C++ BasicBlock类代码示例

/// Unroll the given loop by Count. The loop must be in LCSSA form. Returns true
/// if unrolling was successful, or false if the loop was unmodified. Unrolling
/// can only fail when the loop's latch block is not terminated by a conditional
/// branch instruction. However, if the trip count (and multiple) are not known,
/// loop unrolling will mostly produce more code that is no faster.
///
/// TripCount is generally defined as the number of times the loop header
/// executes. UnrollLoop relaxes the definition to permit early exits: here
/// TripCount is the iteration on which control exits LatchBlock if no early
/// exits were taken. Note that UnrollLoop assumes that the loop counter test
/// terminates LatchBlock in order to remove unnecesssary instances of the
/// test. In other words, control may exit the loop prior to TripCount
/// iterations via an early branch, but control may not exit the loop from the
/// LatchBlock's terminator prior to TripCount iterations.
///
/// Similarly, TripMultiple divides the number of times that the LatchBlock may
/// execute without exiting the loop.
///
/// The LoopInfo Analysis that is passed will be kept consistent.
///
/// If a LoopPassManager is passed in, and the loop is fully removed, it will be
/// removed from the LoopPassManager as well. LPM can also be NULL.
///
/// This utility preserves LoopInfo. If DominatorTree or ScalarEvolution are
/// available it must also preserve those analyses.
bool llvm::UnrollLoop(Loop *L, unsigned Count, unsigned TripCount,
                      bool AllowRuntime, unsigned TripMultiple,
                      LoopInfo *LI, LPPassManager *LPM) {
  BasicBlock *Preheader = L->getLoopPreheader();
  if (!Preheader) {
    DEBUG(dbgs() << "  Can't unroll; loop preheader-insertion failed.\n");
    return false;
  }

  BasicBlock *LatchBlock = L->getLoopLatch();
  if (!LatchBlock) {
    DEBUG(dbgs() << "  Can't unroll; loop exit-block-insertion failed.\n");
    return false;
  }

  BasicBlock *Header = L->getHeader();
  BranchInst *BI = dyn_cast<BranchInst>(LatchBlock->getTerminator());

  if (!BI || BI->isUnconditional()) {
    // The loop-rotate pass can be helpful to avoid this in many cases.
    DEBUG(dbgs() <<
             "  Can't unroll; loop not terminated by a conditional branch.\n");
    return false;
  }

  if (Header->hasAddressTaken()) {
    // The loop-rotate pass can be helpful to avoid this in many cases.
    DEBUG(dbgs() <<
          "  Won't unroll loop: address of header block is taken.\n");
    return false;
  }

  if (TripCount != 0)
    DEBUG(dbgs() << "  Trip Count = " << TripCount << "\n");
  if (TripMultiple != 1)
    DEBUG(dbgs() << "  Trip Multiple = " << TripMultiple << "\n");

  // Effectively "DCE" unrolled iterations that are beyond the tripcount
  // and will never be executed.
  if (TripCount != 0 && Count > TripCount)
    Count = TripCount;

  // Don't enter the unroll code if there is nothing to do. This way we don't
  // need to support "partial unrolling by 1".
  if (TripCount == 0 && Count < 2)
    return false;

  assert(Count > 0);
  assert(TripMultiple > 0);
  assert(TripCount == 0 || TripCount % TripMultiple == 0);

  // Are we eliminating the loop control altogether?
  bool CompletelyUnroll = Count == TripCount;

  // We assume a run-time trip count if the compiler cannot
  // figure out the loop trip count and the unroll-runtime
  // flag is specified.
  bool RuntimeTripCount = (TripCount == 0 && Count > 0 && AllowRuntime);

  if (RuntimeTripCount && !UnrollRuntimeLoopProlog(L, Count, LI, LPM))
    return false;

  // Notify ScalarEvolution that the loop will be substantially changed,
  // if not outright eliminated.
  ScalarEvolution *SE = LPM->getAnalysisIfAvailable<ScalarEvolution>();
  if (SE)
    SE->forgetLoop(L);

  // If we know the trip count, we know the multiple...
  unsigned BreakoutTrip = 0;
  if (TripCount != 0) {
    BreakoutTrip = TripCount % Count;
    TripMultiple = 0;
  } else {
    // Figure out what multiple to use.
//.........这里部分代码省略.........

/// \brief Recursively handle the condition leading to a loop
Value *SIAnnotateControlFlow::handleLoopCondition(Value *Cond, PHINode *Broken,
                                             llvm::Loop *L, BranchInst *Term) {

  // Only search through PHI nodes which are inside the loop.  If we try this
  // with PHI nodes that are outside of the loop, we end up inserting new PHI
  // nodes outside of the loop which depend on values defined inside the loop.
  // This will break the module with
  // 'Instruction does not dominate all users!' errors.
  PHINode *Phi = nullptr;
  if ((Phi = dyn_cast<PHINode>(Cond)) && L->contains(Phi)) {

    BasicBlock *Parent = Phi->getParent();
    PHINode *NewPhi = PHINode::Create(Int64, 0, "", &Parent->front());
    Value *Ret = NewPhi;

    // Handle all non-constant incoming values first
    for (unsigned i = 0, e = Phi->getNumIncomingValues(); i != e; ++i) {
      Value *Incoming = Phi->getIncomingValue(i);
      BasicBlock *From = Phi->getIncomingBlock(i);
      if (isa<ConstantInt>(Incoming)) {
        NewPhi->addIncoming(Broken, From);
        continue;
      }

      Phi->setIncomingValue(i, BoolFalse);
      Value *PhiArg = handleLoopCondition(Incoming, Broken, L, Term);
      NewPhi->addIncoming(PhiArg, From);
    }

    BasicBlock *IDom = DT->getNode(Parent)->getIDom()->getBlock();

    for (unsigned i = 0, e = Phi->getNumIncomingValues(); i != e; ++i) {

      Value *Incoming = Phi->getIncomingValue(i);
      if (Incoming != BoolTrue)
        continue;

      BasicBlock *From = Phi->getIncomingBlock(i);
      if (From == IDom) {
        // We're in the following situation:
        //   IDom/From
        //      |   \
        //      |   If-block
        //      |   /
        //     Parent
        // where we want to break out of the loop if the If-block is not taken.
        // Due to the depth-first traversal, there should be an end.cf
        // intrinsic in Parent, and we insert an else.break before it.
        //
        // Note that the end.cf need not be the first non-phi instruction
        // of parent, particularly when we're dealing with a multi-level
        // break, but it should occur within a group of intrinsic calls
        // at the beginning of the block.
        CallInst *OldEnd = dyn_cast<CallInst>(Parent->getFirstInsertionPt());
        while (OldEnd && OldEnd->getCalledFunction() != EndCf)
          OldEnd = dyn_cast<CallInst>(OldEnd->getNextNode());
        if (OldEnd && OldEnd->getCalledFunction() == EndCf) {
          Value *Args[] = { OldEnd->getArgOperand(0), NewPhi };
          Ret = CallInst::Create(ElseBreak, Args, "", OldEnd);
          continue;
        }
      }
      TerminatorInst *Insert = From->getTerminator();
      Value *PhiArg = CallInst::Create(Break, Broken, "", Insert);
      NewPhi->setIncomingValue(i, PhiArg);
    }
    eraseIfUnused(Phi);
    return Ret;

  } else if (Instruction *Inst = dyn_cast<Instruction>(Cond)) {
    BasicBlock *Parent = Inst->getParent();
    Instruction *Insert;
    if (L->contains(Inst)) {
      Insert = Parent->getTerminator();
    } else {
      Insert = L->getHeader()->getFirstNonPHIOrDbgOrLifetime();
    }
    Value *Args[] = { Cond, Broken };
    return CallInst::Create(IfBreak, Args, "", Insert);

