Example #1
0
/// Merge an autorelease with a retain into a fused call.
bool
ObjCARCContract::ContractAutorelease(Function &F, Instruction *Autorelease,
                                     InstructionClass Class,
                                     SmallPtrSet<Instruction *, 4>
                                       &DependingInstructions,
                                     SmallPtrSet<const BasicBlock *, 4>
                                       &Visited) {
  const Value *Arg = GetObjCArg(Autorelease);

  // Check that there are no instructions between the retain and the autorelease
  // (such as an autorelease_pop) which may change the count.
  CallInst *Retain = 0;
  if (Class == IC_AutoreleaseRV)
    FindDependencies(RetainAutoreleaseRVDep, Arg,
                     Autorelease->getParent(), Autorelease,
                     DependingInstructions, Visited, PA);
  else
    FindDependencies(RetainAutoreleaseDep, Arg,
                     Autorelease->getParent(), Autorelease,
                     DependingInstructions, Visited, PA);

  Visited.clear();
  if (DependingInstructions.size() != 1) {
    DependingInstructions.clear();
    return false;
  }

  Retain = dyn_cast_or_null<CallInst>(*DependingInstructions.begin());
  DependingInstructions.clear();

  if (!Retain ||
      GetBasicInstructionClass(Retain) != IC_Retain ||
      GetObjCArg(Retain) != Arg)
    return false;

  Changed = true;
  ++NumPeeps;

  DEBUG(dbgs() << "ObjCARCContract::ContractAutorelease: Fusing "
                  "retain/autorelease. Erasing: " << *Autorelease << "\n"
                  "                                      Old Retain: "
               << *Retain << "\n");

  if (Class == IC_AutoreleaseRV)
    Retain->setCalledFunction(getRetainAutoreleaseRVCallee(F.getParent()));
  else
    Retain->setCalledFunction(getRetainAutoreleaseCallee(F.getParent()));

  DEBUG(dbgs() << "                                      New Retain: "
               << *Retain << "\n");

  EraseInstruction(Autorelease);
  return true;
}
static bool expandIntrinsic(Module *M, Intrinsic::ID ID) {
  SmallVector<Type *, 3> Types;
  Types.push_back(Type::getInt8PtrTy(M->getContext()));
  if (ID != Intrinsic::memset)
    Types.push_back(Type::getInt8PtrTy(M->getContext()));
  unsigned LengthTypePos = Types.size();
  Types.push_back(Type::getInt64Ty(M->getContext()));

  std::string OldName = Intrinsic::getName(ID, Types);
  Function *OldIntrinsic = M->getFunction(OldName);
  if (!OldIntrinsic)
    return false;

  Types[LengthTypePos] = Type::getInt32Ty(M->getContext());
  Function *NewIntrinsic = Intrinsic::getDeclaration(M, ID, Types);

  for (Value::use_iterator CallIter = OldIntrinsic->use_begin(),
         E = OldIntrinsic->use_end(); CallIter != E; ) {
    CallInst *Call = dyn_cast<CallInst>(*CallIter++);
    if (!Call) {
      report_fatal_error("CanonicalizeMemIntrinsics: Taking the address of an "
                         "intrinsic is not allowed: " + OldName);
    }
    // This temporarily leaves Call non-well-typed.
    Call->setCalledFunction(NewIntrinsic);
    // Truncate the "len" argument.  No overflow check.
    IRBuilder<> Builder(Call);
    Value *Length = Builder.CreateTrunc(Call->getArgOperand(2),
                                        Type::getInt32Ty(M->getContext()),
                                        "mem_len_truncate");
    Call->setArgOperand(2, Length);
  }
  OldIntrinsic->eraseFromParent();
  return true;
}
/*
 * Replace called function of a given call site.
 */
void DeadStoreEliminationPass::replaceCallingInst(Instruction* caller,
    Function* fn) {
  if (isa<CallInst>(caller)) {
    CallInst *callInst = dyn_cast<CallInst>(caller);
    callInst->setCalledFunction(fn);
  } else if (isa<InvokeInst>(caller)) {
    InvokeInst *invokeInst = dyn_cast<InvokeInst>(caller);
    invokeInst->setCalledFunction(fn);
  }
}
Example #4
0
/// Merge an autorelease with a retain into a fused call.
bool ObjCARCContract::contractAutorelease(
    Function &F, Instruction *Autorelease, ARCInstKind Class,
    SmallPtrSetImpl<Instruction *> &DependingInstructions,
    SmallPtrSetImpl<const BasicBlock *> &Visited) {
  const Value *Arg = GetArgRCIdentityRoot(Autorelease);

