Lower signatures exposing struct registers to byval struct pointers

lib/Transforms/NaCl/SimplifyStructRegSignatures.cpp

+//===- SimplifyStructRegSignatures.cpp - struct regs to struct pointers----===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This pass replaces function signatures exposing struct registers
+// to byval pointer-based signatures.
+//
+// There are 2 types of signatures that are thus changed:
+//
+// @foo(%some_struct %val) -> @foo(%some_struct* byval %val)
+//      and
+// %someStruct @bar(<other_args>) -> void @bar(%someStruct* sret, <other_args>)
+//
+// Such function types may appear in other type declarations, for example:
+//
+// %a_struct = type { void (%some_struct)*, i32 }
+//
+// We map such types to corresponding types, mapping the function types
+// appropriately:
+//
+// %a_struct.0 = type { void (%some_struct*)*, i32 }
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ADT/SmallString.h"
+#include "llvm/IR/IRBuilder.h"
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/ilist.h"
+#include "llvm/ADT/SetVector.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/Twine.h"
+#include "llvm/IR/Argument.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/BasicBlock.h"
+#include "llvm/IR/DebugInfo.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/Module.h"
+#include "llvm/IR/Type.h"
+#include "llvm/IR/Use.h"
+#include "llvm/IR/User.h"
+#include "llvm/IR/Value.h"
+#include "llvm/Pass.h"
+#include "llvm/PassInfo.h"
+#include "llvm/PassRegistry.h"
+#include "llvm/PassSupport.h"
+#include "llvm/Transforms/NaCl.h"
+#include "llvm/Support/Debug.h"
+
+#include <cassert>
+#include <cstddef>
+
+using namespace llvm;
+
+namespace {
+
+static const unsigned int TypicalFuncArity = 8;
+static const unsigned int TypicalStructArity = 8;
+
+class MappingResult {
+public:
+  MappingResult(Type *ATy, bool Chg) {
+    Ty = ATy;
+    Changed = Chg;
+  }
+
+  bool isChanged() { return Changed; }
+
+  Type *operator->() { return Ty; }
+
+  operator Type *() { return Ty; }
+
+private:
+  Type *Ty;
+  bool Changed;
+};
+
+// Utility class. For any given type, get the associated type that is free of
+// struct register arguments.
+class TypeMapper {
+public:
+  typedef DenseMap<StructType *, StructType *> StructMap;
+  Type *getSimpleType(LLVMContext &Ctx, Type *Ty);
+
+private:
+  DenseMap<Type *, Type *> MappedTypes;
+  MappingResult getSimpleArgumentType(LLVMContext &Ctx, Type *Ty,
+                                      StructMap &Tentatives);
+  MappingResult getSimpleAggregateTypeInternal(LLVMContext &Ctx, Type *Ty,
+                                               StructMap &Tentatives);
+
+  bool isChangedStruct(LLVMContext &Ctx, StructType *StructTy,
+                       SmallVector<Type *, TypicalStructArity> &ElemTypes,
+                       StructMap &Tentatives);
+};
+
+// This is a ModulePass because the pass recreates functions in
+// order to change their signatures.
+class SimplifyStructRegSignatures : public ModulePass {
+public:
+  static char ID;
+
+  SimplifyStructRegSignatures() : ModulePass(ID) {
+    initializeSimplifyStructRegSignaturesPass(*PassRegistry::getPassRegistry());
+  }
+  virtual bool runOnModule(Module &M);
+
+private:
+  TypeMapper Mapper;
+  DenseSet<Function *> FunctionsToDelete;
+  SetVector<CallInst *> CallsToPatch;
+  SetVector<InvokeInst *> InvokesToPatch;
+  DenseMap<Function *, Function *> FunctionMap;
+  bool
+  simplifyFunction(LLVMContext &Ctx, Function *OldFunc,
+                   DenseMap<const Function *, DISubprogram> &DISubprogramMap);
+  void scheduleInstructionsForCleanup(Function *NewFunc);
+  template <class TCall>
+  void fixCallSite(LLVMContext &Ctx, TCall *Call, unsigned PreferredAlignment);
+  void fixFunctionBody(LLVMContext &Ctx, Function *OldFunc, Function *NewFunc);
+
+  template <class TCall>
+  TCall *fixCallTargetAndArguments(LLVMContext &Ctx, IRBuilder<> &Builder,
+                                   TCall *OldCall, Value *NewTarget,
+                                   FunctionType *NewType,
+                                   BasicBlock::iterator AllocaInsPoint,
+                                   Value *ExtraArg = nullptr);
+  void checkNoUnsupportedInstructions(LLVMContext &Ctx, Function *Fct);
+};
+}
+
+char SimplifyStructRegSignatures::ID = 0;
+
+INITIALIZE_PASS(
+    SimplifyStructRegSignatures, "simplify-struct-reg-signatures",
+    "Simplify function signatures by removing struct register parameters",
