C++ SmallVector::empty方法代码示例

bool Inliner::runOnSCC(CallGraphSCC &SCC) {
  CallGraph &CG = getAnalysis<CallGraphWrapperPass>().getCallGraph();
  AssumptionCacheTracker *ACT = &getAnalysis<AssumptionCacheTracker>();
  auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
  const TargetLibraryInfo *TLI = TLIP ? &TLIP->getTLI() : nullptr;
  AliasAnalysis *AA = &getAnalysis<AliasAnalysis>();

  SmallPtrSet<Function*, 8> SCCFunctions;
  DEBUG(dbgs() << "Inliner visiting SCC:");
  for (CallGraphSCC::iterator I = SCC.begin(), E = SCC.end(); I != E; ++I) {
    Function *F = (*I)->getFunction();
    if (F) SCCFunctions.insert(F);
    DEBUG(dbgs() << " " << (F ? F->getName() : "INDIRECTNODE"));
  }

  // Scan through and identify all call sites ahead of time so that we only
  // inline call sites in the original functions, not call sites that result
  // from inlining other functions.
  SmallVector<std::pair<CallSite, int>, 16> CallSites;
  
  // When inlining a callee produces new call sites, we want to keep track of
  // the fact that they were inlined from the callee.  This allows us to avoid
  // infinite inlining in some obscure cases.  To represent this, we use an
  // index into the InlineHistory vector.
  SmallVector<std::pair<Function*, int>, 8> InlineHistory;

  for (CallGraphSCC::iterator I = SCC.begin(), E = SCC.end(); I != E; ++I) {
    Function *F = (*I)->getFunction();
    if (!F) continue;
    
    for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
      for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
        CallSite CS(cast<Value>(I));
        // If this isn't a call, or it is a call to an intrinsic, it can
        // never be inlined.
        if (!CS || isa<IntrinsicInst>(I))
          continue;
        
        // If this is a direct call to an external function, we can never inline
        // it.  If it is an indirect call, inlining may resolve it to be a
        // direct call, so we keep it.
        if (CS.getCalledFunction() && CS.getCalledFunction()->isDeclaration())
          continue;
        
        CallSites.push_back(std::make_pair(CS, -1));
      }
  }

  DEBUG(dbgs() << ": " << CallSites.size() << " call sites.\n");

  // If there are no calls in this function, exit early.
  if (CallSites.empty())
    return false;
  
  // Now that we have all of the call sites, move the ones to functions in the
  // current SCC to the end of the list.
  unsigned FirstCallInSCC = CallSites.size();
  for (unsigned i = 0; i < FirstCallInSCC; ++i)
    if (Function *F = CallSites[i].first.getCalledFunction())
      if (SCCFunctions.count(F))
        std::swap(CallSites[i--], CallSites[--FirstCallInSCC]);

  
  InlinedArrayAllocasTy InlinedArrayAllocas;
  InlineFunctionInfo InlineInfo(&CG, AA, ACT);

  // Now that we have all of the call sites, loop over them and inline them if
  // it looks profitable to do so.
  bool Changed = false;
  bool LocalChange;
  do {
    LocalChange = false;
    // Iterate over the outer loop because inlining functions can cause indirect
    // calls to become direct calls.
    for (unsigned CSi = 0; CSi != CallSites.size(); ++CSi) {
      CallSite CS = CallSites[CSi].first;
      
      Function *Caller = CS.getCaller();
      Function *Callee = CS.getCalledFunction();

      // If this call site is dead and it is to a readonly function, we should
      // just delete the call instead of trying to inline it, regardless of
      // size.  This happens because IPSCCP propagates the result out of the
      // call and then we're left with the dead call.
      if (isInstructionTriviallyDead(CS.getInstruction(), TLI)) {
        DEBUG(dbgs() << "    -> Deleting dead call: "
                     << *CS.getInstruction() << "\n");
        // Update the call graph by deleting the edge from Callee to Caller.
        CG[Caller]->removeCallEdgeFor(CS);
        CS.getInstruction()->eraseFromParent();
        ++NumCallsDeleted;
      } else {
        // We can only inline direct calls to non-declarations.
        if (!Callee || Callee->isDeclaration()) continue;
      
        // If this call site was obtained by inlining another function, verify
        // that the include path for the function did not include the callee
        // itself.  If so, we'd be recursively inlining the same function,
        // which would provide the same callsites, which would cause us to
        // infinitely inline.
//.........这里部分代码省略.........

void WinEHNumbering::processCallSite(ArrayRef<ActionHandler *> Actions,
                                     ImmutableCallSite CS) {
  int FirstMismatch = 0;
  for (int E = std::min(HandlerStack.size(), Actions.size()); FirstMismatch < E;
       ++FirstMismatch) {
    if (HandlerStack[FirstMismatch]->getHandlerBlockOrFunc() !=
        Actions[FirstMismatch]->getHandlerBlockOrFunc())
      break;
    delete Actions[FirstMismatch];
  }

  bool EnteringScope = (int)Actions.size() > FirstMismatch;

  // Don't recurse while we are looping over the handler stack.  Instead, defer
  // the numbering of the catch handlers until we are done popping.
  SmallVector<CatchHandler *, 4> PoppedCatches;
  for (int I = HandlerStack.size() - 1; I >= FirstMismatch; --I) {
    if (auto *CH = dyn_cast<CatchHandler>(HandlerStack.back())) {
      PoppedCatches.push_back(CH);
    } else {
      // Delete cleanup handlers
      delete HandlerStack.back();
    }
    HandlerStack.pop_back();
  }

  // We need to create a new state number if we are exiting a try scope and we
  // will not push any more actions.
  int TryHigh = NextState - 1;
  if (!EnteringScope && !PoppedCatches.empty()) {
    createUnwindMapEntry(currentEHNumber(), nullptr);
    ++NextState;
  }

  int LastTryLowIdx = 0;
  for (int I = 0, E = PoppedCatches.size(); I != E; ++I) {
    CatchHandler *CH = PoppedCatches[I];
    if (I + 1 == E || CH->getEHState() != PoppedCatches[I + 1]->getEHState()) {
      int TryLow = CH->getEHState();
      auto Handlers =
          makeArrayRef(&PoppedCatches[LastTryLowIdx], I - LastTryLowIdx + 1);
      createTryBlockMapEntry(TryLow, TryHigh, Handlers);
      LastTryLowIdx = I + 1;
    }
  }

  for (CatchHandler *CH : PoppedCatches) {
    if (auto *F = dyn_cast<Function>(CH->getHandlerBlockOrFunc()))
      calculateStateNumbers(*F);
    delete CH;
  }

  bool LastActionWasCatch = false;
  for (size_t I = FirstMismatch; I != Actions.size(); ++I) {
    // We can reuse eh states when pushing two catches for the same invoke.
    bool CurrActionIsCatch = isa<CatchHandler>(Actions[I]);
    // FIXME: Reenable this optimization!
    if (CurrActionIsCatch && LastActionWasCatch && false) {
      Actions[I]->setEHState(currentEHNumber());
    } else {
      createUnwindMapEntry(currentEHNumber(), Actions[I]);
      Actions[I]->setEHState(NextState);
      NextState++;
      DEBUG(dbgs() << "Creating unwind map entry for: (");
      print_name(Actions[I]->getHandlerBlockOrFunc());
      DEBUG(dbgs() << ", " << currentEHNumber() << ")\n");
    }
    HandlerStack.push_back(Actions[I]);
    LastActionWasCatch = CurrActionIsCatch;
  }

