Update LLVM to r108243.

18 files changed, 321 insertions, 516 deletions

diff --git a/lib/Transforms/Utils/AddrModeMatcher.cpp b/lib/Transforms/Utils/AddrModeMatcher.cpp
index ea9d1c1..4d64c85 100644
--- a/lib/Transforms/Utils/AddrModeMatcher.cpp
+++ b/lib/Transforms/Utils/AddrModeMatcher.cpp
@@ -381,29 +381,28 @@ static bool IsOperandAMemoryOperand(CallInst *CI, InlineAsm *IA, Value *OpVal,
                                     const TargetLowering &TLI) {
   std::vector<InlineAsm::ConstraintInfo>
   Constraints = IA->ParseConstraints();
-  
-  unsigned ArgNo = 1;   // ArgNo - The operand of the CallInst.
+
+  unsigned ArgNo = 0;   // The argument of the CallInst.
   for (unsigned i = 0, e = Constraints.size(); i != e; ++i) {
     TargetLowering::AsmOperandInfo OpInfo(Constraints[i]);
-    
+
     // Compute the value type for each operand.
     switch (OpInfo.Type) {
       case InlineAsm::isOutput:
         if (OpInfo.isIndirect)
-          OpInfo.CallOperandVal = CI->getOperand(ArgNo++);
+          OpInfo.CallOperandVal = CI->getArgOperand(ArgNo++);
         break;
       case InlineAsm::isInput:
-        OpInfo.CallOperandVal = CI->getOperand(ArgNo++);
+        OpInfo.CallOperandVal = CI->getArgOperand(ArgNo++);
         break;
       case InlineAsm::isClobber:
         // Nothing to do.
         break;
     }
-    
+
     // Compute the constraint code and ConstraintType to use.
-    TLI.ComputeConstraintToUse(OpInfo, SDValue(),
-                             OpInfo.ConstraintType == TargetLowering::C_Memory);
-    
+    TLI.ComputeConstraintToUse(OpInfo, SDValue());
+
     // If this asm operand is our Value*, and if it isn't an indirect memory
     // operand, we can't fold it!
     if (OpInfo.CallOperandVal == OpVal &&
@@ -411,7 +410,7 @@ static bool IsOperandAMemoryOperand(CallInst *CI, InlineAsm *IA, Value *OpVal,
          !OpInfo.isIndirect))
       return false;
   }
-  
+
   return true;
 }
 
@@ -450,7 +449,7 @@ static bool FindAllMemoryUses(Instruction *I,
     
     if (CallInst *CI = dyn_cast<CallInst>(U)) {
       InlineAsm *IA = dyn_cast<InlineAsm>(CI->getCalledValue());
-      if (IA == 0) return true;
+      if (!IA) return true;
       
       // If this is a memory operand, we're cool, otherwise bail out.
       if (!IsOperandAMemoryOperand(CI, IA, I, TLI))
diff --git a/lib/Transforms/Utils/BasicBlockUtils.cpp b/lib/Transforms/Utils/BasicBlockUtils.cpp
index 2f1ae00..ec625b4 100644
--- a/lib/Transforms/Utils/BasicBlockUtils.cpp
+++ b/lib/Transforms/Utils/BasicBlockUtils.cpp
@@ -558,121 +558,3 @@ void llvm::FindFunctionBackedges(const Function &F,
   
   
 }
-
-
-
-/// AreEquivalentAddressValues - Test if A and B will obviously have the same
-/// value. This includes recognizing that %t0 and %t1 will have the same
-/// value in code like this:
-///   %t0 = getelementptr \@a, 0, 3
-///   store i32 0, i32* %t0
-///   %t1 = getelementptr \@a, 0, 3
-///   %t2 = load i32* %t1
-///
-static bool AreEquivalentAddressValues(const Value *A, const Value *B) {
-  // Test if the values are trivially equivalent.
-  if (A == B) return true;
-  
-  // Test if the values come from identical arithmetic instructions.
-  // Use isIdenticalToWhenDefined instead of isIdenticalTo because
-  // this function is only used when one address use dominates the
-  // other, which means that they'll always either have the same
-  // value or one of them will have an undefined value.
-  if (isa<BinaryOperator>(A) || isa<CastInst>(A) ||
-      isa<PHINode>(A) || isa<GetElementPtrInst>(A))
-    if (const Instruction *BI = dyn_cast<Instruction>(B))
-      if (cast<Instruction>(A)->isIdenticalToWhenDefined(BI))
-        return true;
-  
-  // Otherwise they may not be equivalent.
-  return false;
-}
-
-/// FindAvailableLoadedValue - Scan the ScanBB block backwards (starting at the
-/// instruction before ScanFrom) checking to see if we have the value at the
-/// memory address *Ptr locally available within a small number of instructions.
-/// If the value is available, return it.
-///
-/// If not, return the iterator for the last validated instruction that the 
-/// value would be live through.  If we scanned the entire block and didn't find
-/// something that invalidates *Ptr or provides it, ScanFrom would be left at
-/// begin() and this returns null.  ScanFrom could also be left 
-///
-/// MaxInstsToScan specifies the maximum instructions to scan in the block.  If
-/// it is set to 0, it will scan the whole block. You can also optionally
-/// specify an alias analysis implementation, which makes this more precise.
-Value *llvm::FindAvailableLoadedValue(Value *Ptr, BasicBlock *ScanBB,
-                                      BasicBlock::iterator &ScanFrom,
-                                      unsigned MaxInstsToScan,
-                                      AliasAnalysis *AA) {
-  if (MaxInstsToScan == 0) MaxInstsToScan = ~0U;
-
-  // If we're using alias analysis to disambiguate get the size of *Ptr.
-  unsigned AccessSize = 0;
-  if (AA) {
-    const Type *AccessTy = cast<PointerType>(Ptr->getType())->getElementType();
-    AccessSize = AA->getTypeStoreSize(AccessTy);
-  }
-  
-  while (ScanFrom != ScanBB->begin()) {
-    // We must ignore debug info directives when counting (otherwise they
-    // would affect codegen).
-    Instruction *Inst = --ScanFrom;
-    if (isa<DbgInfoIntrinsic>(Inst))
-      continue;
-
-    // Restore ScanFrom to expected value in case next test succeeds
-    ScanFrom++;
-   
-    // Don't scan huge blocks.
-    if (MaxInstsToScan-- == 0) return 0;
-    
-    --ScanFrom;
-    // If this is a load of Ptr, the loaded value is available.
-    if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
-      if (AreEquivalentAddressValues(LI->getOperand(0), Ptr))
-        return LI;
-    
-    if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
-      // If this is a store through Ptr, the value is available!
-      if (AreEquivalentAddressValues(SI->getOperand(1), Ptr))
-        return SI->getOperand(0);
-      
-      // If Ptr is an alloca and this is a store to a different alloca, ignore
-      // the store.  This is a trivial form of alias analysis that is important
-      // for reg2mem'd code.
-      if ((isa<AllocaInst>(Ptr) || isa<GlobalVariable>(Ptr)) &&
-          (isa<AllocaInst>(SI->getOperand(1)) ||
-           isa<GlobalVariable>(SI->getOperand(1))))
-        continue;
-      
-      // If we have alias analysis and it says the store won't modify the loaded
-      // value, ignore the store.
-      if (AA &&
-          (AA->getModRefInfo(SI, Ptr, AccessSize) & AliasAnalysis::Mod) == 0)
-        continue;
-      
-      // Otherwise the store that may or may not alias the pointer, bail out.
-      ++ScanFrom;
-      return 0;
-    }
-    
-    // If this is some other instruction that may clobber Ptr, bail out.
-    if (Inst->mayWriteToMemory()) {
-      // If alias analysis claims that it really won't modify the load,
-      // ignore it.
-      if (AA &&
-          (AA->getModRefInfo(Inst, Ptr, AccessSize) & AliasAnalysis::Mod) == 0)
-        continue;
-      
-      // May modify the pointer, bail out.
-      ++ScanFrom;
-      return 0;
-    }
-  }
-  
-  // Got to the start of the block, we didn't find it, but are done for this
-  // block.
-  return 0;
-}
-
diff --git a/lib/Transforms/Utils/BreakCriticalEdges.cpp b/lib/Transforms/Utils/BreakCriticalEdges.cpp
index 8c25ad1..26f53c0 100644
--- a/lib/Transforms/Utils/BreakCriticalEdges.cpp
+++ b/lib/Transforms/Utils/BreakCriticalEdges.cpp
@@ -106,11 +106,12 @@ bool llvm::isCriticalEdge(const TerminatorInst *TI, unsigned SuccNum,
   // If AllowIdenticalEdges is true, then we allow this edge to be considered
   // non-critical iff all preds come from TI's block.
   while (I != E) {
-    if (*I != FirstPred)
+    const BasicBlock *P = *I;
+    if (P != FirstPred)
       return true;
     // Note: leave this as is until no one ever compiles with either gcc 4.0.1
     // or Xcode 2. This seems to work around the pred_iterator assert in PR 2207
-    E = pred_end(*I);
+    E = pred_end(P);
     ++I;
   }
   return false;
@@ -277,11 +278,13 @@ BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
         OtherPreds.push_back(PN->getIncomingBlock(i));
   } else {
     for (pred_iterator I = pred_begin(DestBB), E = pred_end(DestBB);
-         I != E; ++I)
-      if (*I != NewBB)
-        OtherPreds.push_back(*I);
+         I != E; ++I) {
+      BasicBlock *P = *I;
+      if (P != NewBB)
+          OtherPreds.push_back(P);
+    }
   }
-  
+
   bool NewBBDominatesDestBB = true;
   
   // Should we update DominatorTree information?
@@ -400,11 +403,13 @@ BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
           bool HasPredOutsideOfLoop = false;
           BasicBlock *Exit = ExitBlocks[i];
           for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit);
-               I != E; ++I)
-            if (TIL->contains(*I))
-              Preds.push_back(*I);
+               I != E; ++I) {
+            BasicBlock *P = *I;
+            if (TIL->contains(P))
+              Preds.push_back(P);
             else
               HasPredOutsideOfLoop = true;
+          }
           // If there are any preds not in the loop, we'll need to split
           // the edges. The Preds.empty() check is needed because a block
           // may appear multiple times in the list. We can't use
diff --git a/lib/Transforms/Utils/BuildLibCalls.cpp b/lib/Transforms/Utils/BuildLibCalls.cpp
index 767fa3a..7a9d007 100644
--- a/lib/Transforms/Utils/BuildLibCalls.cpp
+++ b/lib/Transforms/Utils/BuildLibCalls.cpp
@@ -69,6 +69,31 @@ Value *llvm::EmitStrChr(Value *Ptr, char C, IRBuilder<> &B,
   return CI;
 }
 