  // Insert IfBreak before TERM for constant COND.
  } else if (isa<ConstantInt>(Cond)) {
    Value *Args[] = { Cond, Broken };
    return CallInst::Create(IfBreak, Args, "", Term);

  } else {
    llvm_unreachable("Unhandled loop condition!");
  }
  return nullptr;
}

/// setupEntryBlockAndCallSites - Setup the entry block by creating and filling
/// the function context and marking the call sites with the appropriate
/// values. These values are used by the DWARF EH emitter.
bool SjLjEHPrepare::setupEntryBlockAndCallSites(Function &F) {
  SmallVector<ReturnInst *, 16> Returns;
  SmallVector<InvokeInst *, 16> Invokes;
  SmallSetVector<LandingPadInst *, 16> LPads;

  // Look through the terminators of the basic blocks to find invokes.
  for (BasicBlock &BB : F)
    if (auto *II = dyn_cast<InvokeInst>(BB.getTerminator())) {
      if (Function *Callee = II->getCalledFunction())
        if (Callee->getIntrinsicID() == Intrinsic::donothing) {
          // Remove the NOP invoke.
          BranchInst::Create(II->getNormalDest(), II);
          II->eraseFromParent();
          continue;
        }

      Invokes.push_back(II);
      LPads.insert(II->getUnwindDest()->getLandingPadInst());
    } else if (auto *RI = dyn_cast<ReturnInst>(BB.getTerminator())) {
      Returns.push_back(RI);
    }

  if (Invokes.empty())
    return false;

  NumInvokes += Invokes.size();

  lowerIncomingArguments(F);
  lowerAcrossUnwindEdges(F, Invokes);

  Value *FuncCtx =
      setupFunctionContext(F, makeArrayRef(LPads.begin(), LPads.end()));
  BasicBlock *EntryBB = &F.front();
  IRBuilder<> Builder(EntryBB->getTerminator());

  // Get a reference to the jump buffer.
  Value *JBufPtr =
      Builder.CreateConstGEP2_32(FunctionContextTy, FuncCtx, 0, 5, "jbuf_gep");

  // Save the frame pointer.
  Value *FramePtr = Builder.CreateConstGEP2_32(doubleUnderJBufTy, JBufPtr, 0, 0,
                                               "jbuf_fp_gep");

  Value *Val = Builder.CreateCall(FrameAddrFn, Builder.getInt32(0), "fp");
  Builder.CreateStore(Val, FramePtr, /*isVolatile=*/true);

  // Save the stack pointer.
  Value *StackPtr = Builder.CreateConstGEP2_32(doubleUnderJBufTy, JBufPtr, 0, 2,
                                               "jbuf_sp_gep");

  Val = Builder.CreateCall(StackAddrFn, {}, "sp");
  Builder.CreateStore(Val, StackPtr, /*isVolatile=*/true);

  // Call the setup_dispatch instrinsic. It fills in the rest of the jmpbuf.
  Builder.CreateCall(BuiltinSetupDispatchFn, {});

  // Store a pointer to the function context so that the back-end will know
  // where to look for it.
  Value *FuncCtxArg = Builder.CreateBitCast(FuncCtx, Builder.getInt8PtrTy());
  Builder.CreateCall(FuncCtxFn, FuncCtxArg);

  // At this point, we are all set up, update the invoke instructions to mark
  // their call_site values.
  for (unsigned I = 0, E = Invokes.size(); I != E; ++I) {
    insertCallSiteStore(Invokes[I], I + 1);

    ConstantInt *CallSiteNum =
        ConstantInt::get(Type::getInt32Ty(F.getContext()), I + 1);

    // Record the call site value for the back end so it stays associated with
    // the invoke.
    CallInst::Create(CallSiteFn, CallSiteNum, "", Invokes[I]);
  }

  // Mark call instructions that aren't nounwind as no-action (call_site ==
  // -1). Skip the entry block, as prior to then, no function context has been
  // created for this function and any unexpected exceptions thrown will go
  // directly to the caller's context, which is what we want anyway, so no need
  // to do anything here.
  for (BasicBlock &BB : F) {
    if (&BB == &F.front())
      continue;
    for (Instruction &I : BB)
      if (I.mayThrow())
        insertCallSiteStore(&I, -1);
  }

  // Register the function context and make sure it's known to not throw
  CallInst *Register =
      CallInst::Create(RegisterFn, FuncCtx, "", EntryBB->getTerminator());
  Register->setDoesNotThrow();

  // Following any allocas not in the entry block, update the saved SP in the
  // jmpbuf to the new value.
  for (BasicBlock &BB : F) {
    if (&BB == &F.front())
      continue;
//.........这里部分代码省略.........

// Note: branch conditions, by definition, only have a chain user.
// This is why it should not be saved in a map for recall.
Value* ARMIREmitter::visitBRCOND(const SDNode *N) {
  // Get the address
  const ConstantSDNode *DestNode = dyn_cast<ConstantSDNode>(N->getOperand(0));
  if (!DestNode) {
    printError("visitBRCOND: Not a constant integer for branch!");
    return NULL;
  }

  uint64_t DestInt = DestNode->getSExtValue();
  uint64_t PC = Dec->getDisassembler()->getDebugOffset(N->getDebugLoc());
  // Note: pipeline is 8 bytes
  uint64_t Tgt = PC + 8 + DestInt;

  Function *F = IRB->GetInsertBlock()->getParent();
  BasicBlock *CurBB = IRB->GetInsertBlock();

  BasicBlock *BBTgt = Dec->getOrCreateBasicBlock(Tgt, F);

  // Parse the branch condition code
  const ConstantSDNode *CCNode = dyn_cast<ConstantSDNode>(N->getOperand(1));
  if (!CCNode) {
    printError("visitBRCOND: Condition code is not a constant integer!");
    return NULL;
  }
  ARMCC::CondCodes ARMcc = ARMCC::CondCodes(CCNode->getZExtValue());

  // Unconditional branch
  if (ARMcc == ARMCC::AL) {
    Instruction *Br = IRB->CreateBr(BBTgt);
    Br->setDebugLoc(N->getDebugLoc());
    return Br;
  }

  // If not a conditional branch, find the successor block and look at CC
  BasicBlock *NextBB = NULL;
  Function::iterator BI = F->begin(), BE= F->end();
  while (BI != BE && BI->getName() != CurBB->getName()) ++BI;
  ++BI;
  if (BI == BE) {               // NOTE: This should never happen...
    NextBB = Dec->getOrCreateBasicBlock("end", F);
  } else {
    NextBB = &(*BI);
  }


  SDNode *CPSR = N->getOperand(2)->getOperand(1).getNode();
  SDNode *CMPNode = NULL;
  for (SDNode::use_iterator I = CPSR->use_begin(), E = CPSR->use_end(); I != E;
       ++I) {
    if (I->getOpcode() == ISD::CopyToReg) {
      CMPNode = I->getOperand(2).getNode();
    }
  }

  if (CMPNode == NULL) {
    errs() << "ARMIREmitter ERROR: Could not find CMP SDNode for ARMBRCond!\n";
    return NULL;
  }

  Value *Cmp = NULL;
  Value *LHS = visit(CMPNode->getOperand(0).getNode());
  Value *RHS = visit(CMPNode->getOperand(1).getNode());
  // See ARMCC::CondCodes IntCCToARMCC(ISD::CondCode CC); in ARMISelLowering.cpp
  // TODO: Add support for conditions that handle floating point
  switch(ARMcc) {
    default:
      printError("Unknown condition code");
      return NULL;
    case ARMCC::EQ:
      Cmp = IRB->CreateICmpEQ(LHS, RHS);
      break;
    case ARMCC::NE:
      Cmp = IRB->CreateICmpNE(LHS, RHS);
      break;
    case ARMCC::HS:
      // HS - unsigned higher or same (or carry set)
      Cmp = IRB->CreateICmpUGE(LHS, RHS);
      break;
    case ARMCC::LO:
      // LO - unsigned lower (or carry clear)
      Cmp = IRB->CreateICmpULT(LHS, RHS);
      break;
    case ARMCC::MI:
      // MI - minus (negative)
      printError("Condition code MI is not handled at this time!");
      return NULL;
      // break;
    case ARMCC::PL:
      // PL - plus (positive or zero)
      printError("Condition code PL is not handled at this time!");
      return NULL;
      // break;
    case ARMCC::VS:
      // VS - V Set (signed overflow)
      printError("Condition code VS is not handled at this time!");
      return NULL;
      // break;
    case ARMCC::VC:
//.........这里部分代码省略.........