+    false, false)
+
+// The type is "simple" if it does not recursively reference a
+// function type with at least an operand (arg or return) typed as struct
+// register.
+Type *TypeMapper::getSimpleType(LLVMContext &Ctx, Type *Ty) {
+  auto Found = MappedTypes.find(Ty);
+  if (Found != MappedTypes.end()) {
+    return Found->second;
+  }
+
+  StructMap Tentatives;
+  auto Ret = getSimpleAggregateTypeInternal(Ctx, Ty, Tentatives);
+  assert(Tentatives.size() == 0);
+
+  if (!Ty->isStructTy()) {
+    // Structs are memoized in getSimpleAggregateTypeInternal.
+    MappedTypes[Ty] = Ret;
+  }
+  return Ret;
+}
+
+// Transforms any type that could transitively reference a function pointer
+// into a simplified type.
+// We enter this function trying to determine the mapping of a type. Because
+// of how structs are handled (not interned by llvm - see further comments
+// below) we may be working with temporary types - types (pointers, for example)
+// transitively referencing "tentative" structs. For that reason, we do not
+// memoize anything here, except for structs. The latter is so that we avoid
+// unnecessary repeated creation of types (pointers, function types, etc),
+// as we try to map a given type.
+MappingResult
+TypeMapper::getSimpleAggregateTypeInternal(LLVMContext &Ctx, Type *Ty,
+                                           StructMap &Tentatives) {
+  // Leverage the map for types we encounter on the way.
+  auto Found = MappedTypes.find(Ty);
+  if (Found != MappedTypes.end()) {
+    return {Found->second, Found->second != Ty};
+  }
+
+  if (auto *OldFnTy = dyn_cast<FunctionType>(Ty)) {
+    Type *OldRetType = OldFnTy->getReturnType();
+    Type *NewRetType = OldRetType;
+    Type *Void = Type::getVoidTy(Ctx);
+    SmallVector<Type *, TypicalFuncArity> NewArgs;
+    bool Changed = false;
+    // Struct register returns become the first parameter of the new FT.
+    // The new FT has void for the return type
+    if (OldRetType->isAggregateType()) {
+      NewRetType = Void;
+      Changed = true;
+      NewArgs.push_back(getSimpleArgumentType(Ctx, OldRetType, Tentatives));
+    }
+    for (auto OldParam : OldFnTy->params()) {
+      auto NewType = getSimpleArgumentType(Ctx, OldParam, Tentatives);
+      Changed |= NewType.isChanged();
+      NewArgs.push_back(NewType);
+    }
+    Type *NewFuncType =
+        FunctionType::get(NewRetType, NewArgs, OldFnTy->isVarArg());
+    return {NewFuncType, Changed};
+  }
+
+  if (auto PtrTy = dyn_cast<PointerType>(Ty)) {
+    auto NewTy = getSimpleAggregateTypeInternal(
+        Ctx, PtrTy->getPointerElementType(), Tentatives);
+
+    return {NewTy->getPointerTo(PtrTy->getAddressSpace()), NewTy.isChanged()};
+  }
+
+  if (auto ArrTy = dyn_cast<ArrayType>(Ty)) {
+    auto NewTy = getSimpleAggregateTypeInternal(
+        Ctx, ArrTy->getArrayElementType(), Tentatives);
+    return {ArrayType::get(NewTy, ArrTy->getArrayNumElements()),
+            NewTy.isChanged()};
+  }
+
+  if (auto VecTy = dyn_cast<VectorType>(Ty)) {
+    auto NewTy = getSimpleAggregateTypeInternal(
+        Ctx, VecTy->getVectorElementType(), Tentatives);
+    return {VectorType::get(NewTy, VecTy->getVectorNumElements()),
+            NewTy.isChanged()};
+  }
+
+  // LLVM doesn't intern identified structs (the ones with a name). This,
+  // together with the fact that such structs can be recursive,
+  // complicates things a bit. We want to make sure that we only change
+  // "unsimplified" structs (those that somehow reference funcs that
+  // are not simple).
+  // We don't want to change "simplified" structs, otherwise converting
+  // instruction types will become trickier.
+  if (auto StructTy = dyn_cast<StructType>(Ty)) {
+    SmallVector<Type *, TypicalStructArity> ElemTypes;
+    if (!StructTy->isLiteral()) {
+      // Literals - struct without a name - cannot be recursive, so we
+      // don't need to form tentatives.
+      auto Found = Tentatives.find(StructTy);
+
+      // Having a tentative means we are in a recursion trying to map this
+      // particular struct, so arriving back to it is not a change.
+      // We will determine if this struct is actually
+      // changed by checking its other fields.
+      if (Found != Tentatives.end()) {
+        return {Found->second, false};
+      }
+      // We have never seen this struct, so we start a tentative.
+      std::string NewName = StructTy->getStructName();