+/// EmitStrNCmp - Emit a call to the strncmp function to the builder.
+Value *llvm::EmitStrNCmp(Value *Ptr1, Value *Ptr2, Value *Len,
+                         IRBuilder<> &B, const TargetData *TD) {
+  Module *M = B.GetInsertBlock()->getParent()->getParent();
+  AttributeWithIndex AWI[3];
+  AWI[0] = AttributeWithIndex::get(1, Attribute::NoCapture);
+  AWI[1] = AttributeWithIndex::get(2, Attribute::NoCapture);
+  AWI[2] = AttributeWithIndex::get(~0u, Attribute::ReadOnly |
+                                   Attribute::NoUnwind);
+
+  LLVMContext &Context = B.GetInsertBlock()->getContext();
+  Value *StrNCmp = M->getOrInsertFunction("strncmp", AttrListPtr::get(AWI, 3),
+                                          B.getInt32Ty(),
+                                          B.getInt8PtrTy(),
+                                          B.getInt8PtrTy(),
+                                          TD->getIntPtrType(Context), NULL);
+  CallInst *CI = B.CreateCall3(StrNCmp, CastToCStr(Ptr1, B),
+                               CastToCStr(Ptr2, B), Len, "strncmp");
+
+  if (const Function *F = dyn_cast<Function>(StrNCmp->stripPointerCasts()))
+    CI->setCallingConv(F->getCallingConv());
+
+  return CI;
+}
+
 /// EmitStrCpy - Emit a call to the strcpy function to the builder, for the
 /// specified pointer arguments.
 Value *llvm::EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
@@ -112,10 +137,10 @@ Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,
 Value *llvm::EmitMemCpy(Value *Dst, Value *Src, Value *Len, unsigned Align,
                         bool isVolatile, IRBuilder<> &B, const TargetData *TD) {
   Module *M = B.GetInsertBlock()->getParent()->getParent();
-  const Type *ArgTys[3] = { Dst->getType(), Src->getType(), Len->getType() };
-  Value *MemCpy = Intrinsic::getDeclaration(M, Intrinsic::memcpy, ArgTys, 3);
   Dst = CastToCStr(Dst, B);
   Src = CastToCStr(Src, B);
+  const Type *ArgTys[3] = { Dst->getType(), Src->getType(), Len->getType() };
+  Value *MemCpy = Intrinsic::getDeclaration(M, Intrinsic::memcpy, ArgTys, 3);
   return B.CreateCall5(MemCpy, Dst, Src, Len,
                        ConstantInt::get(B.getInt32Ty(), Align),
                        ConstantInt::get(B.getInt1Ty(), isVolatile));
@@ -395,11 +420,11 @@ bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
         FT->getParamType(2) != TD->getIntPtrType(Context) ||
         FT->getParamType(3) != TD->getIntPtrType(Context))
       return false;
-    
-    if (isFoldable(4, 3, false)) {
-      EmitMemCpy(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+
+    if (isFoldable(3 + CallInst::ArgOffset, 2 + CallInst::ArgOffset, false)) {
+      EmitMemCpy(CI->getArgOperand(0), CI->getArgOperand(1), CI->getArgOperand(2),
                  1, false, B, TD);
-      replaceCall(CI->getOperand(1));
+      replaceCall(CI->getArgOperand(0));
       return true;
     }
     return false;
@@ -418,11 +443,11 @@ bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
         FT->getParamType(2) != TD->getIntPtrType(Context) ||
         FT->getParamType(3) != TD->getIntPtrType(Context))
       return false;
-    
-    if (isFoldable(4, 3, false)) {
-      EmitMemMove(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+
+    if (isFoldable(3 + CallInst::ArgOffset, 2 + CallInst::ArgOffset, false)) {
+      EmitMemMove(CI->getArgOperand(0), CI->getArgOperand(1), CI->getArgOperand(2),
                   1, false, B, TD);
-      replaceCall(CI->getOperand(1));
+      replaceCall(CI->getArgOperand(0));
       return true;
     }
     return false;
@@ -436,12 +461,12 @@ bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
         FT->getParamType(2) != TD->getIntPtrType(Context) ||
         FT->getParamType(3) != TD->getIntPtrType(Context))
       return false;
-    
-    if (isFoldable(4, 3, false)) {
-      Value *Val = B.CreateIntCast(CI->getOperand(2), B.getInt8Ty(),
+
+    if (isFoldable(3 + CallInst::ArgOffset, 2 + CallInst::ArgOffset, false)) {
+      Value *Val = B.CreateIntCast(CI->getArgOperand(1), B.getInt8Ty(),
                                    false);
-      EmitMemSet(CI->getOperand(1), Val,  CI->getOperand(3), false, B, TD);
-      replaceCall(CI->getOperand(1));
+      EmitMemSet(CI->getArgOperand(0), Val,  CI->getArgOperand(2), false, B, TD);
+      replaceCall(CI->getArgOperand(0));
       return true;
     }
     return false;
@@ -462,8 +487,8 @@ bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
     // st[rp]cpy_chk call which may fail at runtime if the size is too long.
     // TODO: It might be nice to get a maximum length out of the possible
     // string lengths for varying.
-    if (isFoldable(3, 2, true)) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
+    if (isFoldable(2 + CallInst::ArgOffset, 1 + CallInst::ArgOffset, true)) {
+      Value *Ret = EmitStrCpy(CI->getArgOperand(0), CI->getArgOperand(1), B, TD,
                               Name.substr(2, 6));
       replaceCall(Ret);
       return true;
@@ -479,10 +504,10 @@ bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
         !FT->getParamType(2)->isIntegerTy() ||
         FT->getParamType(3) != TD->getIntPtrType(Context))
       return false;
-    
-    if (isFoldable(4, 3, false)) {
-      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
-                               CI->getOperand(3), B, TD, Name.substr(2, 7));
+
+    if (isFoldable(3 + CallInst::ArgOffset, 2 + CallInst::ArgOffset, false)) {
+      Value *Ret = EmitStrNCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+                               CI->getArgOperand(2), B, TD, Name.substr(2, 7));
       replaceCall(Ret);
       return true;
     }
diff --git a/lib/Transforms/Utils/CloneFunction.cpp b/lib/Transforms/Utils/CloneFunction.cpp
index 6d4fe4b..1dcfd57 100644
--- a/lib/Transforms/Utils/CloneFunction.cpp
+++ b/lib/Transforms/Utils/CloneFunction.cpp
@@ -32,7 +32,7 @@ using namespace llvm;
 
 // CloneBasicBlock - See comments in Cloning.h
 BasicBlock *llvm::CloneBasicBlock(const BasicBlock *BB,
-                                  DenseMap<const Value*, Value*> &ValueMap,
+                                  ValueToValueMapTy &VMap,
                                   const Twine &NameSuffix, Function *F,
                                   ClonedCodeInfo *CodeInfo) {
   BasicBlock *NewBB = BasicBlock::Create(BB->getContext(), "", F);
@@ -47,7 +47,7 @@ BasicBlock *llvm::CloneBasicBlock(const BasicBlock *BB,
     if (II->hasName())
       NewInst->setName(II->getName()+NameSuffix);
     NewBB->getInstList().push_back(NewInst);
-    ValueMap[II] = NewInst;                // Add instruction map to value.
+    VMap[II] = NewInst;                // Add instruction map to value.
     
     hasCalls |= (isa<CallInst>(II) && !isa<DbgInfoIntrinsic>(II));
     if (const AllocaInst *AI = dyn_cast<AllocaInst>(II)) {
@@ -72,7 +72,7 @@ BasicBlock *llvm::CloneBasicBlock(const BasicBlock *BB,
 // ArgMap values.
 //
 void llvm::CloneFunctionInto(Function *NewFunc, const Function *OldFunc,
-                             DenseMap<const Value*, Value*> &ValueMap,
+                             ValueToValueMapTy &VMap,
                              SmallVectorImpl<ReturnInst*> &Returns,
                              const char *NameSuffix, ClonedCodeInfo *CodeInfo) {
   assert(NameSuffix && "NameSuffix cannot be null!");
@@ -80,17 +80,17 @@ void llvm::CloneFunctionInto(Function *NewFunc, const Function *OldFunc,
 #ifndef NDEBUG
   for (Function::const_arg_iterator I = OldFunc->arg_begin(), 
        E = OldFunc->arg_end(); I != E; ++I)
-    assert(ValueMap.count(I) && "No mapping from source argument specified!");
+    assert(VMap.count(I) && "No mapping from source argument specified!");
 #endif
 
   // Clone any attributes.
   if (NewFunc->arg_size() == OldFunc->arg_size())
     NewFunc->copyAttributesFrom(OldFunc);
   else {
-    //Some arguments were deleted with the ValueMap. Copy arguments one by one
+    //Some arguments were deleted with the VMap. Copy arguments one by one
     for (Function::const_arg_iterator I = OldFunc->arg_begin(), 
            E = OldFunc->arg_end(); I != E; ++I)
-      if (Argument* Anew = dyn_cast<Argument>(ValueMap[I]))
+      if (Argument* Anew = dyn_cast<Argument>(VMap[I]))
         Anew->addAttr( OldFunc->getAttributes()
                        .getParamAttributes(I->getArgNo() + 1));
     NewFunc->setAttributes(NewFunc->getAttributes()
@@ -111,43 +111,43 @@ void llvm::CloneFunctionInto(Function *NewFunc, const Function *OldFunc,
     const BasicBlock &BB = *BI;
 
     // Create a new basic block and copy instructions into it!
-    BasicBlock *CBB = CloneBasicBlock(&BB, ValueMap, NameSuffix, NewFunc,
+    BasicBlock *CBB = CloneBasicBlock(&BB, VMap, NameSuffix, NewFunc,
                                       CodeInfo);
-    ValueMap[&BB] = CBB;                       // Add basic block mapping.
+    VMap[&BB] = CBB;                       // Add basic block mapping.
 