/// NormalizeLandingPads - Normalize and discover landing pads, noting them
/// in the LandingPads set.  A landing pad is normal if the only CFG edges
/// that end at it are unwind edges from invoke instructions. If we inlined
/// through an invoke we could have a normal branch from the previous
/// unwind block through to the landing pad for the original invoke.
/// Abnormal landing pads are fixed up by redirecting all unwind edges to
/// a new basic block which falls through to the original.
bool DwarfEHPrepare::NormalizeLandingPads() {
  bool Changed = false;

  const MCAsmInfo *MAI = TM->getMCAsmInfo();
  bool usingSjLjEH = MAI->getExceptionHandlingType() == ExceptionHandling::SjLj;

  for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I) {
    TerminatorInst *TI = I->getTerminator();
    if (!isa<InvokeInst>(TI))
      continue;
    BasicBlock *LPad = TI->getSuccessor(1);
    // Skip landing pads that have already been normalized.
    if (LandingPads.count(LPad))
      continue;

    // Check that only invoke unwind edges end at the landing pad.
    bool OnlyUnwoundTo = true;
    bool SwitchOK = usingSjLjEH;
    for (pred_iterator PI = pred_begin(LPad), PE = pred_end(LPad);
         PI != PE; ++PI) {
      TerminatorInst *PT = (*PI)->getTerminator();
      // The SjLj dispatch block uses a switch instruction. This is effectively
      // an unwind edge, so we can disregard it here. There will only ever
      // be one dispatch, however, so if there are multiple switches, one
      // of them truly is a normal edge, not an unwind edge.
      if (SwitchOK && isa<SwitchInst>(PT)) {
        SwitchOK = false;
        continue;
      }
      if (!isa<InvokeInst>(PT) || LPad == PT->getSuccessor(0)) {
        OnlyUnwoundTo = false;
        break;
      }
    }

    if (OnlyUnwoundTo) {
      // Only unwind edges lead to the landing pad.  Remember the landing pad.
      LandingPads.insert(LPad);
      continue;
    }

    // At least one normal edge ends at the landing pad.  Redirect the unwind
    // edges to a new basic block which falls through into this one.

    // Create the new basic block.
    BasicBlock *NewBB = BasicBlock::Create(F->getContext(),
                                           LPad->getName() + "_unwind_edge");

    // Insert it into the function right before the original landing pad.
    LPad->getParent()->getBasicBlockList().insert(LPad, NewBB);

    // Redirect unwind edges from the original landing pad to NewBB.
    for (pred_iterator PI = pred_begin(LPad), PE = pred_end(LPad); PI != PE; ) {
      TerminatorInst *PT = (*PI++)->getTerminator();
      if (isa<InvokeInst>(PT) && PT->getSuccessor(1) == LPad)
        // Unwind to the new block.
        PT->setSuccessor(1, NewBB);
    }

    // If there are any PHI nodes in LPad, we need to update them so that they
    // merge incoming values from NewBB instead.
    for (BasicBlock::iterator II = LPad->begin(); isa<PHINode>(II); ++II) {
      PHINode *PN = cast<PHINode>(II);
      pred_iterator PB = pred_begin(NewBB), PE = pred_end(NewBB);

      // Check to see if all of the values coming in via unwind edges are the
      // same.  If so, we don't need to create a new PHI node.
      Value *InVal = PN->getIncomingValueForBlock(*PB);
      for (pred_iterator PI = PB; PI != PE; ++PI) {
        if (PI != PB && InVal != PN->getIncomingValueForBlock(*PI)) {
          InVal = 0;
          break;
        }
      }

      if (InVal == 0) {
        // Different unwind edges have different values.  Create a new PHI node
        // in NewBB.
        PHINode *NewPN = PHINode::Create(PN->getType(), PN->getName()+".unwind",
                                         NewBB);
        // Add an entry for each unwind edge, using the value from the old PHI.
        for (pred_iterator PI = PB; PI != PE; ++PI)
          NewPN->addIncoming(PN->getIncomingValueForBlock(*PI), *PI);

        // Now use this new PHI as the common incoming value for NewBB in PN.
        InVal = NewPN;
      }

      // Revector exactly one entry in the PHI node to come from NewBB
      // and delete all other entries that come from unwind edges.  If
      // there are both normal and unwind edges from the same predecessor,
      // this leaves an entry for the normal edge.
      for (pred_iterator PI = PB; PI != PE; ++PI)
//.........这里部分代码省略.........

bool LoopInstSimplify::runOnLoop(Loop *L, LPPassManager &LPM) {
  DominatorTree *DT = getAnalysisIfAvailable<DominatorTree>();
  LoopInfo *LI = &getAnalysis<LoopInfo>();
  const TargetData *TD = getAnalysisIfAvailable<TargetData>();
  const TargetLibraryInfo *TLI = &getAnalysis<TargetLibraryInfo>();

  SmallVector<BasicBlock*, 8> ExitBlocks;
  L->getUniqueExitBlocks(ExitBlocks);
  array_pod_sort(ExitBlocks.begin(), ExitBlocks.end());

  SmallPtrSet<const Instruction*, 8> S1, S2, *ToSimplify = &S1, *Next = &S2;

  // The bit we are stealing from the pointer represents whether this basic
  // block is the header of a subloop, in which case we only process its phis.
  typedef PointerIntPair<BasicBlock*, 1> WorklistItem;
  SmallVector<WorklistItem, 16> VisitStack;
  SmallPtrSet<BasicBlock*, 32> Visited;

  bool Changed = false;
  bool LocalChanged;
  do {
    LocalChanged = false;

    VisitStack.clear();
    Visited.clear();

    VisitStack.push_back(WorklistItem(L->getHeader(), false));

    while (!VisitStack.empty()) {
      WorklistItem Item = VisitStack.pop_back_val();
      BasicBlock *BB = Item.getPointer();
      bool IsSubloopHeader = Item.getInt();

      // Simplify instructions in the current basic block.
      for (BasicBlock::iterator BI = BB->begin(), BE = BB->end(); BI != BE;) {
        Instruction *I = BI++;

        // The first time through the loop ToSimplify is empty and we try to
        // simplify all instructions. On later iterations ToSimplify is not
        // empty and we only bother simplifying instructions that are in it.
        if (!ToSimplify->empty() && !ToSimplify->count(I))
          continue;

        // Don't bother simplifying unused instructions.
        if (!I->use_empty()) {
          Value *V = SimplifyInstruction(I, TD, TLI, DT);
          if (V && LI->replacementPreservesLCSSAForm(I, V)) {
            // Mark all uses for resimplification next time round the loop.
            for (Value::use_iterator UI = I->use_begin(), UE = I->use_end();
                 UI != UE; ++UI)
              Next->insert(cast<Instruction>(*UI));

