the source.
      DstModFlags->setOperand(DstIndex, SrcOp);
      Flags[ID].first = SrcOp;
      continue;
    }

    // Diagnose inconsistent merge behavior types.
    if (SrcBehaviorValue != DstBehaviorValue)
      return stringErr("linking module flags '" + ID->getString() +
                       "': IDs have conflicting behaviors");

    auto replaceDstValue = [&](MDNode *New) {
      Metadata *FlagOps[] = {DstOp->getOperand(0), ID, New};
      MDNode *Flag = MDNode::get(DstM.getContext(), FlagOps);
      DstModFlags->setOperand(DstIndex, Flag);
      Flags[ID].first = Flag;
    };

    // Perform the merge for standard behavior types.
    switch (SrcBehaviorValue) {
    case Module::Require:
    case Module::Override:
      llvm_unreachable("not possible");
    case Module::Error: {
      // Emit an error if the values differ.
      if (SrcOp->getOperand(2) != DstOp->getOperand(2))
        return stringErr("linking module flags '" + ID->getString() +
                         "': IDs have conflicting values");
      continue;
    }
    case Module::Warning: {
      // Emit a warning if the values differ.
      if (SrcOp->getOperand(2) != DstOp->getOperand(2)) {
        emitWarning("linking module flags '" + ID->getString() +
                    "': IDs have conflicting values");
      }
      continue;
    }
    case Module::Append: {
      MDNode *DstValue = cast<MDNode>(DstOp->getOperand(2));
      MDNode *SrcValue = cast<MDNode>(SrcOp->getOperand(2));
      SmallVector<Metadata *, 8> MDs;
      MDs.reserve(DstValue->getNumOperands() + SrcValue->getNumOperands());
      MDs.append(DstValue->op_begin(), DstValue->op_end());
      MDs.append(SrcValue->op_begin(), SrcValue->op_end());

      replaceDstValue(MDNode::get(DstM.getContext(), MDs));
      break;
    }
    case Module::AppendUnique: {
      SmallSetVector<Metadata *, 16> Elts;
      MDNode *DstValue = cast<MDNode>(DstOp->getOperand(2));
      MDNode *SrcValue = cast<MDNode>(SrcOp->getOperand(2));
      Elts.insert(DstValue->op_begin(), DstValue->op_end());
      Elts.insert(SrcValue->op_begin(), SrcValue->op_end());

      replaceDstValue(MDNode::get(DstM.getContext(),
                                  makeArrayRef(Elts.begin(), Elts.end())));
      break;
    }
    }
  }

  // Check all of the requirements.
  for (unsigned I = 0, E = Requirements.size(); I != E; ++I) {
    MDNode *Requirement = Requirements[I];
    MDString *Flag = cast<MDString>(Requirement->getOperand(0));
    Metadata *ReqValue = Requirement->getOperand(1);

    MDNode *Op = Flags[Flag].first;
    if (!Op || Op->getOperand(2) != ReqValue)
      return stringErr("linking module flags '" + Flag->getString() +
                       "': does not have the required value");
  }
  return Error::success();
}

/// IsFunctionMallocLike - A function is malloc-like if it returns either null
/// or a pointer that doesn't alias any other pointer visible to the caller.
bool FunctionAttrs::IsFunctionMallocLike(Function *F,
                              SmallPtrSet<Function*, 8> &SCCNodes) const {
  SmallSetVector<Value *, 8> FlowsToReturn;
  for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I)
    if (ReturnInst *Ret = dyn_cast<ReturnInst>(I->getTerminator()))
      FlowsToReturn.insert(Ret->getReturnValue());

  for (unsigned i = 0; i != FlowsToReturn.size(); ++i) {
    Value *RetVal = FlowsToReturn[i];

    if (Constant *C = dyn_cast<Constant>(RetVal)) {
      if (!C->isNullValue() && !isa<UndefValue>(C))
        return false;

      continue;
    }

    if (isa<Argument>(RetVal))
      return false;

    if (Instruction *RVI = dyn_cast<Instruction>(RetVal))
      switch (RVI->getOpcode()) {
        // Extend the analysis by looking upwards.
        case Instruction::BitCast:
        case Instruction::GetElementPtr:
          FlowsToReturn.insert(RVI->getOperand(0));
          continue;
        case Instruction::Select: {
          SelectInst *SI = cast<SelectInst>(RVI);
          FlowsToReturn.insert(SI->getTrueValue());
          FlowsToReturn.insert(SI->getFalseValue());
          continue;
        }
        case Instruction::PHI: {
          PHINode *PN = cast<PHINode>(RVI);
          for (int i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
            FlowsToReturn.insert(PN->getIncomingValue(i));
          continue;
        }