  DEBUG(dbgs() << "In EHState " << currentEHNumber() << " for CallSite: ");
  print_name(CS ? CS.getCalledValue() : nullptr);
  DEBUG(dbgs() << '\n');
}

MachineBasicBlock *
MachineBasicBlock::SplitCriticalEdge(MachineBasicBlock *Succ, Pass *P) {
  // Splitting the critical edge to a landing pad block is non-trivial. Don't do
  // it in this generic function.
  if (Succ->isLandingPad())
    return NULL;

  MachineFunction *MF = getParent();
  DebugLoc dl;  // FIXME: this is nowhere

  // We may need to update this's terminator, but we can't do that if
  // AnalyzeBranch fails. If this uses a jump table, we won't touch it.
  const TargetInstrInfo *TII = MF->getTarget().getInstrInfo();
  MachineBasicBlock *TBB = 0, *FBB = 0;
  SmallVector<MachineOperand, 4> Cond;
  if (TII->AnalyzeBranch(*this, TBB, FBB, Cond))
    return NULL;

  // Avoid bugpoint weirdness: A block may end with a conditional branch but
  // jumps to the same MBB is either case. We have duplicate CFG edges in that
  // case that we can't handle. Since this never happens in properly optimized
  // code, just skip those edges.
  if (TBB && TBB == FBB) {
    DEBUG(dbgs() << "Won't split critical edge after degenerate BB#"
                 << getNumber() << '\n');
    return NULL;
  }

  MachineBasicBlock *NMBB = MF->CreateMachineBasicBlock();
  MF->insert(llvm::next(MachineFunction::iterator(this)), NMBB);
  DEBUG(dbgs() << "Splitting critical edge:"
        " BB#" << getNumber()
        << " -- BB#" << NMBB->getNumber()
        << " -- BB#" << Succ->getNumber() << '\n');

  // On some targets like Mips, branches may kill virtual registers. Make sure
  // that LiveVariables is properly updated after updateTerminator replaces the
  // terminators.
  LiveVariables *LV = P->getAnalysisIfAvailable<LiveVariables>();

  // Collect a list of virtual registers killed by the terminators.
  SmallVector<unsigned, 4> KilledRegs;
  if (LV)
    for (instr_iterator I = getFirstInstrTerminator(), E = instr_end();
         I != E; ++I) {
      MachineInstr *MI = I;
      for (MachineInstr::mop_iterator OI = MI->operands_begin(),
           OE = MI->operands_end(); OI != OE; ++OI) {
        if (!OI->isReg() || OI->getReg() == 0 ||
            !OI->isUse() || !OI->isKill() || OI->isUndef())
          continue;
        unsigned Reg = OI->getReg();
        if (TargetRegisterInfo::isPhysicalRegister(Reg) ||
            LV->getVarInfo(Reg).removeKill(MI)) {
          KilledRegs.push_back(Reg);
          DEBUG(dbgs() << "Removing terminator kill: " << *MI);
          OI->setIsKill(false);
        }
      }
    }

  ReplaceUsesOfBlockWith(Succ, NMBB);
  updateTerminator();

  // Insert unconditional "jump Succ" instruction in NMBB if necessary.
  NMBB->addSuccessor(Succ);
  if (!NMBB->isLayoutSuccessor(Succ)) {
    Cond.clear();
    MF->getTarget().getInstrInfo()->InsertBranch(*NMBB, Succ, NULL, Cond, dl);
  }

  // Fix PHI nodes in Succ so they refer to NMBB instead of this
  for (MachineBasicBlock::instr_iterator
         i = Succ->instr_begin(),e = Succ->instr_end();
       i != e && i->isPHI(); ++i)
    for (unsigned ni = 1, ne = i->getNumOperands(); ni != ne; ni += 2)
      if (i->getOperand(ni+1).getMBB() == this)
        i->getOperand(ni+1).setMBB(NMBB);

  // Inherit live-ins from the successor
  for (MachineBasicBlock::livein_iterator I = Succ->livein_begin(),
         E = Succ->livein_end(); I != E; ++I)
    NMBB->addLiveIn(*I);

  // Update LiveVariables.
  const TargetRegisterInfo *TRI = MF->getTarget().getRegisterInfo();
  if (LV) {
    // Restore kills of virtual registers that were killed by the terminators.
    while (!KilledRegs.empty()) {
      unsigned Reg = KilledRegs.pop_back_val();
      for (instr_iterator I = instr_end(), E = instr_begin(); I != E;) {
        if (!(--I)->addRegisterKilled(Reg, TRI, /* addIfNotFound= */ false))
          continue;
        if (TargetRegisterInfo::isVirtualRegister(Reg))
          LV->getVarInfo(Reg).Kills.push_back(I);
        DEBUG(dbgs() << "Restored terminator kill: " << *I);
        break;
      }
    }
    // Update relevant live-through information.
//.........这里部分代码省略.........

void cl::ParseCommandLineOptions(int argc, char **argv,
                                 const char *Overview, bool ReadResponseFiles) {
  // Process all registered options.
  SmallVector<Option*, 4> PositionalOpts;
  SmallVector<Option*, 4> SinkOpts;
  StringMap<Option*> Opts;
  GetOptionInfo(PositionalOpts, SinkOpts, Opts);

  assert((!Opts.empty() || !PositionalOpts.empty()) &&
         "No options specified!");

  // Expand response files.
  std::vector<char*> newArgv;
  if (ReadResponseFiles) {
    newArgv.push_back(strdup(argv[0]));
    ExpandResponseFiles(argc, argv, newArgv);
    argv = &newArgv[0];
    argc = static_cast<int>(newArgv.size());
  }

  // Copy the program name into ProgName, making sure not to overflow it.
  std::string ProgName = sys::Path(argv[0]).getLast();
  size_t Len = std::min(ProgName.size(), size_t(79));
  memcpy(ProgramName, ProgName.data(), Len);
  ProgramName[Len] = '\0';

  ProgramOverview = Overview;
  bool ErrorParsing = false;

  // Check out the positional arguments to collect information about them.
  unsigned NumPositionalRequired = 0;

  // Determine whether or not there are an unlimited number of positionals
  bool HasUnlimitedPositionals = false;

  Option *ConsumeAfterOpt = 0;
  if (!PositionalOpts.empty()) {
    if (PositionalOpts[0]->getNumOccurrencesFlag() == cl::ConsumeAfter) {
      assert(PositionalOpts.size() > 1 &&
             "Cannot specify cl::ConsumeAfter without a positional argument!");
      ConsumeAfterOpt = PositionalOpts[0];
    }