     if (ReturnInst *RI = dyn_cast<ReturnInst>(CBB->getTerminator()))
       Returns.push_back(RI);
   }
 
   // Loop over all of the instructions in the function, fixing up operand
-  // references as we go.  This uses ValueMap to do all the hard work.
+  // references as we go.  This uses VMap to do all the hard work.
   //
-  for (Function::iterator BB = cast<BasicBlock>(ValueMap[OldFunc->begin()]),
+  for (Function::iterator BB = cast<BasicBlock>(VMap[OldFunc->begin()]),
          BE = NewFunc->end(); BB != BE; ++BB)
     // Loop over all instructions, fixing each one as we find it...
     for (BasicBlock::iterator II = BB->begin(); II != BB->end(); ++II)
-      RemapInstruction(II, ValueMap);
+      RemapInstruction(II, VMap);
 }
 
 /// CloneFunction - Return a copy of the specified function, but without
 /// embedding the function into another module.  Also, any references specified
-/// in the ValueMap are changed to refer to their mapped value instead of the
-/// original one.  If any of the arguments to the function are in the ValueMap,
-/// the arguments are deleted from the resultant function.  The ValueMap is
+/// in the VMap are changed to refer to their mapped value instead of the
+/// original one.  If any of the arguments to the function are in the VMap,
+/// the arguments are deleted from the resultant function.  The VMap is
 /// updated to include mappings from all of the instructions and basicblocks in
 /// the function from their old to new values.
 ///
 Function *llvm::CloneFunction(const Function *F,
-                              DenseMap<const Value*, Value*> &ValueMap,
+                              ValueToValueMapTy &VMap,
                               ClonedCodeInfo *CodeInfo) {
   std::vector<const Type*> ArgTypes;
 
   // The user might be deleting arguments to the function by specifying them in
-  // the ValueMap.  If so, we need to not add the arguments to the arg ty vector
+  // the VMap.  If so, we need to not add the arguments to the arg ty vector
   //
   for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end();
        I != E; ++I)
-    if (ValueMap.count(I) == 0)  // Haven't mapped the argument to anything yet?
+    if (VMap.count(I) == 0)  // Haven't mapped the argument to anything yet?
       ArgTypes.push_back(I->getType());
 
   // Create a new function type...
@@ -161,13 +161,13 @@ Function *llvm::CloneFunction(const Function *F,
   Function::arg_iterator DestI = NewF->arg_begin();
   for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end();
        I != E; ++I)
-    if (ValueMap.count(I) == 0) {   // Is this argument preserved?
+    if (VMap.count(I) == 0) {   // Is this argument preserved?
       DestI->setName(I->getName()); // Copy the name over...
-      ValueMap[I] = DestI++;        // Add mapping to ValueMap
+      VMap[I] = DestI++;        // Add mapping to VMap
     }
 
   SmallVector<ReturnInst*, 8> Returns;  // Ignore returns cloned.
-  CloneFunctionInto(NewF, F, ValueMap, Returns, "", CodeInfo);
+  CloneFunctionInto(NewF, F, VMap, Returns, "", CodeInfo);
   return NewF;
 }
 
@@ -179,19 +179,19 @@ namespace {
   struct PruningFunctionCloner {
     Function *NewFunc;
     const Function *OldFunc;
-    DenseMap<const Value*, Value*> &ValueMap;
+    ValueToValueMapTy &VMap;
     SmallVectorImpl<ReturnInst*> &Returns;
     const char *NameSuffix;
     ClonedCodeInfo *CodeInfo;
     const TargetData *TD;
   public:
     PruningFunctionCloner(Function *newFunc, const Function *oldFunc,
-                          DenseMap<const Value*, Value*> &valueMap,
+                          ValueToValueMapTy &valueMap,
                           SmallVectorImpl<ReturnInst*> &returns,
                           const char *nameSuffix, 
                           ClonedCodeInfo *codeInfo,
                           const TargetData *td)
-    : NewFunc(newFunc), OldFunc(oldFunc), ValueMap(valueMap), Returns(returns),
+    : NewFunc(newFunc), OldFunc(oldFunc), VMap(valueMap), Returns(returns),
       NameSuffix(nameSuffix), CodeInfo(codeInfo), TD(td) {
     }
 
@@ -202,7 +202,7 @@ namespace {
     
   public:
     /// ConstantFoldMappedInstruction - Constant fold the specified instruction,
-    /// mapping its operands through ValueMap if they are available.
+    /// mapping its operands through VMap if they are available.
     Constant *ConstantFoldMappedInstruction(const Instruction *I);
   };
 }
@@ -211,7 +211,7 @@ namespace {
 /// anything that it can reach.
 void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
                                        std::vector<const BasicBlock*> &ToClone){
-  Value *&BBEntry = ValueMap[BB];
+  Value *&BBEntry = VMap[BB];
 
   // Have we already cloned this block?
   if (BBEntry) return;
@@ -230,7 +230,7 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
     // If this instruction constant folds, don't bother cloning the instruction,
     // instead, just add the constant to the value map.
     if (Constant *C = ConstantFoldMappedInstruction(II)) {
-      ValueMap[II] = C;
+      VMap[II] = C;
       continue;
     }
 
@@ -238,7 +238,7 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
     if (II->hasName())
       NewInst->setName(II->getName()+NameSuffix);
     NewBB->getInstList().push_back(NewInst);
-    ValueMap[II] = NewInst;                // Add instruction map to value.
+    VMap[II] = NewInst;                // Add instruction map to value.
     
     hasCalls |= (isa<CallInst>(II) && !isa<DbgInfoIntrinsic>(II));
     if (const AllocaInst *AI = dyn_cast<AllocaInst>(II)) {
@@ -258,12 +258,12 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
       ConstantInt *Cond = dyn_cast<ConstantInt>(BI->getCondition());
       // Or is a known constant in the caller...
       if (Cond == 0)  
-        Cond = dyn_cast_or_null<ConstantInt>(ValueMap[BI->getCondition()]);
+        Cond = dyn_cast_or_null<ConstantInt>(VMap[BI->getCondition()]);
 
       // Constant fold to uncond branch!
       if (Cond) {
         BasicBlock *Dest = BI->getSuccessor(!Cond->getZExtValue());
-        ValueMap[OldTI] = BranchInst::Create(Dest, NewBB);
+        VMap[OldTI] = BranchInst::Create(Dest, NewBB);
         ToClone.push_back(Dest);
         TerminatorDone = true;
       }
@@ -272,10 +272,10 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
     // If switching on a value known constant in the caller.
     ConstantInt *Cond = dyn_cast<ConstantInt>(SI->getCondition());
     if (Cond == 0)  // Or known constant after constant prop in the callee...
-      Cond = dyn_cast_or_null<ConstantInt>(ValueMap[SI->getCondition()]);
+      Cond = dyn_cast_or_null<ConstantInt>(VMap[SI->getCondition()]);
     if (Cond) {     // Constant fold to uncond branch!
       BasicBlock *Dest = SI->getSuccessor(SI->findCaseValue(Cond));
-      ValueMap[OldTI] = BranchInst::Create(Dest, NewBB);
+      VMap[OldTI] = BranchInst::Create(Dest, NewBB);
       ToClone.push_back(Dest);
       TerminatorDone = true;
     }
@@ -286,7 +286,7 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
     if (OldTI->hasName())
       NewInst->setName(OldTI->getName()+NameSuffix);
     NewBB->getInstList().push_back(NewInst);
-    ValueMap[OldTI] = NewInst;             // Add instruction map to value.
+    VMap[OldTI] = NewInst;             // Add instruction map to value.
     
     // Recursively clone any reachable successor blocks.
     const TerminatorInst *TI = BB->getTerminator();
@@ -307,13 +307,13 @@ void PruningFunctionCloner::CloneBlock(const BasicBlock *BB,
 }
 
 /// ConstantFoldMappedInstruction - Constant fold the specified instruction,
-/// mapping its operands through ValueMap if they are available.
+/// mapping its operands through VMap if they are available.
 Constant *PruningFunctionCloner::
 ConstantFoldMappedInstruction(const Instruction *I) {
   SmallVector<Constant*, 8> Ops;
   for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
     if (Constant *Op = dyn_cast_or_null<Constant>(MapValue(I->getOperand(i),
-                                                           ValueMap)))
+                                                           VMap)))
       Ops.push_back(Op);
     else
       return 0;  // All operands not constant!
@@ -363,7 +363,7 @@ static MDNode *UpdateInlinedAtInfo(MDNode *InsnMD, MDNode *TheCallMD) {
 /// dead.  Since this doesn't produce an exact copy of the input, it can't be
 /// used for things like CloneFunction or CloneModule.
 void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
-                                     DenseMap<const Value*, Value*> &ValueMap,
+                                     ValueToValueMapTy &VMap,
                                      SmallVectorImpl<ReturnInst*> &Returns,
                                      const char *NameSuffix, 
                                      ClonedCodeInfo *CodeInfo,
@@ -374,10 +374,10 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
 #ifndef NDEBUG
   for (Function::const_arg_iterator II = OldFunc->arg_begin(), 
        E = OldFunc->arg_end(); II != E; ++II)
-    assert(ValueMap.count(II) && "No mapping from source argument specified!");
+    assert(VMap.count(II) && "No mapping from source argument specified!");
 #endif
 
-  PruningFunctionCloner PFC(NewFunc, OldFunc, ValueMap, Returns,
+  PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, Returns,
                             NameSuffix, CodeInfo, TD);
 
   // Clone the entry block, and anything recursively reachable from it.
@@ -397,14 +397,14 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
   SmallVector<const PHINode*, 16> PHIToResolve;
   for (Function::const_iterator BI = OldFunc->begin(), BE = OldFunc->end();
        BI != BE; ++BI) {
-    BasicBlock *NewBB = cast_or_null<BasicBlock>(ValueMap[BI]);
+    BasicBlock *NewBB = cast_or_null<BasicBlock>(VMap[BI]);
     if (NewBB == 0) continue;  // Dead block.
 