  // Check that there are no instructions between the retain and the autorelease
  // (such as an autorelease_pop) which may change the count.
  CallInst *Retain = nullptr;
  if (Class == ARCInstKind::AutoreleaseRV)
    FindDependencies(RetainAutoreleaseRVDep, Arg,
                     Autorelease->getParent(), Autorelease,
                     DependingInstructions, Visited, PA);
  else
    FindDependencies(RetainAutoreleaseDep, Arg,
                     Autorelease->getParent(), Autorelease,
                     DependingInstructions, Visited, PA);

  Visited.clear();
  if (DependingInstructions.size() != 1) {
    DependingInstructions.clear();
    return false;
  }

  Retain = dyn_cast_or_null<CallInst>(*DependingInstructions.begin());
  DependingInstructions.clear();

  if (!Retain || GetBasicARCInstKind(Retain) != ARCInstKind::Retain ||
      GetArgRCIdentityRoot(Retain) != Arg)
    return false;

  Changed = true;
  ++NumPeeps;

  LLVM_DEBUG(dbgs() << "    Fusing retain/autorelease!\n"
                       "        Autorelease:"
                    << *Autorelease
                    << "\n"
                       "        Retain: "
                    << *Retain << "\n");

  Function *Decl = EP.get(Class == ARCInstKind::AutoreleaseRV
                              ? ARCRuntimeEntryPointKind::RetainAutoreleaseRV
                              : ARCRuntimeEntryPointKind::RetainAutorelease);
  Retain->setCalledFunction(Decl);

  LLVM_DEBUG(dbgs() << "        New RetainAutorelease: " << *Retain << "\n");

  EraseInstruction(Autorelease);
  return true;
}
Example #5
0
void insertCallToAccessFunctionSequential(Function *F, Function *cF) {
  CallInst *I;
  BasicBlock *b;

  Value::user_iterator i = F->user_begin(), e = F->user_end();
  while (i != e) {
    if (isa<CallInst>(*i)) {

      I = dyn_cast<CallInst>(*i);
      b = I->getParent();
      BasicBlock::iterator helper(I);
      CallInst *ci = dyn_cast<CallInst>(I->clone());
      ci->setCalledFunction(cF);
      b->getInstList().insertAfter(helper, ci);

      i++;
      I->replaceAllUsesWith(ci);

      insertCallToPAPI(I, ci);
    }
  }
}
Example #6
0
//
// Method: runOnModule()
//
// Description:
//  Entry point for this LLVM pass.  Search for functions which could be called
//  indirectly and create clones for them which are only called by direct
//  calls.
//
// Inputs:
//  M - A reference to the LLVM module to transform.
//
// Outputs:
//  M - The transformed LLVM module.
//
// Return value:
//  true  - The module was modified.
//  false - The module was not modified.
//
bool
IndClone::runOnModule(Module& M) {
  // Set of functions to clone
  std::vector<Function*> toClone;

  //
  // Check all of the functions in the module.  If the function could be called
  // by an indirect function call, add it to our worklist of functions to
  // clone.
  //
  for (Module::iterator I = M.begin(); I != M.end(); ++I) {
    // Flag whether the function should be cloned
    bool pleaseCloneTheFunction = false;