    // Calculate how many positional values are _required_.
    bool UnboundedFound = false;
    for (size_t i = ConsumeAfterOpt != 0, e = PositionalOpts.size();
         i != e; ++i) {
      Option *Opt = PositionalOpts[i];
      if (RequiresValue(Opt))
        ++NumPositionalRequired;
      else if (ConsumeAfterOpt) {
        // ConsumeAfter cannot be combined with "optional" positional options
        // unless there is only one positional argument...
        if (PositionalOpts.size() > 2)
          ErrorParsing |=
            Opt->error("error - this positional option will never be matched, "
                       "because it does not Require a value, and a "
                       "cl::ConsumeAfter option is active!");
      } else if (UnboundedFound && !Opt->ArgStr[0]) {
        // This option does not "require" a value...  Make sure this option is
        // not specified after an option that eats all extra arguments, or this
        // one will never get any!
        //
        ErrorParsing |= Opt->error("error - option can never match, because "
                                   "another positional argument will match an "
                                   "unbounded number of values, and this option"
                                   " does not require a value!");
      }
      UnboundedFound |= EatsUnboundedNumberOfValues(Opt);
    }
    HasUnlimitedPositionals = UnboundedFound || ConsumeAfterOpt;
  }

  // PositionalVals - A vector of "positional" arguments we accumulate into
  // the process at the end.
  //
  SmallVector<std::pair<StringRef,unsigned>, 4> PositionalVals;

  // If the program has named positional arguments, and the name has been run
  // across, keep track of which positional argument was named.  Otherwise put
  // the positional args into the PositionalVals list...
  Option *ActivePositionalArg = 0;

  // Loop over all of the arguments... processing them.
  bool DashDashFound = false;  // Have we read '--'?
  for (int i = 1; i < argc; ++i) {
    Option *Handler = 0;
    StringRef Value;
    StringRef ArgName = "";

    // If the option list changed, this means that some command line
    // option has just been registered or deregistered.  This can occur in
    // response to things like -load, etc.  If this happens, rescan the options.
    if (OptionListChanged) {
      PositionalOpts.clear();
      SinkOpts.clear();
      Opts.clear();
      GetOptionInfo(PositionalOpts, SinkOpts, Opts);
      OptionListChanged = false;
    }
//.........这里部分代码省略.........

bool
TailDuplicatePass::duplicateSimpleBB(MachineBasicBlock *TailBB,
                                     SmallVector<MachineBasicBlock*, 8> &TDBBs,
                                     const DenseSet<unsigned> &UsedByPhi,
                                     SmallVector<MachineInstr*, 16> &Copies) {
    SmallPtrSet<MachineBasicBlock*, 8> Succs(TailBB->succ_begin(),
            TailBB->succ_end());
    SmallVector<MachineBasicBlock*, 8> Preds(TailBB->pred_begin(),
            TailBB->pred_end());
    bool Changed = false;
    for (SmallSetVector<MachineBasicBlock *, 8>::iterator PI = Preds.begin(),
            PE = Preds.end(); PI != PE; ++PI) {
        MachineBasicBlock *PredBB = *PI;

        if (PredBB->getLandingPadSuccessor())
            continue;

        if (bothUsedInPHI(*PredBB, Succs))
            continue;

        MachineBasicBlock *PredTBB = NULL, *PredFBB = NULL;
        SmallVector<MachineOperand, 4> PredCond;
        if (TII->AnalyzeBranch(*PredBB, PredTBB, PredFBB, PredCond, true))
            continue;

        Changed = true;
        DEBUG(dbgs() << "\nTail-duplicating into PredBB: " << *PredBB
              << "From simple Succ: " << *TailBB);

        MachineBasicBlock *NewTarget = *TailBB->succ_begin();
        MachineBasicBlock *NextBB = llvm::next(MachineFunction::iterator(PredBB));

        // Make PredFBB explicit.
        if (PredCond.empty())
            PredFBB = PredTBB;

        // Make fall through explicit.
        if (!PredTBB)
            PredTBB = NextBB;
        if (!PredFBB)
            PredFBB = NextBB;

        // Redirect
        if (PredFBB == TailBB)
            PredFBB = NewTarget;
        if (PredTBB == TailBB)
            PredTBB = NewTarget;

        // Make the branch unconditional if possible
        if (PredTBB == PredFBB) {
            PredCond.clear();
            PredFBB = NULL;
        }

        // Avoid adding fall through branches.
        if (PredFBB == NextBB)
            PredFBB = NULL;
        if (PredTBB == NextBB && PredFBB == NULL)
            PredTBB = NULL;

        TII->RemoveBranch(*PredBB);

        if (PredTBB)
            TII->InsertBranch(*PredBB, PredTBB, PredFBB, PredCond, DebugLoc());

        PredBB->removeSuccessor(TailBB);
        unsigned NumSuccessors = PredBB->succ_size();
        assert(NumSuccessors <= 1);
        if (NumSuccessors == 0 || *PredBB->succ_begin() != NewTarget)
            PredBB->addSuccessor(NewTarget);

        TDBBs.push_back(PredBB);
    }
    return Changed;
}

// Test whether it's safe to move Def to just before Insert.
// TODO: Compute memory dependencies in a way that doesn't require always
// walking the block.
// TODO: Compute memory dependencies in a way that uses AliasAnalysis to be
// more precise.
static bool IsSafeToMove(const MachineInstr *Def, const MachineInstr *Insert,
                         AliasAnalysis &AA, const MachineRegisterInfo &MRI) {
  assert(Def->getParent() == Insert->getParent());

  // Check for register dependencies.
  SmallVector<unsigned, 4> MutableRegisters;
  for (const MachineOperand &MO : Def->operands()) {
    if (!MO.isReg() || MO.isUndef())
      continue;
    unsigned Reg = MO.getReg();

    // If the register is dead here and at Insert, ignore it.
    if (MO.isDead() && Insert->definesRegister(Reg) &&
        !Insert->readsRegister(Reg))
      continue;

    if (TargetRegisterInfo::isPhysicalRegister(Reg)) {
      // Ignore ARGUMENTS; it's just used to keep the ARGUMENT_* instructions
      // from moving down, and we've already checked for that.
      if (Reg == WebAssembly::ARGUMENTS)
        continue;
      // If the physical register is never modified, ignore it.
      if (!MRI.isPhysRegModified(Reg))
        continue;
      // Otherwise, it's a physical register with unknown liveness.
      return false;
    }

    // If one of the operands isn't in SSA form, it has different values at
    // different times, and we need to make sure we don't move our use across
    // a different def.
    if (!MO.isDef() && !MRI.hasOneDef(Reg))
      MutableRegisters.push_back(Reg);
  }

  bool Read = false, Write = false, Effects = false, StackPointer = false;
  Query(*Def, AA, Read, Write, Effects, StackPointer);

  // If the instruction does not access memory and has no side effects, it has
  // no additional dependencies.
  bool HasMutableRegisters = !MutableRegisters.empty();
  if (!Read && !Write && !Effects && !StackPointer && !HasMutableRegisters)
    return true;