     // Add the new block to the new function.
     NewFunc->getBasicBlockList().push_back(NewBB);
     
     // Loop over all of the instructions in the block, fixing up operand
-    // references as we go.  This uses ValueMap to do all the hard work.
+    // references as we go.  This uses VMap to do all the hard work.
     //
     BasicBlock::iterator I = NewBB->begin();
 
@@ -455,7 +455,7 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
           I->setMetadata(DbgKind, 0);
         }
       }
-      RemapInstruction(I, ValueMap);
+      RemapInstruction(I, VMap);
     }
   }
   
@@ -465,19 +465,19 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
     const PHINode *OPN = PHIToResolve[phino];
     unsigned NumPreds = OPN->getNumIncomingValues();
     const BasicBlock *OldBB = OPN->getParent();
-    BasicBlock *NewBB = cast<BasicBlock>(ValueMap[OldBB]);
+    BasicBlock *NewBB = cast<BasicBlock>(VMap[OldBB]);
 
     // Map operands for blocks that are live and remove operands for blocks
     // that are dead.
     for (; phino != PHIToResolve.size() &&
          PHIToResolve[phino]->getParent() == OldBB; ++phino) {
       OPN = PHIToResolve[phino];
-      PHINode *PN = cast<PHINode>(ValueMap[OPN]);
+      PHINode *PN = cast<PHINode>(VMap[OPN]);
       for (unsigned pred = 0, e = NumPreds; pred != e; ++pred) {
         if (BasicBlock *MappedBlock = 
-            cast_or_null<BasicBlock>(ValueMap[PN->getIncomingBlock(pred)])) {
+            cast_or_null<BasicBlock>(VMap[PN->getIncomingBlock(pred)])) {
           Value *InVal = MapValue(PN->getIncomingValue(pred),
-                                  ValueMap);
+                                  VMap);
           assert(InVal && "Unknown input value?");
           PN->setIncomingValue(pred, InVal);
           PN->setIncomingBlock(pred, MappedBlock);
@@ -531,15 +531,15 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
       while ((PN = dyn_cast<PHINode>(I++))) {
         Value *NV = UndefValue::get(PN->getType());
         PN->replaceAllUsesWith(NV);
-        assert(ValueMap[OldI] == PN && "ValueMap mismatch");
-        ValueMap[OldI] = NV;
+        assert(VMap[OldI] == PN && "VMap mismatch");
+        VMap[OldI] = NV;
         PN->eraseFromParent();
         ++OldI;
       }
     }
     // NOTE: We cannot eliminate single entry phi nodes here, because of
-    // ValueMap.  Single entry phi nodes can have multiple ValueMap entries
-    // pointing at them.  Thus, deleting one would require scanning the ValueMap
+    // VMap.  Single entry phi nodes can have multiple VMap entries
+    // pointing at them.  Thus, deleting one would require scanning the VMap
     // to update any entries in it that would require that.  This would be
     // really slow.
   }
@@ -548,14 +548,14 @@ void llvm::CloneAndPruneFunctionInto(Function *NewFunc, const Function *OldFunc,
   // and zap unconditional fall-through branches.  This happen all the time when
   // specializing code: code specialization turns conditional branches into
   // uncond branches, and this code folds them.
-  Function::iterator I = cast<BasicBlock>(ValueMap[&OldFunc->getEntryBlock()]);
+  Function::iterator I = cast<BasicBlock>(VMap[&OldFunc->getEntryBlock()]);
   while (I != NewFunc->end()) {
     BranchInst *BI = dyn_cast<BranchInst>(I->getTerminator());
     if (!BI || BI->isConditional()) { ++I; continue; }
     
     // Note that we can't eliminate uncond branches if the destination has
     // single-entry PHI nodes.  Eliminating the single-entry phi nodes would
-    // require scanning the ValueMap to update any entries that point to the phi
+    // require scanning the VMap to update any entries that point to the phi
     // node.
     BasicBlock *Dest = BI->getSuccessor(0);
     if (!Dest->getSinglePredecessor() || isa<PHINode>(Dest->begin())) {
diff --git a/lib/Transforms/Utils/CloneLoop.cpp b/lib/Transforms/Utils/CloneLoop.cpp
index 38928dc..551b630 100644
--- a/lib/Transforms/Utils/CloneLoop.cpp
+++ b/lib/Transforms/Utils/CloneLoop.cpp
@@ -15,7 +15,6 @@
 #include "llvm/BasicBlock.h"
 #include "llvm/Analysis/LoopPass.h"
 #include "llvm/Analysis/Dominators.h"
-#include "llvm/ADT/DenseMap.h"
 
 
 using namespace llvm;
@@ -23,13 +22,13 @@ using namespace llvm;
 /// CloneDominatorInfo - Clone basicblock's dominator tree and, if available,
 /// dominance info. It is expected that basic block is already cloned.
 static void CloneDominatorInfo(BasicBlock *BB, 
-                               DenseMap<const Value *, Value *> &ValueMap,
+                               ValueMap<const Value *, Value *> &VMap,
                                DominatorTree *DT,
                                DominanceFrontier *DF) {
 
   assert (DT && "DominatorTree is not available");
-  DenseMap<const Value *, Value*>::iterator BI = ValueMap.find(BB);
-  assert (BI != ValueMap.end() && "BasicBlock clone is missing");
+  ValueMap<const Value *, Value*>::iterator BI = VMap.find(BB);
+  assert (BI != VMap.end() && "BasicBlock clone is missing");
   BasicBlock *NewBB = cast<BasicBlock>(BI->second);
 
   // NewBB already got dominator info.
@@ -43,11 +42,11 @@ static void CloneDominatorInfo(BasicBlock *BB,
 
   // NewBB's dominator is either BB's dominator or BB's dominator's clone.
   BasicBlock *NewBBDom = BBDom;
-  DenseMap<const Value *, Value*>::iterator BBDomI = ValueMap.find(BBDom);
-  if (BBDomI != ValueMap.end()) {
+  ValueMap<const Value *, Value*>::iterator BBDomI = VMap.find(BBDom);
+  if (BBDomI != VMap.end()) {
     NewBBDom = cast<BasicBlock>(BBDomI->second);
     if (!DT->getNode(NewBBDom))
-      CloneDominatorInfo(BBDom, ValueMap, DT, DF);
+      CloneDominatorInfo(BBDom, VMap, DT, DF);
   }
   DT->addNewBlock(NewBB, NewBBDom);
 
@@ -60,8 +59,8 @@ static void CloneDominatorInfo(BasicBlock *BB,
         for (DominanceFrontier::DomSetType::iterator I = S.begin(), E = S.end();
              I != E; ++I) {
           BasicBlock *DB = *I;
-          DenseMap<const Value*, Value*>::iterator IDM = ValueMap.find(DB);
-          if (IDM != ValueMap.end())
+          ValueMap<const Value*, Value*>::iterator IDM = VMap.find(DB);
+          if (IDM != VMap.end())
             NewDFSet.insert(cast<BasicBlock>(IDM->second));
           else
             NewDFSet.insert(DB);
@@ -71,10 +70,10 @@ static void CloneDominatorInfo(BasicBlock *BB,
   }
 }
 
-/// CloneLoop - Clone Loop. Clone dominator info. Populate ValueMap
+/// CloneLoop - Clone Loop. Clone dominator info. Populate VMap
 /// using old blocks to new blocks mapping.
 Loop *llvm::CloneLoop(Loop *OrigL, LPPassManager  *LPM, LoopInfo *LI,
-                      DenseMap<const Value *, Value *> &ValueMap, Pass *P) {
+                      ValueMap<const Value *, Value *> &VMap, Pass *P) {
   
   DominatorTree *DT = NULL;
   DominanceFrontier *DF = NULL;
@@ -104,8 +103,8 @@ Loop *llvm::CloneLoop(Loop *OrigL, LPPassManager  *LPM, LoopInfo *LI,
     for (Loop::block_iterator I = L->block_begin(), E = L->block_end();
          I != E; ++I) {
       BasicBlock *BB = *I;
-      BasicBlock *NewBB = CloneBasicBlock(BB, ValueMap, ".clone");
-      ValueMap[BB] = NewBB;
+      BasicBlock *NewBB = CloneBasicBlock(BB, VMap, ".clone");
+      VMap[BB] = NewBB;
       if (P)
         LPM->cloneBasicBlockSimpleAnalysis(BB, NewBB, L);
       NewLoop->addBasicBlockToLoop(NewBB, LI->getBase());
@@ -117,7 +116,7 @@ Loop *llvm::CloneLoop(Loop *OrigL, LPPassManager  *LPM, LoopInfo *LI,
       for (Loop::block_iterator I = L->block_begin(), E = L->block_end();
            I != E; ++I) {
         BasicBlock *BB = *I;
-        CloneDominatorInfo(BB, ValueMap, DT, DF);
+        CloneDominatorInfo(BB, VMap, DT, DF);
       }
 
     // Process sub loops
@@ -125,7 +124,7 @@ Loop *llvm::CloneLoop(Loop *OrigL, LPPassManager  *LPM, LoopInfo *LI,
       LoopNest.push_back(*I);
   } while (!LoopNest.empty());
 
-  // Remap instructions to reference operands from ValueMap.
+  // Remap instructions to reference operands from VMap.
   for(SmallVector<BasicBlock *, 16>::iterator NBItr = NewBlocks.begin(), 
         NBE = NewBlocks.end();  NBItr != NBE; ++NBItr) {
     BasicBlock *NB = *NBItr;
@@ -135,8 +134,8 @@ Loop *llvm::CloneLoop(Loop *OrigL, LPPassManager  *LPM, LoopInfo *LI,
       for (unsigned index = 0, num_ops = Insn->getNumOperands(); 
            index != num_ops; ++index) {
         Value *Op = Insn->getOperand(index);
-        DenseMap<const Value *, Value *>::iterator OpItr = ValueMap.find(Op);
-        if (OpItr != ValueMap.end())
+        ValueMap<const Value *, Value *>::iterator OpItr = VMap.find(Op);
+        if (OpItr != VMap.end())
           Insn->setOperand(index, OpItr->second);
       }
     }
diff --git a/lib/Transforms/Utils/CloneModule.cpp b/lib/Transforms/Utils/CloneModule.cpp
index b87c082..fc603d2 100644
--- a/lib/Transforms/Utils/CloneModule.cpp
+++ b/lib/Transforms/Utils/CloneModule.cpp
@@ -28,12 +28,12 @@ using namespace llvm;
 Module *llvm::CloneModule(const Module *M) {
   // Create the value map that maps things from the old module over to the new
   // module.
-  DenseMap<const Value*, Value*> ValueMap;
-  return CloneModule(M, ValueMap);
+  ValueToValueMapTy VMap;
+  return CloneModule(M, VMap);
 }
 