    //
    // Only clone functions which are defined and cannot be replaced by another
    // function by the linker.
    //
    if (!I->isDeclaration() && !I->mayBeOverridden()) {
      for (Value::use_iterator ui = I->use_begin(), ue = I->use_end();
          ui != ue; ++ui) {
        if (!isa<CallInst>(*ui) && !isa<InvokeInst>(*ui)) {
          if(!ui->use_empty())
          //
          // If this function is used for anything other than a direct function
          // call, then we want to clone it.
          //
          pleaseCloneTheFunction = true;
        } else {
          //
          // This is a call instruction, but hold up ranger!  We need to make
          // sure that the function isn't passed as an argument to *another*
          // function.  That would make the function usable in an indirect
          // function call.
          //
          for (unsigned index = 1; index < ui->getNumOperands(); ++index) {
            if (ui->getOperand(index)->stripPointerCasts() == I) {
              pleaseCloneTheFunction = true;
              break;
            }
          }
        }

        //
        // If we've discovered that the function could be used by an indirect
        // call site, schedule it for cloning.
        //
        if (pleaseCloneTheFunction) {
          toClone.push_back(I);
          break;
        }
      }
    }
  }

  //
  // Update the statistics on the number of functions we'll be cloning.
  // We only update the statistic if we want to clone one or more functions;
  // due to the magic of how statistics work, avoiding assignment prevents it
  // from needlessly showing up.
  //
  if (toClone.size())
    numCloned += toClone.size();

  //
  // Go through the worklist and clone each function.  After cloning a
  // function, change all direct calls to use the clone instead of using the
  // original function.
  //
  for (unsigned index = 0; index < toClone.size(); ++index) {
    //
    // Clone the function and give it a name indicating that it is a clone to
    // be used for direct function calls.
    //
    Function * Original = toClone[index];
    Function* DirectF = CloneFunction(Original);
    DirectF->setName(Original->getName() + "_DIRECT");

    //
    // Make the clone internal; external code can use the original function.
    //
    DirectF->setLinkage(GlobalValue::InternalLinkage);

    //
    // Link the cloned function into the set of functions belonging to the
    // module.
    //
    Original->getParent()->getFunctionList().push_back(DirectF);

    //
    // Find all uses of the function that use it as a direct call.  Change
    // them to use the clone.
    //
    for (Value::use_iterator ui = Original->use_begin(),
                             ue = Original->use_end();
        ui != ue; ) {
      CallInst *CI = dyn_cast<CallInst>(*ui);
      ui++;
      if (CI) {
        if (CI->getCalledFunction() == Original) {
          ++numReplaced;
          CI->setCalledFunction(DirectF);
        }
      }
    }
  }
  
  //
  // Assume that we've cloned at least one function.
  //
  return true;
}
Example #7
0
void AMDGPUPromoteAlloca::handleAlloca(AllocaInst &I) {
  // Array allocations are probably not worth handling, since an allocation of
  // the array type is the canonical form.
  if (!I.isStaticAlloca() || I.isArrayAllocation())
    return;

  IRBuilder<> Builder(&I);

  // First try to replace the alloca with a vector
  Type *AllocaTy = I.getAllocatedType();

  DEBUG(dbgs() << "Trying to promote " << I << '\n');

  if (tryPromoteAllocaToVector(&I))
    return;

  DEBUG(dbgs() << " alloca is not a candidate for vectorization.\n");

  const Function &ContainingFunction = *I.getParent()->getParent();

  // FIXME: We should also try to get this value from the reqd_work_group_size
  // function attribute if it is available.
  unsigned WorkGroupSize = AMDGPU::getMaximumWorkGroupSize(ContainingFunction);

  int AllocaSize =
      WorkGroupSize * Mod->getDataLayout().getTypeAllocSize(AllocaTy);

  if (AllocaSize > LocalMemAvailable) {
    DEBUG(dbgs() << " Not enough local memory to promote alloca.\n");
    return;
  }

  std::vector<Value*> WorkList;

  if (!collectUsesWithPtrTypes(&I, WorkList)) {
    DEBUG(dbgs() << " Do not know how to convert all uses\n");
    return;
  }

  DEBUG(dbgs() << "Promoting alloca to local memory\n");
  LocalMemAvailable -= AllocaSize;