  // Scan through the intervening instructions between Def and Insert.
  MachineBasicBlock::const_iterator D(Def), I(Insert);
  for (--I; I != D; --I) {
    bool InterveningRead = false;
    bool InterveningWrite = false;
    bool InterveningEffects = false;
    bool InterveningStackPointer = false;
    Query(*I, AA, InterveningRead, InterveningWrite, InterveningEffects,
          InterveningStackPointer);
    if (Effects && InterveningEffects)
      return false;
    if (Read && InterveningWrite)
      return false;
    if (Write && (InterveningRead || InterveningWrite))
      return false;
    if (StackPointer && InterveningStackPointer)
      return false;

    for (unsigned Reg : MutableRegisters)
      for (const MachineOperand &MO : I->operands())
        if (MO.isReg() && MO.isDef() && MO.getReg() == Reg)
          return false;
  }

  return true;
}

/// splitLiveRangesAcrossInvokes - Each value that is live across an unwind edge
/// we spill into a stack location, guaranteeing that there is nothing live
/// across the unwind edge.  This process also splits all critical edges
/// coming out of invoke's.
/// FIXME: Move this function to a common utility file (Local.cpp?) so
/// both SjLj and LowerInvoke can use it.
void SjLjEHPass::
splitLiveRangesAcrossInvokes(SmallVector<InvokeInst*,16> &Invokes) {
  // First step, split all critical edges from invoke instructions.
  for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
    InvokeInst *II = Invokes[i];
    SplitCriticalEdge(II, 0, this);
    SplitCriticalEdge(II, 1, this);
    assert(!isa<PHINode>(II->getNormalDest()) &&
           !isa<PHINode>(II->getUnwindDest()) &&
           "critical edge splitting left single entry phi nodes?");
  }

  Function *F = Invokes.back()->getParent()->getParent();

  // To avoid having to handle incoming arguments specially, we lower each arg
  // to a copy instruction in the entry block.  This ensures that the argument
  // value itself cannot be live across the entry block.
  BasicBlock::iterator AfterAllocaInsertPt = F->begin()->begin();
  while (isa<AllocaInst>(AfterAllocaInsertPt) &&
        isa<ConstantInt>(cast<AllocaInst>(AfterAllocaInsertPt)->getArraySize()))
    ++AfterAllocaInsertPt;
  for (Function::arg_iterator AI = F->arg_begin(), E = F->arg_end();
       AI != E; ++AI) {
    const Type *Ty = AI->getType();
    // Aggregate types can't be cast, but are legal argument types, so we have
    // to handle them differently. We use an extract/insert pair as a
    // lightweight method to achieve the same goal.
    if (isa<StructType>(Ty) || isa<ArrayType>(Ty) || isa<VectorType>(Ty)) {
      Instruction *EI = ExtractValueInst::Create(AI, 0, "",AfterAllocaInsertPt);
      Instruction *NI = InsertValueInst::Create(AI, EI, 0);
      NI->insertAfter(EI);
      AI->replaceAllUsesWith(NI);
      // Set the operand of the instructions back to the AllocaInst.
      EI->setOperand(0, AI);
      NI->setOperand(0, AI);
    } else {
      // This is always a no-op cast because we're casting AI to AI->getType()
      // so src and destination types are identical. BitCast is the only
      // possibility.
      CastInst *NC = new BitCastInst(
        AI, AI->getType(), AI->getName()+".tmp", AfterAllocaInsertPt);
      AI->replaceAllUsesWith(NC);
      // Set the operand of the cast instruction back to the AllocaInst.
      // Normally it's forbidden to replace a CastInst's operand because it
      // could cause the opcode to reflect an illegal conversion. However,
      // we're replacing it here with the same value it was constructed with.
      // We do this because the above replaceAllUsesWith() clobbered the
      // operand, but we want this one to remain.
      NC->setOperand(0, AI);
    }
  }

  // Finally, scan the code looking for instructions with bad live ranges.
  for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
    for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
      // Ignore obvious cases we don't have to handle.  In particular, most
      // instructions either have no uses or only have a single use inside the
      // current block.  Ignore them quickly.
      Instruction *Inst = II;
      if (Inst->use_empty()) continue;
      if (Inst->hasOneUse() &&
          cast<Instruction>(Inst->use_back())->getParent() == BB &&
          !isa<PHINode>(Inst->use_back())) continue;

      // If this is an alloca in the entry block, it's not a real register
      // value.
      if (AllocaInst *AI = dyn_cast<AllocaInst>(Inst))
        if (isa<ConstantInt>(AI->getArraySize()) && BB == F->begin())
          continue;

      // Avoid iterator invalidation by copying users to a temporary vector.
      SmallVector<Instruction*,16> Users;
      for (Value::use_iterator UI = Inst->use_begin(), E = Inst->use_end();
           UI != E; ++UI) {
        Instruction *User = cast<Instruction>(*UI);
        if (User->getParent() != BB || isa<PHINode>(User))
          Users.push_back(User);
      }

      // Find all of the blocks that this value is live in.
      std::set<BasicBlock*> LiveBBs;
      LiveBBs.insert(Inst->getParent());
      while (!Users.empty()) {
        Instruction *U = Users.back();
        Users.pop_back();

        if (!isa<PHINode>(U)) {
          MarkBlocksLiveIn(U->getParent(), LiveBBs);
        } else {
          // Uses for a PHI node occur in their predecessor block.
          PHINode *PN = cast<PHINode>(U);
          for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
            if (PN->getIncomingValue(i) == Inst)
              MarkBlocksLiveIn(PN->getIncomingBlock(i), LiveBBs);
//.........这里部分代码省略.........

void ScheduleDAGSDNodes::BuildSchedUnits() {
  // During scheduling, the NodeId field of SDNode is used to map SDNodes
  // to their associated SUnits by holding SUnits table indices. A value
  // of -1 means the SDNode does not yet have an associated SUnit.
  unsigned NumNodes = 0;
  for (SelectionDAG::allnodes_iterator NI = DAG->allnodes_begin(),
       E = DAG->allnodes_end(); NI != E; ++NI) {
    NI->setNodeId(-1);
    ++NumNodes;
  }

  // Reserve entries in the vector for each of the SUnits we are creating.  This
  // ensure that reallocation of the vector won't happen, so SUnit*'s won't get
  // invalidated.
  // FIXME: Multiply by 2 because we may clone nodes during scheduling.
  // This is a temporary workaround.
  SUnits.reserve(NumNodes * 2);

  // Add all nodes in depth first order.
  SmallVector<SDNode*, 64> Worklist;
  SmallPtrSet<SDNode*, 64> Visited;
  Worklist.push_back(DAG->getRoot().getNode());
  Visited.insert(DAG->getRoot().getNode());

  SmallVector<SUnit*, 8> CallSUnits;
  while (!Worklist.empty()) {
    SDNode *NI = Worklist.pop_back_val();

    // Add all operands to the worklist unless they've already been added.
    for (unsigned i = 0, e = NI->getNumOperands(); i != e; ++i)
      if (Visited.insert(NI->getOperand(i).getNode()).second)
        Worklist.push_back(NI->getOperand(i).getNode());

    if (isPassiveNode(NI))  // Leaf node, e.g. a TargetImmediate.
      continue;

    // If this node has already been processed, stop now.
    if (NI->getNodeId() != -1) continue;

    SUnit *NodeSUnit = newSUnit(NI);

    // See if anything is glued to this node, if so, add them to glued
    // nodes.  Nodes can have at most one glue input and one glue output.  Glue
    // is required to be the last operand and result of a node.