 Module *llvm::CloneModule(const Module *M,
-                          DenseMap<const Value*, Value*> &ValueMap) {
+                          ValueToValueMapTy &VMap) {
   // First off, we need to create the new module...
   Module *New = new Module(M->getModuleIdentifier(), M->getContext());
   New->setDataLayout(M->getDataLayout());
@@ -51,7 +51,7 @@ Module *llvm::CloneModule(const Module *M,
     New->addLibrary(*I);
 
   // Loop over all of the global variables, making corresponding globals in the
-  // new module.  Here we add them to the ValueMap and to the new Module.  We
+  // new module.  Here we add them to the VMap and to the new Module.  We
   // don't worry about attributes or initializers, they will come later.
   //
   for (Module::const_global_iterator I = M->global_begin(), E = M->global_end();
@@ -62,7 +62,7 @@ Module *llvm::CloneModule(const Module *M,
                                             GlobalValue::ExternalLinkage, 0,
                                             I->getName());
     GV->setAlignment(I->getAlignment());
-    ValueMap[I] = GV;
+    VMap[I] = GV;
   }
 
   // Loop over the functions in the module, making external functions as before
@@ -71,13 +71,13 @@ Module *llvm::CloneModule(const Module *M,
       Function::Create(cast<FunctionType>(I->getType()->getElementType()),
                        GlobalValue::ExternalLinkage, I->getName(), New);
     NF->copyAttributesFrom(I);
-    ValueMap[I] = NF;
+    VMap[I] = NF;
   }
 
   // Loop over the aliases in the module
   for (Module::const_alias_iterator I = M->alias_begin(), E = M->alias_end();
        I != E; ++I)
-    ValueMap[I] = new GlobalAlias(I->getType(), GlobalAlias::ExternalLinkage,
+    VMap[I] = new GlobalAlias(I->getType(), GlobalAlias::ExternalLinkage,
                                   I->getName(), NULL, New);
   
   // Now that all of the things that global variable initializer can refer to
@@ -86,10 +86,10 @@ Module *llvm::CloneModule(const Module *M,
   //
   for (Module::const_global_iterator I = M->global_begin(), E = M->global_end();
        I != E; ++I) {
-    GlobalVariable *GV = cast<GlobalVariable>(ValueMap[I]);
+    GlobalVariable *GV = cast<GlobalVariable>(VMap[I]);
     if (I->hasInitializer())
       GV->setInitializer(cast<Constant>(MapValue(I->getInitializer(),
-                                                 ValueMap)));
+                                                 VMap)));
     GV->setLinkage(I->getLinkage());
     GV->setThreadLocal(I->isThreadLocal());
     GV->setConstant(I->isConstant());
@@ -98,17 +98,17 @@ Module *llvm::CloneModule(const Module *M,
   // Similarly, copy over function bodies now...
   //
   for (Module::const_iterator I = M->begin(), E = M->end(); I != E; ++I) {
-    Function *F = cast<Function>(ValueMap[I]);
+    Function *F = cast<Function>(VMap[I]);
     if (!I->isDeclaration()) {
       Function::arg_iterator DestI = F->arg_begin();
       for (Function::const_arg_iterator J = I->arg_begin(); J != I->arg_end();
            ++J) {
         DestI->setName(J->getName());
-        ValueMap[J] = DestI++;
+        VMap[J] = DestI++;
       }
 
       SmallVector<ReturnInst*, 8> Returns;  // Ignore returns cloned.
-      CloneFunctionInto(F, I, ValueMap, Returns);
+      CloneFunctionInto(F, I, VMap, Returns);
     }
 
     F->setLinkage(I->getLinkage());
@@ -117,11 +117,37 @@ Module *llvm::CloneModule(const Module *M,
   // And aliases
   for (Module::const_alias_iterator I = M->alias_begin(), E = M->alias_end();
        I != E; ++I) {
-    GlobalAlias *GA = cast<GlobalAlias>(ValueMap[I]);
+    GlobalAlias *GA = cast<GlobalAlias>(VMap[I]);
     GA->setLinkage(I->getLinkage());
     if (const Constant* C = I->getAliasee())
-      GA->setAliasee(cast<Constant>(MapValue(C, ValueMap)));
+      GA->setAliasee(cast<Constant>(MapValue(C, VMap)));
   }
-  
+
+  // And named metadata....
+  for (Module::const_named_metadata_iterator I = M->named_metadata_begin(),
+         E = M->named_metadata_end(); I != E; ++I) {
+    const NamedMDNode &NMD = *I;
+    SmallVector<MDNode*, 4> MDs;
+    for (unsigned i = 0, e = NMD.getNumOperands(); i != e; ++i)
+      MDs.push_back(cast<MDNode>(MapValue(NMD.getOperand(i), VMap)));
+    NamedMDNode::Create(New->getContext(), NMD.getName(),
+                        MDs.data(), MDs.size(), New);
+  }
+
+  // Update metadata attach with instructions.
+  for (Module::iterator MI = New->begin(), ME = New->end(); MI != ME; ++MI)   
+    for (Function::iterator FI = MI->begin(), FE = MI->end(); 
+         FI != FE; ++FI)
+      for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); 
+           BI != BE; ++BI) {
+        SmallVector<std::pair<unsigned, MDNode *>, 4 > MDs;
+        BI->getAllMetadata(MDs);
+        for (SmallVector<std::pair<unsigned, MDNode *>, 4>::iterator 
+               MDI = MDs.begin(), MDE = MDs.end(); MDI != MDE; ++MDI) {
+          Value *MappedValue = MapValue(MDI->second, VMap);
+          if (MDI->second != MappedValue && MappedValue)
+            BI->setMetadata(MDI->first, cast<MDNode>(MappedValue));
+        }
+      }
   return New;
 }
diff --git a/lib/Transforms/Utils/DemoteRegToStack.cpp b/lib/Transforms/Utils/DemoteRegToStack.cpp
index c908b4a..8e82a02 100644
--- a/lib/Transforms/Utils/DemoteRegToStack.cpp
+++ b/lib/Transforms/Utils/DemoteRegToStack.cpp
@@ -35,7 +35,7 @@ AllocaInst* llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
     I.eraseFromParent();
     return 0;
   }
-  
+
   // Create a stack slot to hold the value.
   AllocaInst *Slot;
   if (AllocaPoint) {
@@ -46,7 +46,7 @@ AllocaInst* llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
     Slot = new AllocaInst(I.getType(), 0, I.getName()+".reg2mem",
                           F->getEntryBlock().begin());
   }
-  
+
   // Change all of the users of the instruction to read from the stack slot
   // instead.
   while (!I.use_empty()) {
@@ -67,7 +67,7 @@ AllocaInst* llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
           Value *&V = Loads[PN->getIncomingBlock(i)];
           if (V == 0) {
             // Insert the load into the predecessor block
-            V = new LoadInst(Slot, I.getName()+".reload", VolatileLoads, 
+            V = new LoadInst(Slot, I.getName()+".reload", VolatileLoads,
                              PN->getIncomingBlock(i)->getTerminator());
           }
           PN->setIncomingValue(i, V);
@@ -110,8 +110,8 @@ AllocaInst* llvm::DemoteRegToStack(Instruction &I, bool VolatileLoads,
 /// The phi node is deleted and it returns the pointer to the alloca inserted.
 AllocaInst* llvm::DemotePHIToStack(PHINode *P, Instruction *AllocaPoint) {
   if (P->use_empty()) {
-    P->eraseFromParent();    
-    return 0;                
+    P->eraseFromParent();
+    return 0;
   }
 
   // Create a stack slot to hold the value.
@@ -124,23 +124,23 @@ AllocaInst* llvm::DemotePHIToStack(PHINode *P, Instruction *AllocaPoint) {
     Slot = new AllocaInst(P->getType(), 0, P->getName()+".reg2mem",
                           F->getEntryBlock().begin());
   }
-  
+
   // Iterate over each operand, insert store in each predecessor.
   for (unsigned i = 0, e = P->getNumIncomingValues(); i < e; ++i) {
     if (InvokeInst *II = dyn_cast<InvokeInst>(P->getIncomingValue(i))) {
-      assert(II->getParent() != P->getIncomingBlock(i) && 
+      assert(II->getParent() != P->getIncomingBlock(i) &&
              "Invoke edge not supported yet"); II=II;
     }
-    new StoreInst(P->getIncomingValue(i), Slot, 
+    new StoreInst(P->getIncomingValue(i), Slot,
                   P->getIncomingBlock(i)->getTerminator());
   }
-  
+
   // Insert load in place of the phi and replace all uses.
   Value *V = new LoadInst(Slot, P->getName()+".reload", P);
   P->replaceAllUsesWith(V);
-  
+
   // Delete phi.
   P->eraseFromParent();
-  
+
   return Slot;
 }
diff --git a/lib/Transforms/Utils/InlineFunction.cpp b/lib/Transforms/Utils/InlineFunction.cpp
index 91390bc..598e7d2 100644
--- a/lib/Transforms/Utils/InlineFunction.cpp
+++ b/lib/Transforms/Utils/InlineFunction.cpp
@@ -63,7 +63,8 @@ static void HandleCallsInBlockInlinedThroughInvoke(BasicBlock *BB,
     