            I->replaceAllUsesWith(V);
            LocalChanged = true;
            ++NumSimplified;
          }
        }
        LocalChanged |= RecursivelyDeleteTriviallyDeadInstructions(I);

        if (IsSubloopHeader && !isa<PHINode>(I))
          break;
      }

      // Add all successors to the worklist, except for loop exit blocks and the
      // bodies of subloops. We visit the headers of loops so that we can process
      // their phis, but we contract the rest of the subloop body and only follow
      // edges leading back to the original loop.
      for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE;
           ++SI) {
        BasicBlock *SuccBB = *SI;
        if (!Visited.insert(SuccBB))
          continue;

        const Loop *SuccLoop = LI->getLoopFor(SuccBB);
        if (SuccLoop && SuccLoop->getHeader() == SuccBB
                     && L->contains(SuccLoop)) {
          VisitStack.push_back(WorklistItem(SuccBB, true));

          SmallVector<BasicBlock*, 8> SubLoopExitBlocks;
          SuccLoop->getExitBlocks(SubLoopExitBlocks);

          for (unsigned i = 0; i < SubLoopExitBlocks.size(); ++i) {
            BasicBlock *ExitBB = SubLoopExitBlocks[i];
            if (LI->getLoopFor(ExitBB) == L && Visited.insert(ExitBB))
              VisitStack.push_back(WorklistItem(ExitBB, false));
          }

          continue;
        }

        bool IsExitBlock = std::binary_search(ExitBlocks.begin(),
                                              ExitBlocks.end(), SuccBB);
        if (IsExitBlock)
          continue;

        VisitStack.push_back(WorklistItem(SuccBB, false));
      }
    }

    // Place the list of instructions to simplify on the next loop iteration
//.........这里部分代码省略.........

/// Create a clone of the blocks in a loop and connect them together.
/// If UnrollProlog is true, loop structure will not be cloned, otherwise a new
/// loop will be created including all cloned blocks, and the iterator of it
/// switches to count NewIter down to 0.
///
static void CloneLoopBlocks(Loop *L, Value *NewIter, const bool UnrollProlog,
                            BasicBlock *InsertTop, BasicBlock *InsertBot,
                            std::vector<BasicBlock *> &NewBlocks,
                            LoopBlocksDFS &LoopBlocks, ValueToValueMapTy &VMap,
                            LoopInfo *LI) {
  BasicBlock *Preheader = L->getLoopPreheader();
  BasicBlock *Header = L->getHeader();
  BasicBlock *Latch = L->getLoopLatch();
  Function *F = Header->getParent();
  LoopBlocksDFS::RPOIterator BlockBegin = LoopBlocks.beginRPO();
  LoopBlocksDFS::RPOIterator BlockEnd = LoopBlocks.endRPO();
  Loop *NewLoop = nullptr;
  Loop *ParentLoop = L->getParentLoop();
  if (!UnrollProlog) {
    NewLoop = new Loop();
    if (ParentLoop)
      ParentLoop->addChildLoop(NewLoop);
    else
      LI->addTopLevelLoop(NewLoop);
  }

  // For each block in the original loop, create a new copy,
  // and update the value map with the newly created values.
  for (LoopBlocksDFS::RPOIterator BB = BlockBegin; BB != BlockEnd; ++BB) {
    BasicBlock *NewBB = CloneBasicBlock(*BB, VMap, ".prol", F);
    NewBlocks.push_back(NewBB);

    if (NewLoop)
      NewLoop->addBasicBlockToLoop(NewBB, *LI);
    else if (ParentLoop)
      ParentLoop->addBasicBlockToLoop(NewBB, *LI);

    VMap[*BB] = NewBB;
    if (Header == *BB) {
      // For the first block, add a CFG connection to this newly
      // created block.
      InsertTop->getTerminator()->setSuccessor(0, NewBB);

    }
    if (Latch == *BB) {
      // For the last block, if UnrollProlog is true, create a direct jump to
      // InsertBot. If not, create a loop back to cloned head.
      VMap.erase((*BB)->getTerminator());
      BasicBlock *FirstLoopBB = cast<BasicBlock>(VMap[Header]);
      BranchInst *LatchBR = cast<BranchInst>(NewBB->getTerminator());
      IRBuilder<> Builder(LatchBR);
      if (UnrollProlog) {
        Builder.CreateBr(InsertBot);
      } else {
        PHINode *NewIdx = PHINode::Create(NewIter->getType(), 2, "prol.iter",
                                          FirstLoopBB->getFirstNonPHI());
        Value *IdxSub =
            Builder.CreateSub(NewIdx, ConstantInt::get(NewIdx->getType(), 1),
                              NewIdx->getName() + ".sub");
        Value *IdxCmp =
            Builder.CreateIsNotNull(IdxSub, NewIdx->getName() + ".cmp");
        Builder.CreateCondBr(IdxCmp, FirstLoopBB, InsertBot);
        NewIdx->addIncoming(NewIter, InsertTop);
        NewIdx->addIncoming(IdxSub, NewBB);
      }
      LatchBR->eraseFromParent();
    }
  }

  // Change the incoming values to the ones defined in the preheader or
  // cloned loop.
  for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
    PHINode *NewPHI = cast<PHINode>(VMap[&*I]);
    if (UnrollProlog) {
      VMap[&*I] = NewPHI->getIncomingValueForBlock(Preheader);
      cast<BasicBlock>(VMap[Header])->getInstList().erase(NewPHI);
    } else {
      unsigned idx = NewPHI->getBasicBlockIndex(Preheader);
      NewPHI->setIncomingBlock(idx, InsertTop);
      BasicBlock *NewLatch = cast<BasicBlock>(VMap[Latch]);
      idx = NewPHI->getBasicBlockIndex(Latch);
      Value *InVal = NewPHI->getIncomingValue(idx);
      NewPHI->setIncomingBlock(idx, NewLatch);
      if (VMap[InVal])
        NewPHI->setIncomingValue(idx, VMap[InVal]);
    }
  }
  if (NewLoop) {
    // Add unroll disable metadata to disable future unrolling for this loop.
    SmallVector<Metadata *, 4> MDs;
    // Reserve first location for self reference to the LoopID metadata node.
    MDs.push_back(nullptr);
    MDNode *LoopID = NewLoop->getLoopID();
    if (LoopID) {
      // First remove any existing loop unrolling metadata.
      for (unsigned i = 1, ie = LoopID->getNumOperands(); i < ie; ++i) {
        bool IsUnrollMetadata = false;
        MDNode *MD = dyn_cast<MDNode>(LoopID->getOperand(i));
        if (MD) {
          const MDString *S = dyn_cast<MDString>(MD->getOperand(0));
//.........这里部分代码省略.........

/// UnswitchNontrivialCondition - We determined that the loop is profitable 
/// to unswitch when LIC equal Val.  Split it into loop versions and test the 
/// condition outside of either loop.  Return the loops created as Out1/Out2.
void LoopUnswitch::UnswitchNontrivialCondition(Value *LIC, Constant *Val, 
                                               Loop *L) {
  Function *F = loopHeader->getParent();
  DEBUG(dbgs() << "loop-unswitch: Unswitching loop %"
        << loopHeader->getName() << " [" << L->getBlocks().size()
        << " blocks] in Function " << F->getName()
        << " when '" << *Val << "' == " << *LIC << "\n");

  if (ScalarEvolution *SE = getAnalysisIfAvailable<ScalarEvolution>())
    SE->forgetLoop(L);

  LoopBlocks.clear();
  NewBlocks.clear();

  // First step, split the preheader and exit blocks, and add these blocks to
  // the LoopBlocks list.
  BasicBlock *NewPreheader = SplitEdge(loopPreheader, loopHeader, this);
  LoopBlocks.push_back(NewPreheader);

  // We want the loop to come after the preheader, but before the exit blocks.
  LoopBlocks.insert(LoopBlocks.end(), L->block_begin(), L->block_end());

  SmallVector<BasicBlock*, 8> ExitBlocks;
  L->getUniqueExitBlocks(ExitBlocks);

  // Split all of the edges from inside the loop to their exit blocks.  Update
  // the appropriate Phi nodes as we do so.
  SplitExitEdges(L, ExitBlocks);