  Function *F = I.getParent()->getParent();

  Type *GVTy = ArrayType::get(I.getAllocatedType(), WorkGroupSize);
  GlobalVariable *GV = new GlobalVariable(
      *Mod, GVTy, false, GlobalValue::InternalLinkage,
      UndefValue::get(GVTy),
      Twine(F->getName()) + Twine('.') + I.getName(),
      nullptr,
      GlobalVariable::NotThreadLocal,
      AMDGPUAS::LOCAL_ADDRESS);
  GV->setUnnamedAddr(true);
  GV->setAlignment(I.getAlignment());

  Value *TCntY, *TCntZ;

  std::tie(TCntY, TCntZ) = getLocalSizeYZ(Builder);
  Value *TIdX = getWorkitemID(Builder, 0);
  Value *TIdY = getWorkitemID(Builder, 1);
  Value *TIdZ = getWorkitemID(Builder, 2);

  Value *Tmp0 = Builder.CreateMul(TCntY, TCntZ, "", true, true);
  Tmp0 = Builder.CreateMul(Tmp0, TIdX);
  Value *Tmp1 = Builder.CreateMul(TIdY, TCntZ, "", true, true);
  Value *TID = Builder.CreateAdd(Tmp0, Tmp1);
  TID = Builder.CreateAdd(TID, TIdZ);

  Value *Indices[] = {
    Constant::getNullValue(Type::getInt32Ty(Mod->getContext())),
    TID
  };

  Value *Offset = Builder.CreateInBoundsGEP(GVTy, GV, Indices);
  I.mutateType(Offset->getType());
  I.replaceAllUsesWith(Offset);
  I.eraseFromParent();

  for (Value *V : WorkList) {
    CallInst *Call = dyn_cast<CallInst>(V);
    if (!Call) {
      Type *EltTy = V->getType()->getPointerElementType();
      PointerType *NewTy = PointerType::get(EltTy, AMDGPUAS::LOCAL_ADDRESS);

      // The operand's value should be corrected on its own.
      if (isa<AddrSpaceCastInst>(V))
        continue;

      // FIXME: It doesn't really make sense to try to do this for all
      // instructions.
      V->mutateType(NewTy);
      continue;
    }

    IntrinsicInst *Intr = dyn_cast<IntrinsicInst>(Call);
    if (!Intr) {
      // FIXME: What is this for? It doesn't make sense to promote arbitrary
      // function calls. If the call is to a defined function that can also be
      // promoted, we should be able to do this once that function is also
      // rewritten.

      std::vector<Type*> ArgTypes;
      for (unsigned ArgIdx = 0, ArgEnd = Call->getNumArgOperands();
                                ArgIdx != ArgEnd; ++ArgIdx) {
        ArgTypes.push_back(Call->getArgOperand(ArgIdx)->getType());
      }
      Function *F = Call->getCalledFunction();
      FunctionType *NewType = FunctionType::get(Call->getType(), ArgTypes,
                                                F->isVarArg());
      Constant *C = Mod->getOrInsertFunction((F->getName() + ".local").str(),
                                             NewType, F->getAttributes());
      Function *NewF = cast<Function>(C);
      Call->setCalledFunction(NewF);
      continue;
    }