+      NewName += ".simplified";
+      StructType *Tentative = StructType::create(Ctx, NewName);
+      Tentatives[StructTy] = Tentative;
+
+      bool Changed = isChangedStruct(Ctx, StructTy, ElemTypes, Tentatives);
+
+      Tentatives.erase(StructTy);
+      // We can now decide the mapping of the struct. We will register it
+      // early with MappedTypes, to avoid leaking tentatives unnecessarily.
+      // We are leaking the created struct here, but there is no way to
+      // correctly delete it.
+      if (!Changed) {
+        return {MappedTypes[StructTy] = StructTy, false};
+      } else {
+        Tentative->setBody(ElemTypes, StructTy->isPacked());
+        return {MappedTypes[StructTy] = Tentative, true};
+      }
+    } else {
+      bool Changed = isChangedStruct(Ctx, StructTy, ElemTypes, Tentatives);
+      return {MappedTypes[StructTy] =
+                  StructType::get(Ctx, ElemTypes, StructTy->isPacked()),
+              Changed};
+    }
+  }
+
+  // Anything else stays the same.
+  return {Ty, false};
+}
+
+bool TypeMapper::isChangedStruct(
+    LLVMContext &Ctx, StructType *StructTy,
+    SmallVector<Type *, TypicalStructArity> &ElemTypes, StructMap &Tentatives) {
+  bool Changed = false;
+  unsigned StructElemCount = StructTy->getStructNumElements();
+  for (unsigned I = 0; I < StructElemCount; I++) {
+    auto NewElem = getSimpleAggregateTypeInternal(
+        Ctx, StructTy->getStructElementType(I), Tentatives);
+    ElemTypes.push_back(NewElem);
+    Changed |= NewElem.isChanged();
+  }
+  return Changed;
+}
+
+// Get the simplified type of a function argument.
+MappingResult TypeMapper::getSimpleArgumentType(LLVMContext &Ctx, Type *Ty,
+                                                StructMap &Tentatives) {
+  // struct registers become pointers to simple structs
+  if (Ty->isAggregateType()) {
+    return MappingResult(
+        PointerType::get(getSimpleAggregateTypeInternal(Ctx, Ty, Tentatives),
+                         0),
+        true);
+  }
+
+  return getSimpleAggregateTypeInternal(Ctx, Ty, Tentatives);
+}
+
+// Apply 'byval' to func arguments that used to be struct regs.
+// Apply 'sret' to the argument corresponding to the return in the old
+// signature.
+static void ApplyByValAndSRet(Function *OldFunc, Function *NewFunc) {
+  // When calling addAttribute, the first one refers to the function, so we
+  // skip past that.
+  unsigned ArgOffset = 1;
+  if (OldFunc->getReturnType()->isAggregateType()) {
+    NewFunc->addAttribute(1, Attribute::AttrKind::StructRet);
+    ArgOffset++;
+  }
+
+  auto &NewArgList = NewFunc->getArgumentList();
+  auto NewArg = NewArgList.begin();
+  for (const Argument &OldArg : OldFunc->getArgumentList()) {
+    if (OldArg.getType()->isAggregateType()) {
+      NewFunc->addAttribute(NewArg->getArgNo() + ArgOffset,
+                            Attribute::AttrKind::ByVal);
+    }
+    NewArg++;
+  }
+}
+
+// Update the arg names for a newly created function.
+static void UpdateArgNames(Function *OldFunc, Function *NewFunc) {
+  auto NewArgIter = NewFunc->arg_begin();
+  if (OldFunc->getReturnType()->isAggregateType()) {
+    NewArgIter->setName("retVal");
+    NewArgIter++;
+  }
+
+  for (const Argument &OldArg : OldFunc->args()) {
+    Argument *NewArg = NewArgIter++;
+    NewArg->setName(OldArg.getName() +
+                    (OldArg.getType()->isAggregateType() ? ".ptr" : ""));
+  }
+}
+
+// Replace all uses of an old value with a new one, disregarding the type. We
+// correct the types after we wire the new parameters in, in fixFunctionBody.
+static void BlindReplace(Value *Old, Value *New) {
+  for (auto UseIter = Old->use_begin(), E = Old->use_end(); E != UseIter;) {
+    Use &AUse = *(UseIter++);
+    AUse.set(New);
+  }
+}
+
+// Adapt the body of a function for the new arguments.
+static void ConvertArgumentValue(Value *Old, Value *New,
+                                 Instruction *InsPoint) {
+  if (Old == New)
+    return;
+
+  if (Old->getType() == New->getType()) {
+    Old->replaceAllUsesWith(New);
+    New->takeName(Old);
+    return;
+  }
+
+  bool IsAggregateToPtr =
+      Old->getType()->isAggregateType() && New->getType()->isPointerTy();
+  BlindReplace(Old, (IsAggregateToPtr
+                         ? new LoadInst(New, Old->getName() + ".sreg", InsPoint)
+                         : New));
+}
+
+// Fix returns. Return true if fixes were needed.
+static void FixReturn(Function *OldFunc, Function *NewFunc) {