  // The exit blocks may have been changed due to edge splitting, recompute.
  ExitBlocks.clear();
  L->getUniqueExitBlocks(ExitBlocks);

  // Add exit blocks to the loop blocks.
  LoopBlocks.insert(LoopBlocks.end(), ExitBlocks.begin(), ExitBlocks.end());

  // Next step, clone all of the basic blocks that make up the loop (including
  // the loop preheader and exit blocks), keeping track of the mapping between
  // the instructions and blocks.
  NewBlocks.reserve(LoopBlocks.size());
  ValueToValueMapTy VMap;
  for (unsigned i = 0, e = LoopBlocks.size(); i != e; ++i) {
    BasicBlock *NewBB = CloneBasicBlock(LoopBlocks[i], VMap, ".us", F);
    NewBlocks.push_back(NewBB);
    VMap[LoopBlocks[i]] = NewBB;  // Keep the BB mapping.
    LPM->cloneBasicBlockSimpleAnalysis(LoopBlocks[i], NewBB, L);
  }

  // Splice the newly inserted blocks into the function right before the
  // original preheader.
  F->getBasicBlockList().splice(NewPreheader, F->getBasicBlockList(),
                                NewBlocks[0], F->end());

  // Now we create the new Loop object for the versioned loop.
  Loop *NewLoop = CloneLoop(L, L->getParentLoop(), VMap, LI, LPM);
  Loop *ParentLoop = L->getParentLoop();
  if (ParentLoop) {
    // Make sure to add the cloned preheader and exit blocks to the parent loop
    // as well.
    ParentLoop->addBasicBlockToLoop(NewBlocks[0], LI->getBase());
  }
  
  for (unsigned i = 0, e = ExitBlocks.size(); i != e; ++i) {
    BasicBlock *NewExit = cast<BasicBlock>(VMap[ExitBlocks[i]]);
    // The new exit block should be in the same loop as the old one.
    if (Loop *ExitBBLoop = LI->getLoopFor(ExitBlocks[i]))
      ExitBBLoop->addBasicBlockToLoop(NewExit, LI->getBase());
    
    assert(NewExit->getTerminator()->getNumSuccessors() == 1 &&
           "Exit block should have been split to have one successor!");
    BasicBlock *ExitSucc = NewExit->getTerminator()->getSuccessor(0);

    // If the successor of the exit block had PHI nodes, add an entry for
    // NewExit.
    PHINode *PN;
    for (BasicBlock::iterator I = ExitSucc->begin(); isa<PHINode>(I); ++I) {
      PN = cast<PHINode>(I);
      Value *V = PN->getIncomingValueForBlock(ExitBlocks[i]);
      ValueToValueMapTy::iterator It = VMap.find(V);
      if (It != VMap.end()) V = It->second;
      PN->addIncoming(V, NewExit);
    }
  }

  // Rewrite the code to refer to itself.
  for (unsigned i = 0, e = NewBlocks.size(); i != e; ++i)
    for (BasicBlock::iterator I = NewBlocks[i]->begin(),
           E = NewBlocks[i]->end(); I != E; ++I)
      RemapInstruction(I, VMap,RF_NoModuleLevelChanges|RF_IgnoreMissingEntries);
  
  // Rewrite the original preheader to select between versions of the loop.
  BranchInst *OldBR = cast<BranchInst>(loopPreheader->getTerminator());
  assert(OldBR->isUnconditional() && OldBR->getSuccessor(0) == LoopBlocks[0] &&
         "Preheader splitting did not work correctly!");

  // Emit the new branch that selects between the two versions of this loop.
  EmitPreheaderBranchOnCondition(LIC, Val, NewBlocks[0], LoopBlocks[0], OldBR);
//.........这里部分代码省略.........

// RewriteLoopBodyWithConditionConstant - We know either that the value LIC has
// the value specified by Val in the specified loop, or we know it does NOT have
// that value.  Rewrite any uses of LIC or of properties correlated to it.
void LoopUnswitch::RewriteLoopBodyWithConditionConstant(Loop *L, Value *LIC,
                                                        Constant *Val,
                                                        bool IsEqual) {
  assert(!isa<Constant>(LIC) && "Why are we unswitching on a constant?");
  
  // FIXME: Support correlated properties, like:
  //  for (...)
  //    if (li1 < li2)
  //      ...
  //    if (li1 > li2)
  //      ...
  
  // FOLD boolean conditions (X|LIC), (X&LIC).  Fold conditional branches,
  // selects, switches.
  std::vector<Instruction*> Worklist;
  LLVMContext &Context = Val->getContext();


  // If we know that LIC == Val, or that LIC == NotVal, just replace uses of LIC
  // in the loop with the appropriate one directly.
  if (IsEqual || (isa<ConstantInt>(Val) &&
      Val->getType()->isIntegerTy(1))) {
    Value *Replacement;
    if (IsEqual)
      Replacement = Val;
    else
      Replacement = ConstantInt::get(Type::getInt1Ty(Val->getContext()), 
                                     !cast<ConstantInt>(Val)->getZExtValue());
    
    for (Value::use_iterator UI = LIC->use_begin(), E = LIC->use_end();
         UI != E; ++UI) {
      Instruction *U = dyn_cast<Instruction>(*UI);
      if (!U || !L->contains(U))
        continue;
      U->replaceUsesOfWith(LIC, Replacement);
      Worklist.push_back(U);
    }
    SimplifyCode(Worklist, L);
    return;
  }
  
  // Otherwise, we don't know the precise value of LIC, but we do know that it
  // is certainly NOT "Val".  As such, simplify any uses in the loop that we
  // can.  This case occurs when we unswitch switch statements.
  for (Value::use_iterator UI = LIC->use_begin(), E = LIC->use_end();
       UI != E; ++UI) {
    Instruction *U = dyn_cast<Instruction>(*UI);
    if (!U || !L->contains(U))
      continue;

    Worklist.push_back(U);

    // TODO: We could do other simplifications, for example, turning 
    // 'icmp eq LIC, Val' -> false.

    // If we know that LIC is not Val, use this info to simplify code.
    SwitchInst *SI = dyn_cast<SwitchInst>(U);
    if (SI == 0 || !isa<ConstantInt>(Val)) continue;
    
    unsigned DeadCase = SI->findCaseValue(cast<ConstantInt>(Val));
    if (DeadCase == 0) continue;  // Default case is live for multiple values.
    
    // Found a dead case value.  Don't remove PHI nodes in the 
    // successor if they become single-entry, those PHI nodes may
    // be in the Users list.

    BasicBlock *Switch = SI->getParent();
    BasicBlock *SISucc = SI->getSuccessor(DeadCase);
    BasicBlock *Latch = L->getLoopLatch();
    if (!SI->findCaseDest(SISucc)) continue;  // Edge is critical.
    // If the DeadCase successor dominates the loop latch, then the
    // transformation isn't safe since it will delete the sole predecessor edge
    // to the latch.
    if (Latch && DT->dominates(SISucc, Latch))
      continue;

    // FIXME: This is a hack.  We need to keep the successor around
    // and hooked up so as to preserve the loop structure, because
    // trying to update it is complicated.  So instead we preserve the
    // loop structure and put the block on a dead code path.
    SplitEdge(Switch, SISucc, this);
    // Compute the successors instead of relying on the return value
    // of SplitEdge, since it may have split the switch successor
    // after PHI nodes.
    BasicBlock *NewSISucc = SI->getSuccessor(DeadCase);
    BasicBlock *OldSISucc = *succ_begin(NewSISucc);
    // Create an "unreachable" destination.
    BasicBlock *Abort = BasicBlock::Create(Context, "us-unreachable",
                                           Switch->getParent(),
                                           OldSISucc);
    new UnreachableInst(Context, Abort);
    // Force the new case destination to branch to the "unreachable"
    // block while maintaining a (dead) CFG edge to the old block.
    NewSISucc->getTerminator()->eraseFromParent();
    BranchInst::Create(Abort, OldSISucc,
                       ConstantInt::getTrue(Context), NewSISucc);
    // Release the PHI operands for this edge.
//.........这里部分代码省略.........