     // Next, create the new invoke instruction, inserting it at the end
     // of the old basic block.
-    SmallVector<Value*, 8> InvokeArgs(CI->op_begin()+1, CI->op_end());
+    ImmutableCallSite CS(CI);
+    SmallVector<Value*, 8> InvokeArgs(CS.arg_begin(), CS.arg_end());
     InvokeInst *II =
       InvokeInst::Create(CI->getCalledValue(), Split, InvokeDest,
                          InvokeArgs.begin(), InvokeArgs.end(),
@@ -169,7 +170,7 @@ static void HandleInlinedInvoke(InvokeInst *II, BasicBlock *FirstNewBlock,
 /// some edges of the callgraph may remain.
 static void UpdateCallGraphAfterInlining(CallSite CS,
                                          Function::iterator FirstNewBlock,
-                                       DenseMap<const Value*, Value*> &ValueMap,
+                                         ValueMap<const Value*, Value*> &VMap,
                                          InlineFunctionInfo &IFI) {
   CallGraph &CG = *IFI.CG;
   const Function *Caller = CS.getInstruction()->getParent()->getParent();
@@ -192,9 +193,9 @@ static void UpdateCallGraphAfterInlining(CallSite CS,
   for (; I != E; ++I) {
     const Value *OrigCall = I->first;
 
-    DenseMap<const Value*, Value*>::iterator VMI = ValueMap.find(OrigCall);
+    ValueMap<const Value*, Value*>::iterator VMI = VMap.find(OrigCall);
     // Only copy the edge if the call was inlined!
-    if (VMI == ValueMap.end() || VMI->second == 0)
+    if (VMI == VMap.end() || VMI->second == 0)
       continue;
     
     // If the call was inlined, but then constant folded, there is no edge to
@@ -285,8 +286,8 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI) {
   ClonedCodeInfo InlinedFunctionInfo;
   Function::iterator FirstNewBlock;
 
-  { // Scope to destroy ValueMap after cloning.
-    DenseMap<const Value*, Value*> ValueMap;
+  { // Scope to destroy VMap after cloning.
+    ValueMap<const Value*, Value*> VMap;
 
     assert(CalledFunc->arg_size() == CS.arg_size() &&
            "No varargs calls can be inlined!");
@@ -351,16 +352,20 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI) {
         // Uses of the argument in the function should use our new alloca
         // instead.
         ActualArg = NewAlloca;
+
+        // Calls that we inline may use the new alloca, so we need to clear
+        // their 'tail' flags.
+        MustClearTailCallFlags = true;
       }
 
-      ValueMap[I] = ActualArg;
+      VMap[I] = ActualArg;
     }
 
     // We want the inliner to prune the code as it copies.  We would LOVE to
     // have no dead or constant instructions leftover after inlining occurs
     // (which can happen, e.g., because an argument was constant), but we'll be
     // happy with whatever the cloner can do.
-    CloneAndPruneFunctionInto(Caller, CalledFunc, ValueMap, Returns, ".i",
+    CloneAndPruneFunctionInto(Caller, CalledFunc, VMap, Returns, ".i",
                               &InlinedFunctionInfo, IFI.TD, TheCall);
 
     // Remember the first block that is newly cloned over.
@@ -368,7 +373,7 @@ bool llvm::InlineFunction(CallSite CS, InlineFunctionInfo &IFI) {
 
     // Update the callgraph if requested.
     if (IFI.CG)
-      UpdateCallGraphAfterInlining(CS, FirstNewBlock, ValueMap, IFI);
+      UpdateCallGraphAfterInlining(CS, FirstNewBlock, VMap, IFI);
   }
 
   // If there are any alloca instructions in the block that used to be the entry
diff --git a/lib/Transforms/Utils/LCSSA.cpp b/lib/Transforms/Utils/LCSSA.cpp
index df6e603..e90c30b 100644
--- a/lib/Transforms/Utils/LCSSA.cpp
+++ b/lib/Transforms/Utils/LCSSA.cpp
@@ -190,14 +190,15 @@ bool LCSSA::ProcessInstruction(Instruction *Inst,
   
   for (Value::use_iterator UI = Inst->use_begin(), E = Inst->use_end();
        UI != E; ++UI) {
-    BasicBlock *UserBB = cast<Instruction>(*UI)->getParent();
-    if (PHINode *PN = dyn_cast<PHINode>(*UI))
+    User *U = *UI;
+    BasicBlock *UserBB = cast<Instruction>(U)->getParent();
+    if (PHINode *PN = dyn_cast<PHINode>(U))
       UserBB = PN->getIncomingBlock(UI);
     
     if (InstBB != UserBB && !inLoop(UserBB))
       UsesToRewrite.push_back(&UI.getUse());
   }
-  
+
   // If there are no uses outside the loop, exit with no change.
   if (UsesToRewrite.empty()) return false;
   
diff --git a/lib/Transforms/Utils/Local.cpp b/lib/Transforms/Utils/Local.cpp
index d03f7a6..0b48a8f 100644
--- a/lib/Transforms/Utils/Local.cpp
+++ b/lib/Transforms/Utils/Local.cpp
@@ -35,111 +35,6 @@
 using namespace llvm;
 
 //===----------------------------------------------------------------------===//
-//  Local analysis.
-//
-
-/// getUnderlyingObjectWithOffset - Strip off up to MaxLookup GEPs and
-/// bitcasts to get back to the underlying object being addressed, keeping
-/// track of the offset in bytes from the GEPs relative to the result.
-/// This is closely related to Value::getUnderlyingObject but is located
-/// here to avoid making VMCore depend on TargetData.
-static Value *getUnderlyingObjectWithOffset(Value *V, const TargetData *TD,
-                                            uint64_t &ByteOffset,
-                                            unsigned MaxLookup = 6) {
-  if (!V->getType()->isPointerTy())
-    return V;
-  for (unsigned Count = 0; MaxLookup == 0 || Count < MaxLookup; ++Count) {
-    if (GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
-      if (!GEP->hasAllConstantIndices())
-        return V;
-      SmallVector<Value*, 8> Indices(GEP->op_begin() + 1, GEP->op_end());
-      ByteOffset += TD->getIndexedOffset(GEP->getPointerOperandType(),
-                                         &Indices[0], Indices.size());
-      V = GEP->getPointerOperand();
-    } else if (Operator::getOpcode(V) == Instruction::BitCast) {
-      V = cast<Operator>(V)->getOperand(0);
-    } else if (GlobalAlias *GA = dyn_cast<GlobalAlias>(V)) {
-      if (GA->mayBeOverridden())
-        return V;
-      V = GA->getAliasee();
-    } else {
-      return V;
-    }
-    assert(V->getType()->isPointerTy() && "Unexpected operand type!");
-  }
-  return V;
-}
-
-/// isSafeToLoadUnconditionally - Return true if we know that executing a load
-/// from this value cannot trap.  If it is not obviously safe to load from the
-/// specified pointer, we do a quick local scan of the basic block containing
-/// ScanFrom, to determine if the address is already accessed.
-bool llvm::isSafeToLoadUnconditionally(Value *V, Instruction *ScanFrom,
-                                       unsigned Align, const TargetData *TD) {
-  uint64_t ByteOffset = 0;
-  Value *Base = V;
-  if (TD)
-    Base = getUnderlyingObjectWithOffset(V, TD, ByteOffset);
-
-  const Type *BaseType = 0;
-  unsigned BaseAlign = 0;
-  if (const AllocaInst *AI = dyn_cast<AllocaInst>(Base)) {
-    // An alloca is safe to load from as load as it is suitably aligned.
-    BaseType = AI->getAllocatedType();
-    BaseAlign = AI->getAlignment();
-  } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(Base)) {
-    // Global variables are safe to load from but their size cannot be
-    // guaranteed if they are overridden.
-    if (!isa<GlobalAlias>(GV) && !GV->mayBeOverridden()) {
-      BaseType = GV->getType()->getElementType();
-      BaseAlign = GV->getAlignment();
-    }
-  }
-
-  if (BaseType && BaseType->isSized()) {
-    if (TD && BaseAlign == 0)
-      BaseAlign = TD->getPrefTypeAlignment(BaseType);
-
-    if (Align <= BaseAlign) {
-      if (!TD)
-        return true; // Loading directly from an alloca or global is OK.
-
-      // Check if the load is within the bounds of the underlying object.
-      const PointerType *AddrTy = cast<PointerType>(V->getType());
-      uint64_t LoadSize = TD->getTypeStoreSize(AddrTy->getElementType());
-      if (ByteOffset + LoadSize <= TD->getTypeAllocSize(BaseType) &&
-          (Align == 0 || (ByteOffset % Align) == 0))
-        return true;
-    }
-  }
-
-  // Otherwise, be a little bit aggressive by scanning the local block where we
-  // want to check to see if the pointer is already being loaded or stored
-  // from/to.  If so, the previous load or store would have already trapped,
-  // so there is no harm doing an extra load (also, CSE will later eliminate
-  // the load entirely).
-  BasicBlock::iterator BBI = ScanFrom, E = ScanFrom->getParent()->begin();
-
-  while (BBI != E) {
-    --BBI;
-
-    // If we see a free or a call which may write to memory (i.e. which might do
-    // a free) the pointer could be marked invalid.
-    if (isa<CallInst>(BBI) && BBI->mayWriteToMemory() &&
-        !isa<DbgInfoIntrinsic>(BBI))
-      return false;
-
-    if (LoadInst *LI = dyn_cast<LoadInst>(BBI)) {
-      if (LI->getOperand(0) == V) return true;
-    } else if (StoreInst *SI = dyn_cast<StoreInst>(BBI)) {
-      if (SI->getOperand(1) == V) return true;
-    }
-  }
-  return false;
-}
-
-
-//===----------------------------------------------------------------------===//
 //  Local constant propagation.
 //
 
@@ -537,9 +432,11 @@ static bool CanPropagatePredecessorsForPHIs(BasicBlock *BB, BasicBlock *Succ) {
   // Use that list to make another list of common predecessors of BB and Succ
   BlockSet CommonPreds;
   for (pred_iterator PI = pred_begin(Succ), PE = pred_end(Succ);
-        PI != PE; ++PI)
-    if (BBPreds.count(*PI))
-      CommonPreds.insert(*PI);
+       PI != PE; ++PI) {
+    BasicBlock *P = *PI;
+    if (BBPreds.count(P))
+      CommonPreds.insert(P);
+  }
 