+
+  Argument *FirstNewArg = NewFunc->getArgumentList().begin();
+
+  for (auto BIter = NewFunc->begin(), LastBlock = NewFunc->end();
+       LastBlock != BIter;) {
+    BasicBlock *BB = BIter++;
+    for (auto IIter = BB->begin(), LastI = BB->end(); LastI != IIter;) {
+      Instruction *Instr = IIter++;
+      if (ReturnInst *Ret = dyn_cast<ReturnInst>(Instr)) {
+        auto RetVal = Ret->getReturnValue();
+        IRBuilder<> Builder(Ret);
+        StoreInst *Store = Builder.CreateStore(RetVal, FirstNewArg);
+        Store->setAlignment(FirstNewArg->getParamAlignment());
+        Builder.CreateRetVoid();
+        Ret->eraseFromParent();
+      }
+    }
+  }
+}
+
+// TODO (mtrofin): is this comprehensive?
+template <class TCall>
+void CopyCallAttributesAndMetadata(TCall *Orig, TCall *NewCall) {
+  NewCall->setCallingConv(Orig->getCallingConv());
+  NewCall->setAttributes(NewCall->getAttributes().addAttributes(
+      Orig->getContext(), AttributeSet::FunctionIndex,
+      Orig->getAttributes().getFnAttributes()));
+  NewCall->takeName(Orig);
+}
+
+static InvokeInst *CreateCallFrom(InvokeInst *Orig, Value *Target,
+                                  ArrayRef<Value *> &Args,
+                                  IRBuilder<> &Builder) {
+  auto Ret = Builder.CreateInvoke(Target, Orig->getNormalDest(),
+                                  Orig->getUnwindDest(), Args);
+  CopyCallAttributesAndMetadata(Orig, Ret);
+  return Ret;
+}
+
+static CallInst *CreateCallFrom(CallInst *Orig, Value *Target,
+                                ArrayRef<Value *> &Args, IRBuilder<> &Builder) {
+
+  CallInst *Ret = Builder.CreateCall(Target, Args);
+  Ret->setTailCallKind(Orig->getTailCallKind());
+  CopyCallAttributesAndMetadata(Orig, Ret);
+  return Ret;
+}
+
+// Fix a call site by handing return type changes and/or parameter type and
+// attribute changes.
+template <class TCall>
+void SimplifyStructRegSignatures::fixCallSite(LLVMContext &Ctx, TCall *OldCall,
+                                              unsigned PreferredAlignment) {
+  Value *NewTarget = OldCall->getCalledValue();
+
+  if (Function *CalledFunc = dyn_cast<Function>(NewTarget)) {
+    NewTarget = this->FunctionMap[CalledFunc];
+  }
+  assert(NewTarget);
+
+  auto *NewType = cast<FunctionType>(
+      Mapper.getSimpleType(Ctx, NewTarget->getType())->getPointerElementType());
+
+  auto *OldRetType = OldCall->getType();
+  const bool IsSRet =
+      !OldCall->getType()->isVoidTy() && NewType->getReturnType()->isVoidTy();
+
+  IRBuilder<> Builder(OldCall);
+  auto AllocaInsPoint =
+      OldCall->getParent()->getParent()->getEntryBlock().getFirstInsertionPt();
+
+  if (IsSRet) {
+    // Insert the Alloca at the beginning of the function, to ensure
+    // we do not leak space if the call is in a loop.
+    auto SavedInsPoint = Builder.GetInsertPoint();
+    Builder.SetInsertPoint(AllocaInsPoint);
+    auto *Alloca = Builder.CreateAlloca(OldRetType);
+    AllocaInsPoint = Builder.GetInsertPoint();
+    Builder.SetInsertPoint(SavedInsPoint);
+
+    Alloca->takeName(OldCall);
+    Alloca->setAlignment(PreferredAlignment);
+
+    auto *NewCall = fixCallTargetAndArguments(Ctx, Builder, OldCall, NewTarget,
+                                              NewType, AllocaInsPoint, Alloca);
+    assert(NewCall);
+    if (auto *Invoke = dyn_cast<InvokeInst>(OldCall))
+      Builder.SetInsertPoint(Invoke->getNormalDest()->getFirstInsertionPt());
+
+    auto *Load = Builder.CreateLoad(Alloca, Alloca->getName() + ".sreg");
+    Load->setAlignment(Alloca->getAlignment());
+    OldCall->replaceAllUsesWith(Load);
+  } else {
+    auto *NewCall = fixCallTargetAndArguments(Ctx, Builder, OldCall, NewTarget,
+                                              NewType, AllocaInsPoint);
+    OldCall->replaceAllUsesWith(NewCall);
+  }
+
+  OldCall->eraseFromParent();
+}
+
+template <class TCall>
+TCall *SimplifyStructRegSignatures::fixCallTargetAndArguments(
+    LLVMContext &Ctx, IRBuilder<> &Builder, TCall *OldCall, Value *NewTarget,
+    FunctionType *NewType, BasicBlock::iterator AllocaInsPoint,
+    Value *ExtraArg) {
+  SmallSetVector<unsigned, TypicalFuncArity> ByRefPlaces;
+  SmallVector<Value *, TypicalFuncArity> NewArgs;
+
+  unsigned argOffset = ExtraArg ? 1 : 0;
+  if (ExtraArg)
+    NewArgs.push_back(ExtraArg);
+
+  // Go over the argument list used in the call/invoke, in order to
+  // correctly deal with varargs scenarios.