//.........这里部分代码省略.........
        RenamePass(RPD.BB, RPD.Pred, RPD.Values, RenamePassWorkList);
    } while (!RenamePassWorkList.empty());

    // The renamer uses the Visited set to avoid infinite loops.  Clear it now.
    Visited.clear();

    // Remove the allocas themselves from the function.
    for (unsigned i = 0, e = Allocas.size(); i != e; ++i) {
        Instruction *A = Allocas[i];

        // If there are any uses of the alloca instructions left, they must be in
        // sections of dead code that were not processed on the dominance frontier.
        // Just delete the users now.
        //
        if (!A->use_empty())
            A->replaceAllUsesWith(UndefValue::get(A->getType()));
        if (AST) AST->deleteValue(A);
        A->eraseFromParent();
    }

    // Remove alloca's dbg.declare instrinsics from the function.
    for (unsigned i = 0, e = AllocaDbgDeclares.size(); i != e; ++i)
        if (DbgDeclareInst *DDI = AllocaDbgDeclares[i])
            DDI->eraseFromParent();

    // Loop over all of the PHI nodes and see if there are any that we can get
    // rid of because they merge all of the same incoming values.  This can
    // happen due to undef values coming into the PHI nodes.  This process is
    // iterative, because eliminating one PHI node can cause others to be removed.
    bool EliminatedAPHI = true;
    while (EliminatedAPHI) {
        EliminatedAPHI = false;

        for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I =
                    NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E;) {
            PHINode *PN = I->second;

            // If this PHI node merges one value and/or undefs, get the value.
            if (Value *V = PN->hasConstantValue(&DT)) {
                if (AST && PN->getType()->isPointerTy())
                    AST->deleteValue(PN);
                PN->replaceAllUsesWith(V);
                PN->eraseFromParent();
                NewPhiNodes.erase(I++);
                EliminatedAPHI = true;
                continue;
            }
            ++I;
        }
    }

    // At this point, the renamer has added entries to PHI nodes for all reachable
    // code.  Unfortunately, there may be unreachable blocks which the renamer
    // hasn't traversed.  If this is the case, the PHI nodes may not
    // have incoming values for all predecessors.  Loop over all PHI nodes we have
    // created, inserting undef values if they are missing any incoming values.
    //
    for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I =
                NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E; ++I) {
        // We want to do this once per basic block.  As such, only process a block
        // when we find the PHI that is the first entry in the block.
        PHINode *SomePHI = I->second;
        BasicBlock *BB = SomePHI->getParent();
        if (&BB->front() != SomePHI)
            continue;

/// ComputeLiveInBlocks - Determine which blocks the value is live in.  These
/// are blocks which lead to uses.  Knowing this allows us to avoid inserting
/// PHI nodes into blocks which don't lead to uses (thus, the inserted phi nodes
/// would be dead).
void PromoteMem2Reg::
ComputeLiveInBlocks(AllocaInst *AI, AllocaInfo &Info,
                    const SmallPtrSet<BasicBlock*, 32> &DefBlocks,
                    SmallPtrSet<BasicBlock*, 32> &LiveInBlocks) {

    // To determine liveness, we must iterate through the predecessors of blocks
    // where the def is live.  Blocks are added to the worklist if we need to
    // check their predecessors.  Start with all the using blocks.
    SmallVector<BasicBlock*, 64> LiveInBlockWorklist;
    LiveInBlockWorklist.insert(LiveInBlockWorklist.end(),
                               Info.UsingBlocks.begin(), Info.UsingBlocks.end());

    // If any of the using blocks is also a definition block, check to see if the
    // definition occurs before or after the use.  If it happens before the use,
    // the value isn't really live-in.
    for (unsigned i = 0, e = LiveInBlockWorklist.size(); i != e; ++i) {
        BasicBlock *BB = LiveInBlockWorklist[i];
        if (!DefBlocks.count(BB)) continue;

        // Okay, this is a block that both uses and defines the value.  If the first
        // reference to the alloca is a def (store), then we know it isn't live-in.
        for (BasicBlock::iterator I = BB->begin(); ; ++I) {
            if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
                if (SI->getOperand(1) != AI) continue;

                // We found a store to the alloca before a load.  The alloca is not
                // actually live-in here.
                LiveInBlockWorklist[i] = LiveInBlockWorklist.back();
                LiveInBlockWorklist.pop_back();
                --i, --e;
                break;
            }

            if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
                if (LI->getOperand(0) != AI) continue;

                // Okay, we found a load before a store to the alloca.  It is actually
                // live into this block.
                break;
            }
        }
    }

    // Now that we have a set of blocks where the phi is live-in, recursively add
    // their predecessors until we find the full region the value is live.
    while (!LiveInBlockWorklist.empty()) {
        BasicBlock *BB = LiveInBlockWorklist.pop_back_val();

        // The block really is live in here, insert it into the set.  If already in
        // the set, then it has already been processed.
        if (!LiveInBlocks.insert(BB))
            continue;

        // Since the value is live into BB, it is either defined in a predecessor or
        // live into it to.  Add the preds to the worklist unless they are a
        // defining block.
        for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
            BasicBlock *P = *PI;

            // The value is not live into a predecessor if it defines the value.
            if (DefBlocks.count(P))
                continue;

            // Otherwise it is, add to the worklist.
            LiveInBlockWorklist.push_back(P);
        }
    }
}

void SanitizerCoverageModule::InjectCoverageAtBlock(Function &F, BasicBlock &BB,
                                                    bool UseCalls) {
  // Don't insert coverage for unreachable blocks: we will never call
  // __sanitizer_cov() for them, so counting them in
  // NumberOfInstrumentedBlocks() might complicate calculation of code coverage
  // percentage. Also, unreachable instructions frequently have no debug
  // locations.
  if (isa<UnreachableInst>(BB.getTerminator()))
    return;
  BasicBlock::iterator IP = BB.getFirstInsertionPt();

  bool IsEntryBB = &BB == &F.getEntryBlock();
  DebugLoc EntryLoc;
  if (IsEntryBB) {
    if (auto SP = getDISubprogram(&F))
      EntryLoc = DebugLoc::get(SP->getScopeLine(), 0, SP);
    // Keep static allocas and llvm.localescape calls in the entry block.  Even
    // if we aren't splitting the block, it's nice for allocas to be before
    // calls.
    IP = PrepareToSplitEntryBlock(BB, IP);
  } else {
    EntryLoc = IP->getDebugLoc();
  }