    Builder.SetInsertPoint(Intr);
    switch (Intr->getIntrinsicID()) {
    case Intrinsic::lifetime_start:
    case Intrinsic::lifetime_end:
      // These intrinsics are for address space 0 only
      Intr->eraseFromParent();
      continue;
    case Intrinsic::memcpy: {
      MemCpyInst *MemCpy = cast<MemCpyInst>(Intr);
      Builder.CreateMemCpy(MemCpy->getRawDest(), MemCpy->getRawSource(),
                           MemCpy->getLength(), MemCpy->getAlignment(),
                           MemCpy->isVolatile());
      Intr->eraseFromParent();
      continue;
    }
    case Intrinsic::memmove: {
      MemMoveInst *MemMove = cast<MemMoveInst>(Intr);
      Builder.CreateMemMove(MemMove->getRawDest(), MemMove->getRawSource(),
                            MemMove->getLength(), MemMove->getAlignment(),
                            MemMove->isVolatile());
      Intr->eraseFromParent();
      continue;
    }
    case Intrinsic::memset: {
      MemSetInst *MemSet = cast<MemSetInst>(Intr);
      Builder.CreateMemSet(MemSet->getRawDest(), MemSet->getValue(),
                           MemSet->getLength(), MemSet->getAlignment(),
                           MemSet->isVolatile());
      Intr->eraseFromParent();
      continue;
    }
    case Intrinsic::invariant_start:
    case Intrinsic::invariant_end:
    case Intrinsic::invariant_group_barrier:
      Intr->eraseFromParent();
      // FIXME: I think the invariant marker should still theoretically apply,
      // but the intrinsics need to be changed to accept pointers with any
      // address space.
      continue;
    case Intrinsic::objectsize: {
      Value *Src = Intr->getOperand(0);
      Type *SrcTy = Src->getType()->getPointerElementType();
      Function *ObjectSize = Intrinsic::getDeclaration(Mod,
        Intrinsic::objectsize,
        { Intr->getType(), PointerType::get(SrcTy, AMDGPUAS::LOCAL_ADDRESS) }
      );

      CallInst *NewCall
        = Builder.CreateCall(ObjectSize, { Src, Intr->getOperand(1) });
      Intr->replaceAllUsesWith(NewCall);
      Intr->eraseFromParent();
      continue;
    }
    default:
      Intr->dump();
      llvm_unreachable("Don't know how to promote alloca intrinsic use.");
    }
  }
}
void AMDGPUPromoteAlloca::visitAlloca(AllocaInst &I) {
  IRBuilder<> Builder(&I);

  // First try to replace the alloca with a vector
  Type *AllocaTy = I.getAllocatedType();

  DEBUG(dbgs() << "Trying to promote " << I << '\n');

  if (tryPromoteAllocaToVector(&I))
    return;

  DEBUG(dbgs() << " alloca is not a candidate for vectorization.\n");

  // FIXME: This is the maximum work group size.  We should try to get
  // value from the reqd_work_group_size function attribute if it is
  // available.
  unsigned WorkGroupSize = 256;
  int AllocaSize = WorkGroupSize *
      Mod->getDataLayout()->getTypeAllocSize(AllocaTy);

  if (AllocaSize > LocalMemAvailable) {
    DEBUG(dbgs() << " Not enough local memory to promote alloca.\n");
    return;
  }

  std::vector<Value*> WorkList;

  if (!collectUsesWithPtrTypes(&I, WorkList)) {
    DEBUG(dbgs() << " Do not know how to convert all uses\n");
    return;
  }

  DEBUG(dbgs() << "Promoting alloca to local memory\n");
  LocalMemAvailable -= AllocaSize;

  GlobalVariable *GV = new GlobalVariable(
      *Mod, ArrayType::get(I.getAllocatedType(), 256), false,
      GlobalValue::ExternalLinkage, 0, I.getName(), 0,
      GlobalVariable::NotThreadLocal, AMDGPUAS::LOCAL_ADDRESS);

  FunctionType *FTy = FunctionType::get(
      Type::getInt32Ty(Mod->getContext()), false);
  AttributeSet AttrSet;
  AttrSet.addAttribute(Mod->getContext(), 0, Attribute::ReadNone);

  Value *ReadLocalSizeY = Mod->getOrInsertFunction(
      "llvm.r600.read.local.size.y", FTy, AttrSet);
  Value *ReadLocalSizeZ = Mod->getOrInsertFunction(
      "llvm.r600.read.local.size.z", FTy, AttrSet);
  Value *ReadTIDIGX = Mod->getOrInsertFunction(
      "llvm.r600.read.tidig.x", FTy, AttrSet);
  Value *ReadTIDIGY = Mod->getOrInsertFunction(
      "llvm.r600.read.tidig.y", FTy, AttrSet);
  Value *ReadTIDIGZ = Mod->getOrInsertFunction(
      "llvm.r600.read.tidig.z", FTy, AttrSet);