[JF, 2015/03/20 16:41:32]

We decided to avoid handling struct passed in vararg because C disallows it. IR
doesn't disallow it, so could you make sure you report fatal error on this?

[Mircea Trofin, 2015/03/20 18:31:32]

Done.

+  unsigned NumActualParams = OldCall->getNumArgOperands();
+  for (unsigned ArgPos = 0; ArgPos < NumActualParams; ArgPos++) {
+
+    Use &OldArgUse = OldCall->getOperandUse(ArgPos);
+    Value *OldArg = OldArgUse;
+    Type *OldArgType = OldArg->getType();
+    unsigned NewArgPos = OldArgUse.getOperandNo() + argOffset;
+    Type *NewArgType = NewType->getFunctionParamType(NewArgPos);
+
+    if (OldArgType != NewArgType && OldArgType->isAggregateType()) {
+      // Insert the Alloca at the beginning of the function, to ensure
+      // we do not leak space if the call is in a loop.
+      auto SavedInsPoint = Builder.GetInsertPoint();
+      Builder.SetInsertPoint(AllocaInsPoint);
+
+      AllocaInst *Alloca =
+          Builder.CreateAlloca(OldArgType, nullptr, OldArg->getName() + ".ptr");
+      AllocaInsPoint = Builder.GetInsertPoint();
+      // Continue with the store at the saved insertion point.
+      Builder.SetInsertPoint(SavedInsPoint);

[JF, 2015/03/20 16:41:32]

I think it would be better to pull this out as a function which creates the
alloca, and returns a tuple with {Alloca, InsPoint}, and use the function here
and above.