   // Shortcut, if there are no common predecessors, merging is always safe
   if (CommonPreds.empty())
diff --git a/lib/Transforms/Utils/LoopSimplify.cpp b/lib/Transforms/Utils/LoopSimplify.cpp
index 1ef3c32..4f4edf3 100644
--- a/lib/Transforms/Utils/LoopSimplify.cpp
+++ b/lib/Transforms/Utils/LoopSimplify.cpp
@@ -142,9 +142,11 @@ ReprocessLoop:
     if (*BB == L->getHeader()) continue;
 
     SmallPtrSet<BasicBlock *, 4> BadPreds;
-    for (pred_iterator PI = pred_begin(*BB), PE = pred_end(*BB); PI != PE; ++PI)
-      if (!L->contains(*PI))
-        BadPreds.insert(*PI);
+    for (pred_iterator PI = pred_begin(*BB), PE = pred_end(*BB); PI != PE; ++PI){
+      BasicBlock *P = *PI;
+      if (!L->contains(P))
+        BadPreds.insert(P);
+    }
 
     // Delete each unique out-of-loop (and thus dead) predecessor.
     for (SmallPtrSet<BasicBlock *, 4>::iterator I = BadPreds.begin(),
@@ -192,7 +194,7 @@ ReprocessLoop:
   if (!Preheader) {
     Preheader = InsertPreheaderForLoop(L);
     if (Preheader) {
-      NumInserted++;
+      ++NumInserted;
       Changed = true;
     }
   }
@@ -215,7 +217,7 @@ ReprocessLoop:
       // allowed.
       if (!L->contains(*PI)) {
         if (RewriteLoopExitBlock(L, ExitBlock)) {
-          NumInserted++;
+          ++NumInserted;
           Changed = true;
         }
         break;
@@ -244,7 +246,7 @@ ReprocessLoop:
     // loop header.
     LoopLatch = InsertUniqueBackedgeBlock(L, Preheader);
     if (LoopLatch) {
-      NumInserted++;
+      ++NumInserted;
       Changed = true;
     }
   }
@@ -353,16 +355,18 @@ BasicBlock *LoopSimplify::InsertPreheaderForLoop(Loop *L) {
   // Compute the set of predecessors of the loop that are not in the loop.
   SmallVector<BasicBlock*, 8> OutsideBlocks;
   for (pred_iterator PI = pred_begin(Header), PE = pred_end(Header);
-       PI != PE; ++PI)
-    if (!L->contains(*PI)) {         // Coming in from outside the loop?
+       PI != PE; ++PI) {
+    BasicBlock *P = *PI;
+    if (!L->contains(P)) {         // Coming in from outside the loop?
       // If the loop is branched to from an indirect branch, we won't
       // be able to fully transform the loop, because it prohibits
       // edge splitting.
-      if (isa<IndirectBrInst>((*PI)->getTerminator())) return 0;
+      if (isa<IndirectBrInst>(P->getTerminator())) return 0;
 
       // Keep track of it.
-      OutsideBlocks.push_back(*PI);
+      OutsideBlocks.push_back(P);
     }
+  }
 
   // Split out the loop pre-header.
   BasicBlock *NewBB =
@@ -385,13 +389,15 @@ BasicBlock *LoopSimplify::InsertPreheaderForLoop(Loop *L) {
 /// outside of the loop.
 BasicBlock *LoopSimplify::RewriteLoopExitBlock(Loop *L, BasicBlock *Exit) {
   SmallVector<BasicBlock*, 8> LoopBlocks;
-  for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit); I != E; ++I)
-    if (L->contains(*I)) {
+  for (pred_iterator I = pred_begin(Exit), E = pred_end(Exit); I != E; ++I) {
+    BasicBlock *P = *I;
+    if (L->contains(P)) {
       // Don't do this if the loop is exited via an indirect branch.
-      if (isa<IndirectBrInst>((*I)->getTerminator())) return 0;
+      if (isa<IndirectBrInst>(P->getTerminator())) return 0;
 
-      LoopBlocks.push_back(*I);
+      LoopBlocks.push_back(P);
     }
+  }
 
   assert(!LoopBlocks.empty() && "No edges coming in from outside the loop?");
   BasicBlock *NewBB = SplitBlockPredecessors(Exit, &LoopBlocks[0], 
@@ -559,10 +565,11 @@ Loop *LoopSimplify::SeparateNestedLoop(Loop *L, LPPassManager &LPM) {
   // Determine which blocks should stay in L and which should be moved out to
   // the Outer loop now.
   std::set<BasicBlock*> BlocksInL;
-  for (pred_iterator PI = pred_begin(Header), E = pred_end(Header); PI!=E; ++PI)
-    if (DT->dominates(Header, *PI))
-      AddBlockAndPredsToSet(*PI, Header, BlocksInL);
-
+  for (pred_iterator PI=pred_begin(Header), E = pred_end(Header); PI!=E; ++PI) {
+    BasicBlock *P = *PI;
+    if (DT->dominates(Header, P))
+      AddBlockAndPredsToSet(P, Header, BlocksInL);
+  }
 
   // Scan all of the loop children of L, moving them to OuterLoop if they are
   // not part of the inner loop.
@@ -610,8 +617,10 @@ LoopSimplify::InsertUniqueBackedgeBlock(Loop *L, BasicBlock *Preheader) {
 
   // Figure out which basic blocks contain back-edges to the loop header.
   std::vector<BasicBlock*> BackedgeBlocks;
-  for (pred_iterator I = pred_begin(Header), E = pred_end(Header); I != E; ++I)
-    if (*I != Preheader) BackedgeBlocks.push_back(*I);
+  for (pred_iterator I = pred_begin(Header), E = pred_end(Header); I != E; ++I){
+    BasicBlock *P = *I;
+    if (P != Preheader) BackedgeBlocks.push_back(P);
+  }
 
   // Create and insert the new backedge block...
   BasicBlock *BEBlock = BasicBlock::Create(Header->getContext(),
diff --git a/lib/Transforms/Utils/LoopUnroll.cpp b/lib/Transforms/Utils/LoopUnroll.cpp
index 84fd1eb..e0e07e7 100644
--- a/lib/Transforms/Utils/LoopUnroll.cpp
+++ b/lib/Transforms/Utils/LoopUnroll.cpp
@@ -37,13 +37,13 @@ STATISTIC(NumCompletelyUnrolled, "Number of loops completely unrolled");
 STATISTIC(NumUnrolled,    "Number of loops unrolled (completely or otherwise)");
 
 /// RemapInstruction - Convert the instruction operands from referencing the
-/// current values into those specified by ValueMap.
+/// current values into those specified by VMap.
 static inline void RemapInstruction(Instruction *I,
-                                    DenseMap<const Value *, Value*> &ValueMap) {
+                                    ValueMap<const Value *, Value*> &VMap) {
   for (unsigned op = 0, E = I->getNumOperands(); op != E; ++op) {
     Value *Op = I->getOperand(op);
-    DenseMap<const Value *, Value*>::iterator It = ValueMap.find(Op);
-    if (It != ValueMap.end())
+    ValueMap<const Value *, Value*>::iterator It = VMap.find(Op);
+    if (It != VMap.end())
       I->setOperand(op, It->second);
   }
 }
@@ -183,7 +183,7 @@ bool llvm::UnrollLoop(Loop *L, unsigned Count, LoopInfo* LI, LPPassManager* LPM)
 
   // For the first iteration of the loop, we should use the precloned values for
   // PHI nodes.  Insert associations now.
-  typedef DenseMap<const Value*, Value*> ValueToValueMapTy;
+  typedef ValueMap<const Value*, Value*> ValueToValueMapTy;
   ValueToValueMapTy LastValueMap;
   std::vector<PHINode*> OrigPHINode;
   for (BasicBlock::iterator I = Header->begin(); isa<PHINode>(I); ++I) {
@@ -205,26 +205,26 @@ bool llvm::UnrollLoop(Loop *L, unsigned Count, LoopInfo* LI, LPPassManager* LPM)
     
     for (std::vector<BasicBlock*>::iterator BB = LoopBlocks.begin(),
          E = LoopBlocks.end(); BB != E; ++BB) {
-      ValueToValueMapTy ValueMap;
-      BasicBlock *New = CloneBasicBlock(*BB, ValueMap, "." + Twine(It));
+      ValueToValueMapTy VMap;
+      BasicBlock *New = CloneBasicBlock(*BB, VMap, "." + Twine(It));
       Header->getParent()->getBasicBlockList().push_back(New);
 
       // Loop over all of the PHI nodes in the block, changing them to use the
       // incoming values from the previous block.
       if (*BB == Header)
         for (unsigned i = 0, e = OrigPHINode.size(); i != e; ++i) {
-          PHINode *NewPHI = cast<PHINode>(ValueMap[OrigPHINode[i]]);
+          PHINode *NewPHI = cast<PHINode>(VMap[OrigPHINode[i]]);
           Value *InVal = NewPHI->getIncomingValueForBlock(LatchBlock);
           if (Instruction *InValI = dyn_cast<Instruction>(InVal))
             if (It > 1 && L->contains(InValI))
               InVal = LastValueMap[InValI];
-          ValueMap[OrigPHINode[i]] = InVal;
+          VMap[OrigPHINode[i]] = InVal;
           New->getInstList().erase(NewPHI);
         }
 
       // Update our running map of newest clones
       LastValueMap[*BB] = New;
-      for (ValueToValueMapTy::iterator VI = ValueMap.begin(), VE = ValueMap.end();
+      for (ValueToValueMapTy::iterator VI = VMap.begin(), VE = VMap.end();
            VI != VE; ++VI)
         LastValueMap[VI->first] = VI->second;
 
diff --git a/lib/Transforms/Utils/LowerInvoke.cpp b/lib/Transforms/Utils/LowerInvoke.cpp
index 0ed8c72..2696e69 100644
--- a/lib/Transforms/Utils/LowerInvoke.cpp
+++ b/lib/Transforms/Utils/LowerInvoke.cpp
@@ -45,6 +45,7 @@
 #include "llvm/Pass.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
+#include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Target/TargetLowering.h"
@@ -62,10 +63,7 @@ static cl::opt<bool> ExpensiveEHSupport("enable-correct-eh-support",
 namespace {
   class LowerInvoke : public FunctionPass {
     // Used for both models.
-    Constant *WriteFn;
     Constant *AbortFn;
-    Value *AbortMessage;
-    unsigned AbortMessageLength;
 