  IRBuilder<> IRB(&*IP);
  IRB.SetCurrentDebugLocation(EntryLoc);
  Value *GuardP = IRB.CreateAdd(
      IRB.CreatePointerCast(GuardArray, IntptrTy),
      ConstantInt::get(IntptrTy, (1 + NumberOfInstrumentedBlocks()) * 4));
  Type *Int32PtrTy = PointerType::getUnqual(IRB.getInt32Ty());
  GuardP = IRB.CreateIntToPtr(GuardP, Int32PtrTy);
  if (Options.TraceBB) {
    IRB.CreateCall(IsEntryBB ? SanCovTraceEnter : SanCovTraceBB, GuardP);
  } else if (UseCalls) {
    IRB.CreateCall(SanCovWithCheckFunction, GuardP);
  } else {
    LoadInst *Load = IRB.CreateLoad(GuardP);
    Load->setAtomic(Monotonic);
    Load->setAlignment(4);
    SetNoSanitizeMetadata(Load);
    Value *Cmp = IRB.CreateICmpSGE(Constant::getNullValue(Load->getType()), Load);
    Instruction *Ins = SplitBlockAndInsertIfThen(
        Cmp, &*IP, false, MDBuilder(*C).createBranchWeights(1, 100000));
    IRB.SetInsertPoint(Ins);
    IRB.SetCurrentDebugLocation(EntryLoc);
    // __sanitizer_cov gets the PC of the instruction using GET_CALLER_PC.
    IRB.CreateCall(SanCovFunction, GuardP);
    IRB.CreateCall(EmptyAsm, {}); // Avoids callback merge.
  }

  if (Options.Use8bitCounters) {
    IRB.SetInsertPoint(&*IP);
    Value *P = IRB.CreateAdd(
        IRB.CreatePointerCast(EightBitCounterArray, IntptrTy),
        ConstantInt::get(IntptrTy, NumberOfInstrumentedBlocks() - 1));
    P = IRB.CreateIntToPtr(P, IRB.getInt8PtrTy());
    LoadInst *LI = IRB.CreateLoad(P);
    Value *Inc = IRB.CreateAdd(LI, ConstantInt::get(IRB.getInt8Ty(), 1));
    StoreInst *SI = IRB.CreateStore(Inc, P);
    SetNoSanitizeMetadata(LI);
    SetNoSanitizeMetadata(SI);
  }
}

/// InsertStackProtectors - Insert code into the prologue and epilogue of the
/// function.
///
///  - The prologue code loads and stores the stack guard onto the stack.
///  - The epilogue checks the value stored in the prologue against the original
///    value. It calls __stack_chk_fail if they differ.
bool StackProtector::InsertStackProtectors() {
  bool HasPrologue = false;
  bool SupportsSelectionDAGSP =
      EnableSelectionDAGSP && !TM->Options.EnableFastISel;
  AllocaInst *AI = nullptr;       // Place on stack that stores the stack guard.
  Value *StackGuardVar = nullptr; // The stack guard variable.

  for (Function::iterator I = F->begin(), E = F->end(); I != E;) {
    BasicBlock *BB = &*I++;
    ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator());
    if (!RI)
      continue;

    if (!HasPrologue) {
      HasPrologue = true;
      SupportsSelectionDAGSP &=
          CreatePrologue(F, M, RI, TLI, Trip, AI, StackGuardVar);
    }

    if (SupportsSelectionDAGSP) {
      // Since we have a potential tail call, insert the special stack check
      // intrinsic.
      Instruction *InsertionPt = nullptr;
      if (CallInst *CI = FindPotentialTailCall(BB, RI, TLI)) {
        InsertionPt = CI;
      } else {
        InsertionPt = RI;
        // At this point we know that BB has a return statement so it *DOES*
        // have a terminator.
        assert(InsertionPt != nullptr &&
               "BB must have a terminator instruction at this point.");
      }

      Function *Intrinsic =
          Intrinsic::getDeclaration(M, Intrinsic::stackprotectorcheck);
      CallInst::Create(Intrinsic, StackGuardVar, "", InsertionPt);
    } else {
      // If we do not support SelectionDAG based tail calls, generate IR level
      // tail calls.
      //
      // For each block with a return instruction, convert this:
      //
      //   return:
      //     ...
      //     ret ...
      //
      // into this:
      //
      //   return:
      //     ...
      //     %1 = load __stack_chk_guard
      //     %2 = load StackGuardSlot
      //     %3 = cmp i1 %1, %2
      //     br i1 %3, label %SP_return, label %CallStackCheckFailBlk
      //
      //   SP_return:
      //     ret ...
      //
      //   CallStackCheckFailBlk:
      //     call void @__stack_chk_fail()
      //     unreachable

      // Create the FailBB. We duplicate the BB every time since the MI tail
      // merge pass will merge together all of the various BB into one including
      // fail BB generated by the stack protector pseudo instruction.
      BasicBlock *FailBB = CreateFailBB();

      // Split the basic block before the return instruction.
      BasicBlock *NewBB = BB->splitBasicBlock(RI->getIterator(), "SP_return");

      // Update the dominator tree if we need to.
      if (DT && DT->isReachableFromEntry(BB)) {
        DT->addNewBlock(NewBB, BB);
        DT->addNewBlock(FailBB, BB);
      }

      // Remove default branch instruction to the new BB.
      BB->getTerminator()->eraseFromParent();

      // Move the newly created basic block to the point right after the old
      // basic block so that it's in the "fall through" position.
      NewBB->moveAfter(BB);

      // Generate the stack protector instructions in the old basic block.
      IRBuilder<> B(BB);
      LoadInst *LI1 = B.CreateLoad(StackGuardVar);
      LoadInst *LI2 = B.CreateLoad(AI);
      Value *Cmp = B.CreateICmpEQ(LI1, LI2);
      unsigned SuccessWeight =
          BranchProbabilityInfo::getBranchWeightStackProtector(true);
      unsigned FailureWeight =
          BranchProbabilityInfo::getBranchWeightStackProtector(false);
      MDNode *Weights = MDBuilder(F->getContext())
                            .createBranchWeights(SuccessWeight, FailureWeight);
//.........这里部分代码省略.........

/// Generates code to divide two unsigned scalar 32-bit integers. Returns the
/// quotient, rounded towards 0. Builder's insert point should be pointing where
/// the caller wants code generated, e.g. at the udiv instruction.
static Value *generateUnsignedDivisionCode(Value *Dividend, Value *Divisor,
                                           IRBuilder<> &Builder) {
  // The basic algorithm can be found in the compiler-rt project's
  // implementation of __udivsi3.c. Here, we do a lower-level IR based approach
  // that's been hand-tuned to lessen the amount of control flow involved.

  // Some helper values
  IntegerType *I32Ty = Builder.getInt32Ty();

  ConstantInt *Zero      = Builder.getInt32(0);
  ConstantInt *One       = Builder.getInt32(1);
  ConstantInt *ThirtyOne = Builder.getInt32(31);
  ConstantInt *NegOne    = ConstantInt::getSigned(I32Ty, -1);
  ConstantInt *True      = Builder.getTrue();

  BasicBlock *IBB = Builder.GetInsertBlock();
  Function *F = IBB->getParent();
  Function *CTLZi32 = Intrinsic::getDeclaration(F->getParent(), Intrinsic::ctlz,
                                                I32Ty);

  // Our CFG is going to look like:
  // +---------------------+
  // | special-cases       |
  // |   ...               |
  // +---------------------+
  //  |       |
  //  |   +----------+
  //  |   |  bb1     |
  //  |   |  ...     |
  //  |   +----------+
  //  |    |      |
  //  |    |  +------------+
  //  |    |  |  preheader |
  //  |    |  |  ...       |
  //  |    |  +------------+
  //  |    |      |
  //  |    |      |      +---+
  //  |    |      |      |   |
  //  |    |  +------------+ |
  //  |    |  |  do-while  | |
  //  |    |  |  ...       | |
  //  |    |  +------------+ |
  //  |    |      |      |   |
  //  |   +-----------+  +---+
  //  |   | loop-exit |
  //  |   |  ...      |
  //  |   +-----------+
  //  |     |
  // +-------+
  // | ...   |
  // | end   |
  // +-------+
  BasicBlock *SpecialCases = Builder.GetInsertBlock();
  SpecialCases->setName(Twine(SpecialCases->getName(), "_udiv-special-cases"));
  BasicBlock *End = SpecialCases->splitBasicBlock(Builder.GetInsertPoint(),
                                                  "udiv-end");
  BasicBlock *LoopExit  = BasicBlock::Create(Builder.getContext(),
                                             "udiv-loop-exit", F, End);
  BasicBlock *DoWhile   = BasicBlock::Create(Builder.getContext(),
                                             "udiv-do-while", F, End);
  BasicBlock *Preheader = BasicBlock::Create(Builder.getContext(),
                                             "udiv-preheader", F, End);
  BasicBlock *BB1       = BasicBlock::Create(Builder.getContext(),
                                             "udiv-bb1", F, End);