  Value *TCntY = Builder.CreateCall(ReadLocalSizeY);
  Value *TCntZ = Builder.CreateCall(ReadLocalSizeZ);
  Value *TIdX  = Builder.CreateCall(ReadTIDIGX);
  Value *TIdY  = Builder.CreateCall(ReadTIDIGY);
  Value *TIdZ  = Builder.CreateCall(ReadTIDIGZ);

  Value *Tmp0 = Builder.CreateMul(TCntY, TCntZ);
  Tmp0 = Builder.CreateMul(Tmp0, TIdX);
  Value *Tmp1 = Builder.CreateMul(TIdY, TCntZ);
  Value *TID = Builder.CreateAdd(Tmp0, Tmp1);
  TID = Builder.CreateAdd(TID, TIdZ);

  std::vector<Value*> Indices;
  Indices.push_back(Constant::getNullValue(Type::getInt32Ty(Mod->getContext())));
  Indices.push_back(TID);

  Value *Offset = Builder.CreateGEP(GV, Indices);
  I.mutateType(Offset->getType());
  I.replaceAllUsesWith(Offset);
  I.eraseFromParent();

  for (std::vector<Value*>::iterator i = WorkList.begin(),
                                     e = WorkList.end(); i != e; ++i) {
    Value *V = *i;
    CallInst *Call = dyn_cast<CallInst>(V);
    if (!Call) {
      Type *EltTy = V->getType()->getPointerElementType();
      PointerType *NewTy = PointerType::get(EltTy, AMDGPUAS::LOCAL_ADDRESS);

      // The operand's value should be corrected on its own.
      if (isa<AddrSpaceCastInst>(V))
        continue;

      // FIXME: It doesn't really make sense to try to do this for all
      // instructions.
      V->mutateType(NewTy);
      continue;
    }

    IntrinsicInst *Intr = dyn_cast<IntrinsicInst>(Call);
    if (!Intr) {
      std::vector<Type*> ArgTypes;
      for (unsigned ArgIdx = 0, ArgEnd = Call->getNumArgOperands();
                                ArgIdx != ArgEnd; ++ArgIdx) {
        ArgTypes.push_back(Call->getArgOperand(ArgIdx)->getType());
      }
      Function *F = Call->getCalledFunction();
      FunctionType *NewType = FunctionType::get(Call->getType(), ArgTypes,
                                                F->isVarArg());
      Constant *C = Mod->getOrInsertFunction(StringRef(F->getName().str() + ".local"), NewType,
                                             F->getAttributes());
      Function *NewF = cast<Function>(C);
      Call->setCalledFunction(NewF);
      continue;
    }

    Builder.SetInsertPoint(Intr);
    switch (Intr->getIntrinsicID()) {
    case Intrinsic::lifetime_start:
    case Intrinsic::lifetime_end:
      // These intrinsics are for address space 0 only
      Intr->eraseFromParent();
      continue;
    case Intrinsic::memcpy: {
      MemCpyInst *MemCpy = cast<MemCpyInst>(Intr);
      Builder.CreateMemCpy(MemCpy->getRawDest(), MemCpy->getRawSource(),
                           MemCpy->getLength(), MemCpy->getAlignment(),
                           MemCpy->isVolatile());
      Intr->eraseFromParent();
      continue;
    }
    case Intrinsic::memset: {
      MemSetInst *MemSet = cast<MemSetInst>(Intr);
      Builder.CreateMemSet(MemSet->getRawDest(), MemSet->getValue(),
                           MemSet->getLength(), MemSet->getAlignment(),
                           MemSet->isVolatile());
      Intr->eraseFromParent();
      continue;
    }
    default:
      Intr->dump();
      llvm_unreachable("Don't know how to promote alloca intrinsic use.");
    }
  }
}
Example #9
0
bool TypeChecksOpt::runOnModule(Module &M) {
  TS = &getAnalysis<dsa::TypeSafety<TDDataStructures> >();