    // Scan up to find glued preds.
    SDNode *N = NI;
    while (N->getNumOperands() &&
           N->getOperand(N->getNumOperands()-1).getValueType() == MVT::Glue) {
      N = N->getOperand(N->getNumOperands()-1).getNode();
      assert(N->getNodeId() == -1 && "Node already inserted!");
      N->setNodeId(NodeSUnit->NodeNum);
      if (N->isMachineOpcode() && TII->get(N->getMachineOpcode()).isCall())
        NodeSUnit->isCall = true;
    }

    // Scan down to find any glued succs.
    N = NI;
    while (N->getValueType(N->getNumValues()-1) == MVT::Glue) {
      SDValue GlueVal(N, N->getNumValues()-1);

      // There are either zero or one users of the Glue result.
      bool HasGlueUse = false;
      for (SDNode::use_iterator UI = N->use_begin(), E = N->use_end();
           UI != E; ++UI)
        if (GlueVal.isOperandOf(*UI)) {
          HasGlueUse = true;
          assert(N->getNodeId() == -1 && "Node already inserted!");
          N->setNodeId(NodeSUnit->NodeNum);
          N = *UI;
          if (N->isMachineOpcode() && TII->get(N->getMachineOpcode()).isCall())
            NodeSUnit->isCall = true;
          break;
        }
      if (!HasGlueUse) break;
    }

    if (NodeSUnit->isCall)
      CallSUnits.push_back(NodeSUnit);

    // Schedule zero-latency TokenFactor below any nodes that may increase the
    // schedule height. Otherwise, ancestors of the TokenFactor may appear to
    // have false stalls.
    if (NI->getOpcode() == ISD::TokenFactor)
      NodeSUnit->isScheduleLow = true;

    // If there are glue operands involved, N is now the bottom-most node
    // of the sequence of nodes that are glued together.
    // Update the SUnit.
    NodeSUnit->setNode(N);
    assert(N->getNodeId() == -1 && "Node already inserted!");
    N->setNodeId(NodeSUnit->NodeNum);

    // Compute NumRegDefsLeft. This must be done before AddSchedEdges.
    InitNumRegDefsLeft(NodeSUnit);

    // Assign the Latency field of NodeSUnit using target-provided information.
    computeLatency(NodeSUnit);
  }

//.........这里部分代码省略.........

/// EmitSchedule - Emit the machine code in scheduled order. Return the new
/// InsertPos and MachineBasicBlock that contains this insertion
/// point. ScheduleDAGSDNodes holds a BB pointer for convenience, but this does
/// not necessarily refer to returned BB. The emitter may split blocks.
MachineBasicBlock *ScheduleDAGSDNodes::
EmitSchedule(MachineBasicBlock::iterator &InsertPos) {
  InstrEmitter Emitter(BB, InsertPos);
  DenseMap<SDValue, unsigned> VRBaseMap;
  DenseMap<SUnit*, unsigned> CopyVRBaseMap;
  SmallVector<std::pair<unsigned, MachineInstr*>, 32> Orders;
  SmallSet<unsigned, 8> Seen;
  bool HasDbg = DAG->hasDebugValues();

  // If this is the first BB, emit byval parameter dbg_value's.
  if (HasDbg && BB->getParent()->begin() == MachineFunction::iterator(BB)) {
    SDDbgInfo::DbgIterator PDI = DAG->ByvalParmDbgBegin();
    SDDbgInfo::DbgIterator PDE = DAG->ByvalParmDbgEnd();
    for (; PDI != PDE; ++PDI) {
      MachineInstr *DbgMI= Emitter.EmitDbgValue(*PDI, VRBaseMap);
      if (DbgMI)
        BB->insert(InsertPos, DbgMI);
    }
  }

  for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
    SUnit *SU = Sequence[i];
    if (!SU) {
      // Null SUnit* is a noop.
      TII->insertNoop(*Emitter.getBlock(), InsertPos);
      continue;
    }

    // For pre-regalloc scheduling, create instructions corresponding to the
    // SDNode and any glued SDNodes and append them to the block.
    if (!SU->getNode()) {
      // Emit a copy.
      EmitPhysRegCopy(SU, CopyVRBaseMap, InsertPos);
      continue;
    }

    SmallVector<SDNode *, 4> GluedNodes;
    for (SDNode *N = SU->getNode()->getGluedNode(); N; N = N->getGluedNode())
      GluedNodes.push_back(N);
    while (!GluedNodes.empty()) {
      SDNode *N = GluedNodes.back();
      Emitter.EmitNode(GluedNodes.back(), SU->OrigNode != SU, SU->isCloned,
                       VRBaseMap);
      // Remember the source order of the inserted instruction.
      if (HasDbg)
        ProcessSourceNode(N, DAG, Emitter, VRBaseMap, Orders, Seen);
      GluedNodes.pop_back();
    }
    Emitter.EmitNode(SU->getNode(), SU->OrigNode != SU, SU->isCloned,
                     VRBaseMap);
    // Remember the source order of the inserted instruction.
    if (HasDbg)
      ProcessSourceNode(SU->getNode(), DAG, Emitter, VRBaseMap, Orders,
                        Seen);
  }

  // Insert all the dbg_values which have not already been inserted in source
  // order sequence.
  if (HasDbg) {
    MachineBasicBlock::iterator BBBegin = BB->getFirstNonPHI();

    // Sort the source order instructions and use the order to insert debug
    // values.
    std::sort(Orders.begin(), Orders.end(), less_first());

    SDDbgInfo::DbgIterator DI = DAG->DbgBegin();
    SDDbgInfo::DbgIterator DE = DAG->DbgEnd();
    // Now emit the rest according to source order.
    unsigned LastOrder = 0;
    for (unsigned i = 0, e = Orders.size(); i != e && DI != DE; ++i) {
      unsigned Order = Orders[i].first;
      MachineInstr *MI = Orders[i].second;
      // Insert all SDDbgValue's whose order(s) are before "Order".
      if (!MI)
        continue;
      for (; DI != DE &&
             (*DI)->getOrder() >= LastOrder && (*DI)->getOrder() < Order; ++DI) {
        if ((*DI)->isInvalidated())
          continue;
        MachineInstr *DbgMI = Emitter.EmitDbgValue(*DI, VRBaseMap);
        if (DbgMI) {
          if (!LastOrder)
            // Insert to start of the BB (after PHIs).
            BB->insert(BBBegin, DbgMI);
          else {
            // Insert at the instruction, which may be in a different
            // block, if the block was split by a custom inserter.
            MachineBasicBlock::iterator Pos = MI;
            MI->getParent()->insert(Pos, DbgMI);
          }
        }
      }
      LastOrder = Order;
    }
    // Add trailing DbgValue's before the terminator. FIXME: May want to add
    // some of them before one or more conditional branches?
//.........这里部分代码省略.........