        // Check whether the pointer came from an allocation.
        case Instruction::Alloca:
          break;
        case Instruction::Call:
        case Instruction::Invoke: {
          CallSite CS(RVI);
          if (CS.paramHasAttr(0, Attribute::NoAlias))
            break;
          if (CS.getCalledFunction() &&
              SCCNodes.count(CS.getCalledFunction()))
            break;
        } // fall-through
        default:
          return false;  // Did not come from an allocation.
      }

    if (PointerMayBeCaptured(RetVal, false, /*StoreCaptures=*/false))
      return false;
  }

  return true;
}

/// Sort the blocks in RPO, taking special care to make sure that loops are
/// contiguous even in the case of split backedges.
///
/// TODO: Determine whether RPO is actually worthwhile, or whether we should
/// move to just a stable-topological-sort-based approach that would preserve
/// more of the original order.
static void SortBlocks(MachineFunction &MF, const MachineLoopInfo &MLI) {
  // Note that we do our own RPO rather than using
  // "llvm/ADT/PostOrderIterator.h" because we want control over the order that
  // successors are visited in (see above). Also, we can sort the blocks in the
  // MachineFunction as we go.
  SmallPtrSet<MachineBasicBlock *, 16> Visited;
  SmallVector<POStackEntry, 16> Stack;

  MachineBasicBlock *EntryBlock = &*MF.begin();
  Visited.insert(EntryBlock);
  Stack.push_back(POStackEntry(EntryBlock, MF, MLI));

  for (;;) {
    POStackEntry &Entry = Stack.back();
    SmallVectorImpl<MachineBasicBlock *> &Succs = Entry.Succs;
    if (!Succs.empty()) {
      MachineBasicBlock *Succ = Succs.pop_back_val();
      if (Visited.insert(Succ).second)
        Stack.push_back(POStackEntry(Succ, MF, MLI));
      continue;
    }

    // Put the block in its position in the MachineFunction.
    MachineBasicBlock &MBB = *Entry.MBB;
    MBB.moveBefore(&*MF.begin());

    // Branch instructions may utilize a fallthrough, so update them if a
    // fallthrough has been added or removed.
    if (!MBB.empty() && MBB.back().isTerminator() && !MBB.back().isBranch() &&
        !MBB.back().isBarrier())
      report_fatal_error(
          "Non-branch terminator with fallthrough cannot yet be rewritten");
    if (MBB.empty() || !MBB.back().isTerminator() || MBB.back().isBranch())
      MBB.updateTerminator();

    Stack.pop_back();
    if (Stack.empty())
      break;
  }

  // Now that we've sorted the blocks in RPO, renumber them.
  MF.RenumberBlocks();

#ifndef NDEBUG
  SmallSetVector<MachineLoop *, 8> OnStack;

  // Insert a sentinel representing the degenerate loop that starts at the
  // function entry block and includes the entire function as a "loop" that
  // executes once.
  OnStack.insert(nullptr);

  for (auto &MBB : MF) {
    assert(MBB.getNumber() >= 0 && "Renumbered blocks should be non-negative.");

    MachineLoop *Loop = MLI.getLoopFor(&MBB);
    if (Loop && &MBB == Loop->getHeader()) {
      // Loop header. The loop predecessor should be sorted above, and the other
      // predecessors should be backedges below.
      for (auto Pred : MBB.predecessors())
        assert(
            (Pred->getNumber() < MBB.getNumber() || Loop->contains(Pred)) &&
            "Loop header predecessors must be loop predecessors or backedges");
      assert(OnStack.insert(Loop) && "Loops should be declared at most once.");
    } else {
      // Not a loop header. All predecessors should be sorted above.
      for (auto Pred : MBB.predecessors())
        assert(Pred->getNumber() < MBB.getNumber() &&
               "Non-loop-header predecessors should be topologically sorted");
      assert(OnStack.count(MLI.getLoopFor(&MBB)) &&
             "Blocks must be nested in their loops");
    }
    while (OnStack.size() > 1 && &MBB == LoopBottom(OnStack.back()))
      OnStack.pop_back();
  }
  assert(OnStack.pop_back_val() == nullptr &&
         "The function entry block shouldn't actually be a loop header");
  assert(OnStack.empty() &&
         "Control flow stack pushes and pops should be balanced.");
#endif
}