  // Create the necessary prototypes
  VoidTy = IntegerType::getVoidTy(M.getContext());
  Int8Ty = IntegerType::getInt8Ty(M.getContext());
  Int32Ty = IntegerType::getInt32Ty(M.getContext());
  Int64Ty = IntegerType::getInt64Ty(M.getContext());
  VoidPtrTy = PointerType::getUnqual(Int8Ty);
  TypeTagTy = Int8Ty;
  TypeTagPtrTy = PointerType::getUnqual(TypeTagTy);

  Constant *memsetF = M.getOrInsertFunction ("llvm.memset.i64", VoidTy,
                                             VoidPtrTy,
                                             Int8Ty,
                                             Int64Ty,
                                             Int32Ty,
                                             NULL);
  trackGlobal = M.getOrInsertFunction("trackGlobal",
                                      VoidTy,
                                      VoidPtrTy,/*ptr*/
                                      TypeTagTy,/*type*/
                                      Int64Ty,/*size*/
                                      Int32Ty,/*tag*/
                                      NULL);
  trackInitInst = M.getOrInsertFunction("trackInitInst",
                                        VoidTy,
                                        VoidPtrTy,/*ptr*/
                                        Int64Ty,/*size*/
                                        Int32Ty,/*tag*/
                                        NULL);
  trackUnInitInst = M.getOrInsertFunction("trackUnInitInst",
                                          VoidTy,
                                          VoidPtrTy,/*ptr*/
                                          Int64Ty,/*size*/
                                          Int32Ty,/*tag*/
                                          NULL);
  trackStoreInst = M.getOrInsertFunction("trackStoreInst",
                                         VoidTy,
                                         VoidPtrTy,/*ptr*/
                                         TypeTagTy,/*type*/
                                         Int64Ty,/*size*/
                                         Int32Ty,/*tag*/
                                         NULL);
  checkTypeInst = M.getOrInsertFunction("checkType",
                                        VoidTy,
                                        TypeTagTy,/*type*/
                                        Int64Ty,/*size*/
                                        TypeTagPtrTy,
                                        VoidPtrTy,/*ptr*/
                                        Int32Ty,/*tag*/
                                        NULL);
  copyTypeInfo = M.getOrInsertFunction("copyTypeInfo",
                                       VoidTy,
                                       VoidPtrTy,/*dest ptr*/
                                       VoidPtrTy,/*src ptr*/
                                       Int64Ty,/*size*/
                                       Int32Ty,/*tag*/
                                       NULL);
  setTypeInfo = M.getOrInsertFunction("setTypeInfo",
                                      VoidTy,
                                      VoidPtrTy,/*dest ptr*/
                                      TypeTagPtrTy,/*metadata*/
                                      Int64Ty,/*size*/
                                      TypeTagTy,
                                      VoidPtrTy,
                                      Int32Ty,/*tag*/
                                      NULL);
  trackStringInput = M.getOrInsertFunction("trackStringInput",
                                           VoidTy,
                                           VoidPtrTy,
                                           Int32Ty,
                                           NULL);
  getTypeTag = M.getOrInsertFunction("getTypeTag",
                                     VoidTy,
                                     VoidPtrTy, /*ptr*/
                                     Int64Ty, /*size*/
                                     TypeTagPtrTy, /*dest for type tag*/
                                     Int32Ty, /*tag*/
                                     NULL);
  MallocFunc = M.getFunction("malloc");

  for(Value::use_iterator User = trackGlobal->use_begin(); User != trackGlobal->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);
    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      std::vector<Value*>Args;
      Args.push_back(CI->getOperand(1));
      Args.push_back(CI->getOperand(3));
      Args.push_back(CI->getOperand(4));
      CallInst::Create(trackInitInst, Args, "", CI);
      toDelete.push_back(CI);
    }
  }

  for(Value::use_iterator User = checkTypeInst->use_begin(); User != checkTypeInst->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);

    if(TS->isTypeSafe(CI->getOperand(4)->stripPointerCasts(), CI->getParent()->getParent())) {
      toDelete.push_back(CI);
    }
  }

  for(Value::use_iterator User = trackStoreInst->use_begin(); User != trackStoreInst->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);