[Mircea Trofin, 2015/03/20 18:31:32]

Done.

+      Builder.CreateStore(OldArg, Alloca);
+      ByRefPlaces.insert(NewArgPos);
+      NewArgs.push_back(Alloca);
+    } else {
+      NewArgs.push_back(OldArg);
+    }
+  }
+
+  ArrayRef<Value *> ArrRef = NewArgs;
+  TCall *NewCall = CreateCallFrom(OldCall, NewTarget, ArrRef, Builder);
+
+  // Copy the attributes over, and add byref/sret as necessary.
+  const AttributeSet &OldAttrSet = OldCall->getAttributes();
+  const AttributeSet &NewAttrSet = NewCall->getAttributes();
+
+  for (unsigned I = 0; I < NewCall->getNumArgOperands(); I++) {
+    NewCall->setAttributes(NewAttrSet.addAttributes(
+        Ctx, I + argOffset + 1, OldAttrSet.getParamAttributes(I + 1)));
+    if (ByRefPlaces.count(I)) {
+      NewCall->addAttribute(I + 1, Attribute::ByVal);
+    }
+  }
+
+  if (ExtraArg) {
+    NewAttrSet.addAttributes(Ctx, 1, OldAttrSet.getRetAttributes());
+    NewCall->addAttribute(1, Attribute::StructRet);
+  } else {
+    NewCall->setAttributes(NewAttrSet.addAttributes(
+        Ctx, AttributeSet::ReturnIndex, OldAttrSet.getRetAttributes()));
+  }
+  return NewCall;
+}
+
+void SimplifyStructRegSignatures::scheduleInstructionsForCleanup(
+    Function *NewFunc) {
+  for (auto &BBIter : NewFunc->getBasicBlockList()) {
+    for (auto &IIter : BBIter.getInstList()) {
+      if (CallInst *Call = dyn_cast<CallInst>(&IIter)) {
+        CallsToPatch.insert(Call);
+      } else if (InvokeInst *Invoke = dyn_cast<InvokeInst>(&IIter)) {
+        InvokesToPatch.insert(Invoke);
+      }
+    }
+  }
+}
+
+// Change function body in the light of type changes.
+void SimplifyStructRegSignatures::fixFunctionBody(LLVMContext &Ctx,
+                                                  Function *OldFunc,
+                                                  Function *NewFunc) {
+  if (NewFunc->empty())
+    return;
+
+  bool returnWasFixed = OldFunc->getReturnType()->isAggregateType();
+
+  Instruction *InsPoint = NewFunc->begin()->begin();
+  auto NewArgIter = NewFunc->arg_begin();
+  // Advance one more if we used to return a struct register.
+  if (returnWasFixed)
+    NewArgIter++;
+
+  // Wire new parameters in.
+  for (auto ArgIter = OldFunc->arg_begin(), E = OldFunc->arg_end();
+       E != ArgIter;) {
+    Argument *OldArg = ArgIter++;
+    Argument *NewArg = NewArgIter++;
+    ConvertArgumentValue(OldArg, NewArg, InsPoint);
+  }
+
+  // Now fix instruction types. We know that each value could only possibly be
+  // of a simplified type. At the end of this, call sites will be invalid, but
+  // we handle that afterwards, to make sure we have all the functions changed
+  // first (so that calls have valid targets)
+  for (auto BBIter = NewFunc->begin(), LBlock = NewFunc->end();
+       LBlock != BBIter;) {
+    auto Block = BBIter++;
+    for (auto IIter = Block->begin(), LIns = Block->end(); LIns != IIter;) {
+      auto Instr = IIter++;
+      Instr->mutateType(Mapper.getSimpleType(Ctx, Instr->getType()));
+    }
+  }
+  if (returnWasFixed)
+    FixReturn(OldFunc, NewFunc);
+}
+
+// Ensure function is simplified, returning true if the function
+// had to be changed.
+bool SimplifyStructRegSignatures::simplifyFunction(
+    LLVMContext &Ctx, Function *OldFunc,
+    DenseMap<const Function *, DISubprogram> &DISubprogramMap) {
+  auto *OldFT = OldFunc->getFunctionType();