//.........这里部分代码省略.........
      break;
    case CCValAssign::AExt:
      Arg = DAG.getNode(ISD::ANY_EXTEND, DL, VA.getLocVT(), Arg);
      break;
    default:
      llvm_unreachable("Unknown loc info!");
    }

    // Use local copy if it is a byval arg.
    if (Flags.isByVal())
      Arg = ByValArgs[J++];

    // Arguments that can be passed on register must be kept at RegsToPass
    // vector
    if (VA.isRegLoc()) {
      RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg));
    } else {
      assert(VA.isMemLoc());

      if (StackPtr.getNode() == 0)
        StackPtr = DAG.getCopyFromReg(Chain, DL, Lanai::SP,
                                      getPointerTy(DAG.getDataLayout()));

      SDValue PtrOff =
          DAG.getNode(ISD::ADD, DL, getPointerTy(DAG.getDataLayout()), StackPtr,
                      DAG.getIntPtrConstant(VA.getLocMemOffset(), DL));

      MemOpChains.push_back(DAG.getStore(
          Chain, DL, Arg, PtrOff, MachinePointerInfo(), false, false, 0));
    }
  }

  // Transform all store nodes into one single node because all store nodes are
  // independent of each other.
  if (!MemOpChains.empty())
    Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other,
                        ArrayRef<SDValue>(&MemOpChains[0], MemOpChains.size()));

  SDValue InFlag;

  // Build a sequence of copy-to-reg nodes chained together with token chain and
  // flag operands which copy the outgoing args into registers.  The InFlag in
  // necessary since all emitted instructions must be stuck together.
  for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I) {
    Chain = DAG.getCopyToReg(Chain, DL, RegsToPass[I].first,
                             RegsToPass[I].second, InFlag);
    InFlag = Chain.getValue(1);
  }

  // If the callee is a GlobalAddress node (quite common, every direct call is)
  // turn it into a TargetGlobalAddress node so that legalize doesn't hack it.
  // Likewise ExternalSymbol -> TargetExternalSymbol.
  uint8_t OpFlag = LanaiII::MO_NO_FLAG;
  if (G) {
    Callee = DAG.getTargetGlobalAddress(
        G->getGlobal(), DL, getPointerTy(DAG.getDataLayout()), 0, OpFlag);
  } else if (ExternalSymbolSDNode *E = dyn_cast<ExternalSymbolSDNode>(Callee)) {
    Callee = DAG.getTargetExternalSymbol(
        E->getSymbol(), getPointerTy(DAG.getDataLayout()), OpFlag);
  }

  // Returns a chain & a flag for retval copy to use.
  SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
  SmallVector<SDValue, 8> Ops;
  Ops.push_back(Chain);
  Ops.push_back(Callee);

  // Add a register mask operand representing the call-preserved registers.
  // TODO: Should return-twice functions be handled?
  const uint32_t *Mask =
      TRI->getCallPreservedMask(DAG.getMachineFunction(), CallConv);
  assert(Mask && "Missing call preserved mask for calling convention");
  Ops.push_back(DAG.getRegisterMask(Mask));

  // Add argument registers to the end of the list so that they are
  // known live into the call.
  for (unsigned I = 0, E = RegsToPass.size(); I != E; ++I)
    Ops.push_back(DAG.getRegister(RegsToPass[I].first,
                                  RegsToPass[I].second.getValueType()));

  if (InFlag.getNode())
    Ops.push_back(InFlag);

  Chain = DAG.getNode(LanaiISD::CALL, DL, NodeTys,
                      ArrayRef<SDValue>(&Ops[0], Ops.size()));
  InFlag = Chain.getValue(1);

  // Create the CALLSEQ_END node.
  Chain = DAG.getCALLSEQ_END(
      Chain,
      DAG.getConstant(NumBytes, DL, getPointerTy(DAG.getDataLayout()), true),
      DAG.getConstant(0, DL, getPointerTy(DAG.getDataLayout()), true), InFlag,
      DL);
  InFlag = Chain.getValue(1);

  // Handle result values, copying them out of physregs into vregs that we
  // return.
  return LowerCallResult(Chain, InFlag, CallConv, IsVarArg, Ins, DL, DAG,
                         InVals);
}

static bool mergeConstants(Module &M) {
  // Find all the globals that are marked "used".  These cannot be merged.
  SmallPtrSet<const GlobalValue*, 8> UsedGlobals;
  FindUsedValues(M.getGlobalVariable("llvm.used"), UsedGlobals);
  FindUsedValues(M.getGlobalVariable("llvm.compiler.used"), UsedGlobals);

  // Map unique constants to globals.
  DenseMap<Constant *, GlobalVariable *> CMap;

  // Replacements - This vector contains a list of replacements to perform.
  SmallVector<std::pair<GlobalVariable*, GlobalVariable*>, 32> Replacements;

  bool MadeChange = false;

  // XXX EMSCRIPTEN: mark @__init_array_start as not to be touched
  const GlobalValue *InitArrayStart = M.getNamedGlobal("__init_array_start");

  // Iterate constant merging while we are still making progress.  Merging two
  // constants together may allow us to merge other constants together if the
  // second level constants have initializers which point to the globals that
  // were just merged.
  while (1) {

    // First: Find the canonical constants others will be merged with.
    for (Module::global_iterator GVI = M.global_begin(), E = M.global_end();
         GVI != E; ) {
      GlobalVariable *GV = &*GVI++;

      // XXX EMSCRIPTEN: mark @__init_array_start as not to be touched
      if (GV == InitArrayStart)
        continue;

      // If this GV is dead, remove it.
      GV->removeDeadConstantUsers();
      if (GV->use_empty() && GV->hasLocalLinkage()) {
        GV->eraseFromParent();
        continue;
      }

      // Only process constants with initializers in the default address space.
      if (!GV->isConstant() || !GV->hasDefinitiveInitializer() ||
          GV->getType()->getAddressSpace() != 0 || GV->hasSection() ||
          // Don't touch values marked with attribute(used).
          UsedGlobals.count(GV))
        continue;

      // This transformation is legal for weak ODR globals in the sense it
      // doesn't change semantics, but we really don't want to perform it
      // anyway; it's likely to pessimize code generation, and some tools
      // (like the Darwin linker in cases involving CFString) don't expect it.
      if (GV->isWeakForLinker())
        continue;

      Constant *Init = GV->getInitializer();

      // Check to see if the initializer is already known.
      GlobalVariable *&Slot = CMap[Init];

      // If this is the first constant we find or if the old one is local,
      // replace with the current one. If the current is externally visible
      // it cannot be replace, but can be the canonical constant we merge with.
      if (!Slot || IsBetterCanonical(*GV, *Slot))
        Slot = GV;
    }

    // Second: identify all globals that can be merged together, filling in
    // the Replacements vector.  We cannot do the replacement in this pass
    // because doing so may cause initializers of other globals to be rewritten,
    // invalidating the Constant* pointers in CMap.
    for (Module::global_iterator GVI = M.global_begin(), E = M.global_end();
         GVI != E; ) {
      GlobalVariable *GV = &*GVI++;

      // Only process constants with initializers in the default address space.
      if (!GV->isConstant() || !GV->hasDefinitiveInitializer() ||
          GV->getType()->getAddressSpace() != 0 || GV->hasSection() ||
          // Don't touch values marked with attribute(used).
          UsedGlobals.count(GV))
        continue;

      // We can only replace constant with local linkage.
      if (!GV->hasLocalLinkage())
        continue;

      Constant *Init = GV->getInitializer();

      // Check to see if the initializer is already known.
      GlobalVariable *Slot = CMap[Init];

      if (!Slot || Slot == GV)
        continue;

      if (!Slot->hasGlobalUnnamedAddr() && !GV->hasGlobalUnnamedAddr())
        continue;

      if (!GV->hasGlobalUnnamedAddr())
        Slot->setUnnamedAddr(GlobalValue::UnnamedAddr::None);

      // Make all uses of the duplicate constant use the canonical version.
      Replacements.push_back(std::make_pair(GV, Slot));
//.........这里部分代码省略.........