  // We'll be overwriting the terminator to insert our extra blocks
  SpecialCases->getTerminator()->eraseFromParent();

  // First off, check for special cases: dividend or divisor is zero, divisor
  // is greater than dividend, and divisor is 1.
  // ; special-cases:
  // ;   %ret0_1      = icmp eq i32 %divisor, 0
  // ;   %ret0_2      = icmp eq i32 %dividend, 0
  // ;   %ret0_3      = or i1 %ret0_1, %ret0_2
  // ;   %tmp0        = tail call i32 @llvm.ctlz.i32(i32 %divisor, i1 true)
  // ;   %tmp1        = tail call i32 @llvm.ctlz.i32(i32 %dividend, i1 true)
  // ;   %sr          = sub nsw i32 %tmp0, %tmp1
  // ;   %ret0_4      = icmp ugt i32 %sr, 31
  // ;   %ret0        = or i1 %ret0_3, %ret0_4
  // ;   %retDividend = icmp eq i32 %sr, 31
  // ;   %retVal      = select i1 %ret0, i32 0, i32 %dividend
  // ;   %earlyRet    = or i1 %ret0, %retDividend
  // ;   br i1 %earlyRet, label %end, label %bb1
  Builder.SetInsertPoint(SpecialCases);
  Value *Ret0_1      = Builder.CreateICmpEQ(Divisor, Zero);
  Value *Ret0_2      = Builder.CreateICmpEQ(Dividend, Zero);
  Value *Ret0_3      = Builder.CreateOr(Ret0_1, Ret0_2);
  Value *Tmp0        = Builder.CreateCall2(CTLZi32, Divisor, True);
  Value *Tmp1        = Builder.CreateCall2(CTLZi32, Dividend, True);
  Value *SR          = Builder.CreateSub(Tmp0, Tmp1);
  Value *Ret0_4      = Builder.CreateICmpUGT(SR, ThirtyOne);
  Value *Ret0        = Builder.CreateOr(Ret0_3, Ret0_4);
  Value *RetDividend = Builder.CreateICmpEQ(SR, ThirtyOne);
  Value *RetVal      = Builder.CreateSelect(Ret0, Zero, Dividend);
  Value *EarlyRet    = Builder.CreateOr(Ret0, RetDividend);
  Builder.CreateCondBr(EarlyRet, End, BB1);

//.........这里部分代码省略.........

bool AtomicExpandLoadLinked::expandAtomicCmpXchg(AtomicCmpXchgInst *CI) {
  AtomicOrdering SuccessOrder = CI->getSuccessOrdering();
  AtomicOrdering FailureOrder = CI->getFailureOrdering();
  Value *Addr = CI->getPointerOperand();
  BasicBlock *BB = CI->getParent();
  Function *F = BB->getParent();
  LLVMContext &Ctx = F->getContext();

  // Given: cmpxchg some_op iN* %addr, iN %desired, iN %new success_ord fail_ord
  //
  // The full expansion we produce is:
  //     [...]
  //     fence?
  // cmpxchg.start:
  //     %loaded = @load.linked(%addr)
  //     %should_store = icmp eq %loaded, %desired
  //     br i1 %should_store, label %cmpxchg.trystore,
  //                          label %cmpxchg.failure
  // cmpxchg.trystore:
  //     %stored = @store_conditional(%new, %addr)
  //     %success = icmp eq i32 %stored, 0
  //     br i1 %success, label %cmpxchg.success, label %loop/%cmpxchg.failure
  // cmpxchg.success:
  //     fence?
  //     br label %cmpxchg.end
  // cmpxchg.failure:
  //     fence?
  //     br label %cmpxchg.end
  // cmpxchg.end:
  //     %success = phi i1 [true, %cmpxchg.success], [false, %cmpxchg.failure]
  //     %restmp = insertvalue { iN, i1 } undef, iN %loaded, 0
  //     %res = insertvalue { iN, i1 } %restmp, i1 %success, 1
  //     [...]
  BasicBlock *ExitBB = BB->splitBasicBlock(CI, "cmpxchg.end");
  auto FailureBB = BasicBlock::Create(Ctx, "cmpxchg.failure", F, ExitBB);
  auto SuccessBB = BasicBlock::Create(Ctx, "cmpxchg.success", F, FailureBB);
  auto TryStoreBB = BasicBlock::Create(Ctx, "cmpxchg.trystore", F, SuccessBB);
  auto LoopBB = BasicBlock::Create(Ctx, "cmpxchg.start", F, TryStoreBB);

  // This grabs the DebugLoc from CI
  IRBuilder<> Builder(CI);

  // The split call above "helpfully" added a branch at the end of BB (to the
  // wrong place), but we might want a fence too. It's easiest to just remove
  // the branch entirely.
  std::prev(BB->end())->eraseFromParent();
  Builder.SetInsertPoint(BB);
  AtomicOrdering MemOpOrder = insertLeadingFence(Builder, SuccessOrder);
  Builder.CreateBr(LoopBB);

  // Start the main loop block now that we've taken care of the preliminaries.
  Builder.SetInsertPoint(LoopBB);
  Value *Loaded = TM->getSubtargetImpl()->getTargetLowering()->emitLoadLinked(
      Builder, Addr, MemOpOrder);
  Value *ShouldStore =
      Builder.CreateICmpEQ(Loaded, CI->getCompareOperand(), "should_store");

  // If the the cmpxchg doesn't actually need any ordering when it fails, we can
  // jump straight past that fence instruction (if it exists).
  Builder.CreateCondBr(ShouldStore, TryStoreBB, FailureBB);

  Builder.SetInsertPoint(TryStoreBB);
  Value *StoreSuccess =
      TM->getSubtargetImpl()->getTargetLowering()->emitStoreConditional(
          Builder, CI->getNewValOperand(), Addr, MemOpOrder);
  StoreSuccess = Builder.CreateICmpEQ(
      StoreSuccess, ConstantInt::get(Type::getInt32Ty(Ctx), 0), "success");
  Builder.CreateCondBr(StoreSuccess, SuccessBB,
                       CI->isWeak() ? FailureBB : LoopBB);

  // Make sure later instructions don't get reordered with a fence if necessary.
  Builder.SetInsertPoint(SuccessBB);
  insertTrailingFence(Builder, SuccessOrder);
  Builder.CreateBr(ExitBB);

  Builder.SetInsertPoint(FailureBB);
  insertTrailingFence(Builder, FailureOrder);
  Builder.CreateBr(ExitBB);

  // Finally, we have control-flow based knowledge of whether the cmpxchg
  // succeeded or not. We expose this to later passes by converting any
  // subsequent "icmp eq/ne %loaded, %oldval" into a use of an appropriate PHI.

  // Setup the builder so we can create any PHIs we need.
  Builder.SetInsertPoint(ExitBB, ExitBB->begin());
  PHINode *Success = Builder.CreatePHI(Type::getInt1Ty(Ctx), 2);
  Success->addIncoming(ConstantInt::getTrue(Ctx), SuccessBB);
  Success->addIncoming(ConstantInt::getFalse(Ctx), FailureBB);

  // Look for any users of the cmpxchg that are just comparing the loaded value
  // against the desired one, and replace them with the CFG-derived version.
  SmallVector<ExtractValueInst *, 2> PrunedInsts;
  for (auto User : CI->users()) {
    ExtractValueInst *EV = dyn_cast<ExtractValueInst>(User);
    if (!EV)
      continue;

    assert(EV->getNumIndices() == 1 && EV->getIndices()[0] <= 1 &&
           "weird extraction from { iN, i1 }");

//.........这里部分代码省略.........