    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      toDelete.push_back(CI);
    }
  }

  // for alloca's if they are type known
  // assume initialized with TOP
  for(Value::use_iterator User = trackUnInitInst->use_begin(); User != trackUnInitInst->use_end(); ) {
    CallInst *CI = dyn_cast<CallInst>(*(User++));
    assert(CI);

    // check if operand is an alloca inst.
    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      CI->setCalledFunction(trackInitInst);

      if(AllocaInst *AI = dyn_cast<AllocaInst>(CI->getOperand(1)->stripPointerCasts())) {
        // Initialize the allocation to NULL
        std::vector<Value *> Args2;
        Args2.push_back(CI->getOperand(1));
        Args2.push_back(ConstantInt::get(Int8Ty, 0));
        Args2.push_back(CI->getOperand(2));
        Args2.push_back(ConstantInt::get(Int32Ty, AI->getAlignment()));
        CallInst::Create(memsetF, Args2, "", CI);
      }
    }
  }

  if(MallocFunc) {
    for(Value::use_iterator User = MallocFunc->use_begin(); User != MallocFunc->use_end(); User ++) {
      CallInst *CI = dyn_cast<CallInst>(*User);
      if(!CI)
        continue;
      if(TS->isTypeSafe(CI, CI->getParent()->getParent())){
        CastInst *BCI = BitCastInst::CreatePointerCast(CI, VoidPtrTy);
        CastInst *Size = CastInst::CreateSExtOrBitCast(CI->getOperand(1), Int64Ty);
        Size->insertAfter(CI);
        BCI->insertAfter(Size);
        std::vector<Value *>Args;
        Args.push_back(BCI);
        Args.push_back(Size);
        Args.push_back(ConstantInt::get(Int32Ty, 0));
        CallInst *CINew = CallInst::Create(trackInitInst, Args);
        CINew->insertAfter(BCI);
      }
    }
  }

  // also do for mallocs/calloc/other allocators???
  // other allocators??

  for(Value::use_iterator User = copyTypeInfo->use_begin(); User != copyTypeInfo->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);

    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      std::vector<Value*> Args;
      Args.push_back(CI->getOperand(1));
      Args.push_back(CI->getOperand(3)); // size
      Args.push_back(CI->getOperand(4));
      CallInst::Create(trackInitInst, Args, "", CI);
      toDelete.push_back(CI);
    }
  }
  for(Value::use_iterator User = setTypeInfo->use_begin(); User != setTypeInfo->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);

    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      std::vector<Value*> Args;
      Args.push_back(CI->getOperand(1));
      Args.push_back(CI->getOperand(3)); // size
      Args.push_back(CI->getOperand(6));
      CallInst::Create(trackInitInst, Args, "", CI);
      toDelete.push_back(CI);
    }
  }

  for(Value::use_iterator User = getTypeTag->use_begin(); User != getTypeTag->use_end(); ++User) {
    CallInst *CI = dyn_cast<CallInst>(*User);
    assert(CI);
    if(TS->isTypeSafe(CI->getOperand(1)->stripPointerCasts(), CI->getParent()->getParent())) {
      AllocaInst *AI = dyn_cast<AllocaInst>(CI->getOperand(3)->stripPointerCasts());
      assert(AI);
      std::vector<Value*>Args;
      Args.push_back(CI->getOperand(3));
      Args.push_back(ConstantInt::get(Int8Ty, 255));
      Args.push_back(CI->getOperand(2));
      Args.push_back(ConstantInt::get(Int32Ty, AI->getAlignment()));
      CallInst::Create(memsetF, Args, "", CI);
      toDelete.push_back(CI);
    }
  }

  numSafe += toDelete.size();

  while(!toDelete.empty()) {
    Instruction *I = toDelete.back();
    toDelete.pop_back();
    I->eraseFromParent();
  }

  return (numSafe > 0);
}