/// \brief Figure out if the loop is worth full unrolling.
///
/// Complete loop unrolling can make some loads constant, and we need to know
/// if that would expose any further optimization opportunities.  This routine
/// estimates this optimization.  It computes cost of unrolled loop
/// (UnrolledCost) and dynamic cost of the original loop (RolledDynamicCost). By
/// dynamic cost we mean that we won't count costs of blocks that are known not
/// to be executed (i.e. if we have a branch in the loop and we know that at the
/// given iteration its condition would be resolved to true, we won't add up the
/// cost of the 'false'-block).
/// \returns Optional value, holding the RolledDynamicCost and UnrolledCost. If
/// the analysis failed (no benefits expected from the unrolling, or the loop is
/// too big to analyze), the returned value is None.
static Optional<EstimatedUnrollCost>
analyzeLoopUnrollCost(const Loop *L, unsigned TripCount, DominatorTree &DT,
                      ScalarEvolution &SE, const TargetTransformInfo &TTI,
                      int MaxUnrolledLoopSize) {
  // We want to be able to scale offsets by the trip count and add more offsets
  // to them without checking for overflows, and we already don't want to
  // analyze *massive* trip counts, so we force the max to be reasonably small.
  assert(UnrollMaxIterationsCountToAnalyze < (INT_MAX / 2) &&
         "The unroll iterations max is too large!");

  // Don't simulate loops with a big or unknown tripcount
  if (!UnrollMaxIterationsCountToAnalyze || !TripCount ||
      TripCount > UnrollMaxIterationsCountToAnalyze)
    return None;

  SmallSetVector<BasicBlock *, 16> BBWorklist;
  DenseMap<Value *, Constant *> SimplifiedValues;
  SmallVector<std::pair<Value *, Constant *>, 4> SimplifiedInputValues;

  // The estimated cost of the unrolled form of the loop. We try to estimate
  // this by simplifying as much as we can while computing the estimate.
  int UnrolledCost = 0;
  // We also track the estimated dynamic (that is, actually executed) cost in
  // the rolled form. This helps identify cases when the savings from unrolling
  // aren't just exposing dead control flows, but actual reduced dynamic
  // instructions due to the simplifications which we expect to occur after
  // unrolling.
  int RolledDynamicCost = 0;

  // Ensure that we don't violate the loop structure invariants relied on by
  // this analysis.
  assert(L->isLoopSimplifyForm() && "Must put loop into normal form first.");
  assert(L->isLCSSAForm(DT) &&
         "Must have loops in LCSSA form to track live-out values.");

  DEBUG(dbgs() << "Starting LoopUnroll profitability analysis...\n");

  // Simulate execution of each iteration of the loop counting instructions,
  // which would be simplified.
  // Since the same load will take different values on different iterations,
  // we literally have to go through all loop's iterations.
  for (unsigned Iteration = 0; Iteration < TripCount; ++Iteration) {
    DEBUG(dbgs() << " Analyzing iteration " << Iteration << "\n");

    // Prepare for the iteration by collecting any simplified entry or backedge
    // inputs.
    for (Instruction &I : *L->getHeader()) {
      auto *PHI = dyn_cast<PHINode>(&I);
      if (!PHI)
        break;

      // The loop header PHI nodes must have exactly two input: one from the
      // loop preheader and one from the loop latch.
      assert(
          PHI->getNumIncomingValues() == 2 &&
          "Must have an incoming value only for the preheader and the latch.");

      Value *V = PHI->getIncomingValueForBlock(
          Iteration == 0 ? L->getLoopPreheader() : L->getLoopLatch());
      Constant *C = dyn_cast<Constant>(V);
      if (Iteration != 0 && !C)
        C = SimplifiedValues.lookup(V);
      if (C)
        SimplifiedInputValues.push_back({PHI, C});
    }

    // Now clear and re-populate the map for the next iteration.
    SimplifiedValues.clear();
    while (!SimplifiedInputValues.empty())
      SimplifiedValues.insert(SimplifiedInputValues.pop_back_val());

    UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, SE);

    BBWorklist.clear();
    BBWorklist.insert(L->getHeader());
    // Note that we *must not* cache the size, this loop grows the worklist.
    for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
      BasicBlock *BB = BBWorklist[Idx];

      // Visit all instructions in the given basic block and try to simplify
      // it.  We don't change the actual IR, just count optimization
      // opportunities.
      for (Instruction &I : *BB) {
        int InstCost = TTI.getUserCost(&I);

        // Visit the instruction to analyze its loop cost after unrolling,
        // and if the visitor returns false, include this instruction in the
//.........这里部分代码省略.........