+  auto *NewFT = cast<FunctionType>(Mapper.getSimpleType(Ctx, OldFT));
+
+  Function *&AssociatedFctLoc = FunctionMap[OldFunc];
+  if (NewFT != OldFT) {
+    auto *NewFunc = Function::Create(NewFT, OldFunc->getLinkage());
+    AssociatedFctLoc = NewFunc;
+
+    NewFunc->copyAttributesFrom(OldFunc);
+    OldFunc->getParent()->getFunctionList().insert(OldFunc, NewFunc);
+    NewFunc->takeName(OldFunc);
+
+    UpdateArgNames(OldFunc, NewFunc);
+    ApplyByValAndSRet(OldFunc, NewFunc);
+
+    NewFunc->getBasicBlockList().splice(NewFunc->begin(),
+                                        OldFunc->getBasicBlockList());
+
+    fixFunctionBody(Ctx, OldFunc, NewFunc);
+    FunctionsToDelete.insert(OldFunc);
+    auto Found = DISubprogramMap.find(OldFunc);
+    if (Found != DISubprogramMap.end())
+      Found->second.replaceFunction(NewFunc);
+  } else {
+    AssociatedFctLoc = OldFunc;
+  }
+  scheduleInstructionsForCleanup(AssociatedFctLoc);
+  return NewFT != OldFT;
+}
+
+bool SimplifyStructRegSignatures::runOnModule(Module &M) {
+  bool Changed = false;
+
+  const DataLayout *DL = M.getDataLayout();
+  unsigned PreferredAlignment = 0;
+  if (DL)
+    PreferredAlignment = DL->getStackAlignment();
+
+  LLVMContext &Ctx = M.getContext();
+  auto DISubprogramMap = makeSubprogramMap(M);
+
+  // Change function signatures and fix a changed function body by
+  // wiring the new arguments. Call sites are unchanged at this point.
+  for (Module::iterator Iter = M.begin(), E = M.end(); Iter != E;) {
+    Function *Func = Iter++;
+    checkNoUnsupportedInstructions(Ctx, Func);
+    Changed |= simplifyFunction(Ctx, Func, DISubprogramMap);
+  }
+
+  // Fix call sites.
+  for (auto &CallToFix : CallsToPatch) {
+    fixCallSite(Ctx, CallToFix, PreferredAlignment);
+  }
+
+  for (auto &InvokeToFix : InvokesToPatch) {
+    fixCallSite(Ctx, InvokeToFix, PreferredAlignment);
+  }
+
+  // Delete leftover functions - the ones with old signatures.
+  for (auto &ToDelete : FunctionsToDelete) {
+    ToDelete->eraseFromParent();
+  }
+
+  return Changed;
+}
+
+void SimplifyStructRegSignatures::checkNoUnsupportedInstructions(
+    LLVMContext &Ctx, Function *Fct) {
+  for (auto &BB : Fct->getBasicBlockList())
+    for (auto &Inst : BB.getInstList())
+      if (auto *Landing = dyn_cast<LandingPadInst>(&Inst)) {
+        auto *LType = Landing->getPersonalityFn()->getType();
+        if (LType != Mapper.getSimpleType(Ctx, LType))

[JF, 2015/03/20 16:41:32]

Before report_fatal_error you should:
errs() << *Landing << '\n';

Same below.

[Mircea Trofin, 2015/03/20 18:31:32]

Done.

+          llvm::report_fatal_error("Landing pads with aggregate register "
+                                   "signatures are not supported.",
+                                   true);

[JF, 2015/03/20 16:41:32]

You don't need llvm:: here, and true is the default so you can drop it.

[Mircea Trofin, 2015/03/20 18:31:32]

Done.

+      } else if (auto *Resume = dyn_cast<ResumeInst>(&Inst)) {
+        auto *RType = Resume->getValue()->getType();
+        if (RType != Mapper.getSimpleType(Ctx, RType))
+          llvm::report_fatal_error(
+              "Resumes with aggregate register signatures are not supported.",
+              true);
+      }
+}
+
+ModulePass *llvm::createSimplifyStructRegSignaturesPass() {
+  return new SimplifyStructRegSignatures();
+}