/// isSafeToPromoteArgument - As you might guess from the name of this method,
/// it checks to see if it is both safe and useful to promote the argument.
/// This method limits promotion of aggregates to only promote up to three
/// elements of the aggregate in order to avoid exploding the number of
/// arguments passed in.
bool ArgPromotion::isSafeToPromoteArgument(Argument *Arg, bool isByVal) const {
  typedef std::set<IndicesVector> GEPIndicesSet;

  // Quick exit for unused arguments
  if (Arg->use_empty())
    return true;

  // We can only promote this argument if all of the uses are loads, or are GEP
  // instructions (with constant indices) that are subsequently loaded.
  //
  // Promoting the argument causes it to be loaded in the caller
  // unconditionally. This is only safe if we can prove that either the load
  // would have happened in the callee anyway (ie, there is a load in the entry
  // block) or the pointer passed in at every call site is guaranteed to be
  // valid.
  // In the former case, invalid loads can happen, but would have happened
  // anyway, in the latter case, invalid loads won't happen. This prevents us
  // from introducing an invalid load that wouldn't have happened in the
  // original code.
  //
  // This set will contain all sets of indices that are loaded in the entry
  // block, and thus are safe to unconditionally load in the caller.
  GEPIndicesSet SafeToUnconditionallyLoad;

  // This set contains all the sets of indices that we are planning to promote.
  // This makes it possible to limit the number of arguments added.
  GEPIndicesSet ToPromote;

  // If the pointer is always valid, any load with first index 0 is valid.
  if (isByVal || AllCalleesPassInValidPointerForArgument(Arg))
    SafeToUnconditionallyLoad.insert(IndicesVector(1, 0));

  // First, iterate the entry block and mark loads of (geps of) arguments as
  // safe.
  BasicBlock *EntryBlock = Arg->getParent()->begin();
  // Declare this here so we can reuse it
  IndicesVector Indices;
  for (BasicBlock::iterator I = EntryBlock->begin(), E = EntryBlock->end();
       I != E; ++I)
    if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
      Value *V = LI->getPointerOperand();
      if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(V)) {
        V = GEP->getPointerOperand();
        if (V == Arg) {
          // This load actually loads (part of) Arg? Check the indices then.
          Indices.reserve(GEP->getNumIndices());
          for (User::op_iterator II = GEP->idx_begin(), IE = GEP->idx_end();
               II != IE; ++II)
            if (ConstantInt *CI = dyn_cast<ConstantInt>(*II))
              Indices.push_back(CI->getSExtValue());
            else
              // We found a non-constant GEP index for this argument? Bail out
              // right away, can't promote this argument at all.
              return false;

          // Indices checked out, mark them as safe
          MarkIndicesSafe(Indices, SafeToUnconditionallyLoad);
          Indices.clear();
        }
      } else if (V == Arg) {
        // Direct loads are equivalent to a GEP with a single 0 index.
        MarkIndicesSafe(IndicesVector(1, 0), SafeToUnconditionallyLoad);
      }
    }

  // Now, iterate all uses of the argument to see if there are any uses that are
  // not (GEP+)loads, or any (GEP+)loads that are not safe to promote.
  SmallVector<LoadInst*, 16> Loads;
  IndicesVector Operands;
  for (Value::use_iterator UI = Arg->use_begin(), E = Arg->use_end();
       UI != E; ++UI) {
    User *U = *UI;
    Operands.clear();
    if (LoadInst *LI = dyn_cast<LoadInst>(U)) {
      if (LI->isVolatile()) return false;  // Don't hack volatile loads
      Loads.push_back(LI);
      // Direct loads are equivalent to a GEP with a zero index and then a load.
      Operands.push_back(0);
    } else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(U)) {
      if (GEP->use_empty()) {
        // Dead GEP's cause trouble later.  Just remove them if we run into
        // them.
        getAnalysis<AliasAnalysis>().deleteValue(GEP);
        GEP->eraseFromParent();
        // TODO: This runs the above loop over and over again for dead GEPs
        // Couldn't we just do increment the UI iterator earlier and erase the
        // use?
        return isSafeToPromoteArgument(Arg, isByVal);
      }

      // Ensure that all of the indices are constants.
      for (User::op_iterator i = GEP->idx_begin(), e = GEP->idx_end();
        i != e; ++i)
        if (ConstantInt *C = dyn_cast<ConstantInt>(*i))
          Operands.push_back(C->getSExtValue());
//.........这里部分代码省略.........

/// PromoteArguments - This method checks the specified function to see if there
/// are any promotable arguments and if it is safe to promote the function (for
/// example, all callers are direct).  If safe to promote some arguments, it
/// calls the DoPromotion method.
///
CallGraphNode *ArgPromotion::PromoteArguments(CallGraphNode *CGN) {
  Function *F = CGN->getFunction();

  // Make sure that it is local to this module.
  if (!F || !F->hasLocalLinkage()) return 0;

  // First check: see if there are any pointer arguments!  If not, quick exit.
  SmallVector<std::pair<Argument*, unsigned>, 16> PointerArgs;
  unsigned ArgNo = 0;
  for (Function::arg_iterator I = F->arg_begin(), E = F->arg_end();
       I != E; ++I, ++ArgNo)
    if (I->getType()->isPointerTy())
      PointerArgs.push_back(std::pair<Argument*, unsigned>(I, ArgNo));
  if (PointerArgs.empty()) return 0;

  // Second check: make sure that all callers are direct callers.  We can't
  // transform functions that have indirect callers.
  if (F->hasAddressTaken())
    return 0;

  // Check to see which arguments are promotable.  If an argument is promotable,
  // add it to ArgsToPromote.
  SmallPtrSet<Argument*, 8> ArgsToPromote;
  SmallPtrSet<Argument*, 8> ByValArgsToTransform;
  for (unsigned i = 0; i != PointerArgs.size(); ++i) {
    bool isByVal = F->paramHasAttr(PointerArgs[i].second+1, Attribute::ByVal);

    // If this is a byval argument, and if the aggregate type is small, just
    // pass the elements, which is always safe.
    Argument *PtrArg = PointerArgs[i].first;
    if (isByVal) {
      const Type *AgTy = cast<PointerType>(PtrArg->getType())->getElementType();
      if (const StructType *STy = dyn_cast<StructType>(AgTy)) {
        if (maxElements > 0 && STy->getNumElements() > maxElements) {
          DEBUG(dbgs() << "argpromotion disable promoting argument '"
                << PtrArg->getName() << "' because it would require adding more"
                << " than " << maxElements << " arguments to the function.\n");
        } else {
          // If all the elements are single-value types, we can promote it.
          bool AllSimple = true;
          for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i)
            if (!STy->getElementType(i)->isSingleValueType()) {
              AllSimple = false;
              break;
            }

          // Safe to transform, don't even bother trying to "promote" it.
          // Passing the elements as a scalar will allow scalarrepl to hack on
          // the new alloca we introduce.
          if (AllSimple) {
            ByValArgsToTransform.insert(PtrArg);
            continue;
          }
        }
      }
    }

    // Otherwise, see if we can promote the pointer to its value.
    if (isSafeToPromoteArgument(PtrArg, isByVal))
      ArgsToPromote.insert(PtrArg);
  }

  // No promotable pointer arguments.
  if (ArgsToPromote.empty() && ByValArgsToTransform.empty()) 
    return 0;

  return DoPromotion(F, ArgsToPromote, ByValArgsToTransform);
}

/// Remove dead functions that are not included in DNR (Do Not Remove) list.
bool Inliner::removeDeadFunctions(CallGraph &CG, bool AlwaysInlineOnly) {
  SmallVector<CallGraphNode*, 16> FunctionsToRemove;

  // Scan for all of the functions, looking for ones that should now be removed
  // from the program.  Insert the dead ones in the FunctionsToRemove set.
  for (CallGraph::iterator I = CG.begin(), E = CG.end(); I != E; ++I) {
    CallGraphNode *CGN = I->second;
    Function *F = CGN->getFunction();
    if (!F || F->isDeclaration())
      continue;

    // Handle the case when this function is called and we only want to care
    // about always-inline functions. This is a bit of a hack to share code
    // between here and the InlineAlways pass.
    if (AlwaysInlineOnly && !F->hasFnAttribute(Attribute::AlwaysInline))
      continue;

    // If the only remaining users of the function are dead constants, remove
    // them.
    F->removeDeadConstantUsers();

    if (!F->isDefTriviallyDead())
      continue;

    // It is unsafe to drop a function with discardable linkage from a COMDAT
    // without also dropping the other members of the COMDAT.
    // The inliner doesn't visit non-function entities which are in COMDAT
    // groups so it is unsafe to do so *unless* the linkage is local.
    if (!F->hasLocalLinkage() && F->hasComdat())
      continue;
    
    // Remove any call graph edges from the function to its callees.
    CGN->removeAllCalledFunctions();

    // Remove any edges from the external node to the function's call graph
    // node.  These edges might have been made irrelegant due to
    // optimization of the program.
    CG.getExternalCallingNode()->removeAnyCallEdgeTo(CGN);

    // Removing the node for callee from the call graph and delete it.
    FunctionsToRemove.push_back(CGN);
  }
  if (FunctionsToRemove.empty())
    return false;

  // Now that we know which functions to delete, do so.  We didn't want to do
  // this inline, because that would invalidate our CallGraph::iterator
  // objects. :(
  //
  // Note that it doesn't matter that we are iterating over a non-stable order
  // here to do this, it doesn't matter which order the functions are deleted
  // in.
  array_pod_sort(FunctionsToRemove.begin(), FunctionsToRemove.end());
  FunctionsToRemove.erase(std::unique(FunctionsToRemove.begin(),
                                      FunctionsToRemove.end()),
                          FunctionsToRemove.end());
  for (SmallVectorImpl<CallGraphNode *>::iterator I = FunctionsToRemove.begin(),
                                                  E = FunctionsToRemove.end();
       I != E; ++I) {
    delete CG.removeFunctionFromModule(*I);
    ++NumDeleted;
  }
  return true;
}

/// Returns Attribute::None, Attribute::ReadOnly or Attribute::ReadNone.
static Attribute::AttrKind
determinePointerReadAttrs(Argument *A,
                          const SmallPtrSet<Argument *, 8> &SCCNodes) {

  SmallVector<Use *, 32> Worklist;
  SmallSet<Use *, 32> Visited;

  // inalloca arguments are always clobbered by the call.
  if (A->hasInAllocaAttr())
    return Attribute::None;

  bool IsRead = false;
  // We don't need to track IsWritten. If A is written to, return immediately.

  for (Use &U : A->uses()) {
    Visited.insert(&U);
    Worklist.push_back(&U);
  }

  while (!Worklist.empty()) {
    Use *U = Worklist.pop_back_val();
    Instruction *I = cast<Instruction>(U->getUser());

    switch (I->getOpcode()) {
    case Instruction::BitCast:
    case Instruction::GetElementPtr:
    case Instruction::PHI:
    case Instruction::Select:
    case Instruction::AddrSpaceCast:
      // The original value is not read/written via this if the new value isn't.
      for (Use &UU : I->uses())
        if (Visited.insert(&UU).second)
          Worklist.push_back(&UU);
      break;

    case Instruction::Call:
    case Instruction::Invoke: {
      bool Captures = true;

      if (I->getType()->isVoidTy())
        Captures = false;

      auto AddUsersToWorklistIfCapturing = [&] {
        if (Captures)
          for (Use &UU : I->uses())
            if (Visited.insert(&UU).second)
              Worklist.push_back(&UU);
      };

      CallSite CS(I);
      if (CS.doesNotAccessMemory()) {
        AddUsersToWorklistIfCapturing();
        continue;
      }

      Function *F = CS.getCalledFunction();
      if (!F) {
        if (CS.onlyReadsMemory()) {
          IsRead = true;
          AddUsersToWorklistIfCapturing();
          continue;
        }
        return Attribute::None;
      }

      // Note: the callee and the two successor blocks *follow* the argument
      // operands.  This means there is no need to adjust UseIndex to account
      // for these.

      unsigned UseIndex = std::distance(CS.arg_begin(), U);

      // U cannot be the callee operand use: since we're exploring the
      // transitive uses of an Argument, having such a use be a callee would
      // imply the CallSite is an indirect call or invoke; and we'd take the
      // early exit above.
      assert(UseIndex < CS.data_operands_size() &&
             "Data operand use expected!");

      bool IsOperandBundleUse = UseIndex >= CS.getNumArgOperands();

      if (UseIndex >= F->arg_size() && !IsOperandBundleUse) {
        assert(F->isVarArg() && "More params than args in non-varargs call");
        return Attribute::None;
      }

      Captures &= !CS.doesNotCapture(UseIndex);

      // Since the optimizer (by design) cannot see the data flow corresponding
      // to a operand bundle use, these cannot participate in the optimistic SCC
      // analysis.  Instead, we model the operand bundle uses as arguments in
      // call to a function external to the SCC.
      if (!SCCNodes.count(&*std::next(F->arg_begin(), UseIndex)) ||
          IsOperandBundleUse) {

        // The accessors used on CallSite here do the right thing for calls and
        // invokes with operand bundles.

        if (!CS.onlyReadsMemory() && !CS.onlyReadsMemory(UseIndex))
          return Attribute::None;
//.........这里部分代码省略.........