     // Used for expensive EH support.
     const Type *JBLinkTy;
@@ -92,10 +90,8 @@ namespace {
     }
 
   private:
-    void createAbortMessage(Module *M);
-    void writeAbortMessage(Instruction *IB);
     bool insertCheapEHSupport(Function &F);
-    void splitLiveRangesLiveAcrossInvokes(std::vector<InvokeInst*> &Invokes);
+    void splitLiveRangesLiveAcrossInvokes(SmallVectorImpl<InvokeInst*>&Invokes);
     void rewriteExpensiveInvoke(InvokeInst *II, unsigned InvokeNo,
                                 AllocaInst *InvokeNum, AllocaInst *StackPtr,
                                 SwitchInst *CatchSwitch);
@@ -123,7 +119,6 @@ FunctionPass *llvm::createLowerInvokePass(const TargetLowering *TLI,
 bool LowerInvoke::doInitialization(Module &M) {
   const Type *VoidPtrTy =
           Type::getInt8PtrTy(M.getContext());
-  AbortMessage = 0;
   if (useExpensiveEHSupport) {
     // Insert a type for the linked list of jump buffers.
     unsigned JBSize = TLI ? TLI->getJumpBufSize() : 0;
@@ -175,68 +170,14 @@ bool LowerInvoke::doInitialization(Module &M) {
   // We need the 'write' and 'abort' functions for both models.
   AbortFn = M.getOrInsertFunction("abort", Type::getVoidTy(M.getContext()),
                                   (Type *)0);
-#if 0 // "write" is Unix-specific.. code is going away soon anyway.
-  WriteFn = M.getOrInsertFunction("write", Type::VoidTy, Type::Int32Ty,
-                                  VoidPtrTy, Type::Int32Ty, (Type *)0);
-#else
-  WriteFn = 0;
-#endif
   return true;
 }
 
-void LowerInvoke::createAbortMessage(Module *M) {
-  if (useExpensiveEHSupport) {
-    // The abort message for expensive EH support tells the user that the
-    // program 'unwound' without an 'invoke' instruction.
-    Constant *Msg =
-      ConstantArray::get(M->getContext(),
-                         "ERROR: Exception thrown, but not caught!\n");
-    AbortMessageLength = Msg->getNumOperands()-1;  // don't include \0
-
-    GlobalVariable *MsgGV = new GlobalVariable(*M, Msg->getType(), true,
-                                               GlobalValue::InternalLinkage,
-                                               Msg, "abortmsg");
-    std::vector<Constant*> GEPIdx(2,
-                     Constant::getNullValue(Type::getInt32Ty(M->getContext())));
-    AbortMessage = ConstantExpr::getGetElementPtr(MsgGV, &GEPIdx[0], 2);
-  } else {
-    // The abort message for cheap EH support tells the user that EH is not
-    // enabled.
-    Constant *Msg =
-      ConstantArray::get(M->getContext(), 
-                        "Exception handler needed, but not enabled."      
-                        "Recompile program with -enable-correct-eh-support.\n");
-    AbortMessageLength = Msg->getNumOperands()-1;  // don't include \0
-
-    GlobalVariable *MsgGV = new GlobalVariable(*M, Msg->getType(), true,
-                                               GlobalValue::InternalLinkage,
-                                               Msg, "abortmsg");
-    std::vector<Constant*> GEPIdx(2, Constant::getNullValue(
-                                            Type::getInt32Ty(M->getContext())));
-    AbortMessage = ConstantExpr::getGetElementPtr(MsgGV, &GEPIdx[0], 2);
-  }
-}
-
-
-void LowerInvoke::writeAbortMessage(Instruction *IB) {
-#if 0
-  if (AbortMessage == 0)
-    createAbortMessage(IB->getParent()->getParent()->getParent());
-
-  // These are the arguments we WANT...
-  Value* Args[3];
-  Args[0] = ConstantInt::get(Type::Int32Ty, 2);
-  Args[1] = AbortMessage;
-  Args[2] = ConstantInt::get(Type::Int32Ty, AbortMessageLength);
-  (new CallInst(WriteFn, Args, 3, "", IB))->setTailCall();
-#endif
-}
-
 bool LowerInvoke::insertCheapEHSupport(Function &F) {
   bool Changed = false;
   for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
     if (InvokeInst *II = dyn_cast<InvokeInst>(BB->getTerminator())) {
-      std::vector<Value*> CallArgs(II->op_begin(), II->op_end() - 3);
+      SmallVector<Value*,16> CallArgs(II->op_begin(), II->op_end() - 3);
       // Insert a normal call instruction...
       CallInst *NewCall = CallInst::Create(II->getCalledValue(),
                                            CallArgs.begin(), CallArgs.end(),
@@ -257,9 +198,6 @@ bool LowerInvoke::insertCheapEHSupport(Function &F) {
 
       ++NumInvokes; Changed = true;
     } else if (UnwindInst *UI = dyn_cast<UnwindInst>(BB->getTerminator())) {
-      // Insert a new call to write(2, AbortMessage, AbortMessageLength);
-      writeAbortMessage(UI);
-
       // Insert a call to abort()
       CallInst::Create(AbortFn, "", UI)->setTailCall();
 
@@ -320,7 +258,7 @@ void LowerInvoke::rewriteExpensiveInvoke(InvokeInst *II, unsigned InvokeNo,
   CatchSwitch->addCase(InvokeNoC, II->getUnwindDest());
 
   // Insert a normal call instruction.
-  std::vector<Value*> CallArgs(II->op_begin(), II->op_end() - 3);
+  SmallVector<Value*,16> CallArgs(II->op_begin(), II->op_end() - 3);
   CallInst *NewCall = CallInst::Create(II->getCalledValue(),
                                        CallArgs.begin(), CallArgs.end(), "",
                                        II);
@@ -349,7 +287,7 @@ static void MarkBlocksLiveIn(BasicBlock *BB, std::set<BasicBlock*> &LiveBBs) {
 // across the unwind edge.  This process also splits all critical edges
 // coming out of invoke's.
 void LowerInvoke::
-splitLiveRangesLiveAcrossInvokes(std::vector<InvokeInst*> &Invokes) {
+splitLiveRangesLiveAcrossInvokes(SmallVectorImpl<InvokeInst*> &Invokes) {
   // First step, split all critical edges from invoke instructions.
   for (unsigned i = 0, e = Invokes.size(); i != e; ++i) {
     InvokeInst *II = Invokes[i];
@@ -371,16 +309,33 @@ splitLiveRangesLiveAcrossInvokes(std::vector<InvokeInst*> &Invokes) {
     ++AfterAllocaInsertPt;
   for (Function::arg_iterator AI = F->arg_begin(), E = F->arg_end();
        AI != E; ++AI) {
-    // 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);
-    // Normally its is 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 to simply
-    // make NC its user.
-    NC->setOperand(0, 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.
@@ -402,7 +357,7 @@ splitLiveRangesLiveAcrossInvokes(std::vector<InvokeInst*> &Invokes) {
           continue;
 
       // Avoid iterator invalidation by copying users to a temporary vector.
-      std::vector<Instruction*> Users;
+      SmallVector<Instruction*,16> Users;
       for (Value::use_iterator UI = Inst->use_begin(), E = Inst->use_end();
            UI != E; ++UI) {
         Instruction *User = cast<Instruction>(*UI);
@@ -452,9 +407,9 @@ splitLiveRangesLiveAcrossInvokes(std::vector<InvokeInst*> &Invokes) {
 }
 
 bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
-  std::vector<ReturnInst*> Returns;
-  std::vector<UnwindInst*> Unwinds;
-  std::vector<InvokeInst*> Invokes;
+  SmallVector<ReturnInst*,16> Returns;
+  SmallVector<UnwindInst*,16> Unwinds;
+  SmallVector<InvokeInst*,16> Invokes;
 
   for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
     if (ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator())) {
@@ -502,12 +457,11 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
       new AllocaInst(JBLinkTy, 0, Align,
                      "jblink", F.begin()->begin());
 
-    std::vector<Value*> Idx;
-    Idx.push_back(Constant::getNullValue(Type::getInt32Ty(F.getContext())));
-    Idx.push_back(ConstantInt::get(Type::getInt32Ty(F.getContext()), 1));
-    OldJmpBufPtr = GetElementPtrInst::Create(JmpBuf, Idx.begin(), Idx.end(),
+    Value *Idx[] = { Constant::getNullValue(Type::getInt32Ty(F.getContext())),
+                     ConstantInt::get(Type::getInt32Ty(F.getContext()), 1) };
+    OldJmpBufPtr = GetElementPtrInst::Create(JmpBuf, &Idx[0], &Idx[2],
                                              "OldBuf",
-                                              EntryBB->getTerminator());
+                                             EntryBB->getTerminator());
 
     // Copy the JBListHead to the alloca.
     Value *OldBuf = new LoadInst(JBListHead, "oldjmpbufptr", true,
@@ -552,7 +506,7 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
                                                      "setjmp.cont");
 
     Idx[1] = ConstantInt::get(Type::getInt32Ty(F.getContext()), 0);
-    Value *JmpBufPtr = GetElementPtrInst::Create(JmpBuf, Idx.begin(), Idx.end(),
+    Value *JmpBufPtr = GetElementPtrInst::Create(JmpBuf, &Idx[0], &Idx[2],
                                                  "TheJmpBuf",
                                                  EntryBB->getTerminator());
     JmpBufPtr = new BitCastInst(JmpBufPtr,
@@ -605,24 +559,20 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) {
 
   // Create the block to do the longjmp.
   // Get a pointer to the jmpbuf and longjmp.
-  std::vector<Value*> Idx;
-  Idx.push_back(Constant::getNullValue(Type::getInt32Ty(F.getContext())));
-  Idx.push_back(ConstantInt::get(Type::getInt32Ty(F.getContext()), 0));
-  Idx[0] = GetElementPtrInst::Create(BufPtr, Idx.begin(), Idx.end(), "JmpBuf",
+  Value *Idx[] = { Constant::getNullValue(Type::getInt32Ty(F.getContext())),
+                   ConstantInt::get(Type::getInt32Ty(F.getContext()), 0) };
+  Idx[0] = GetElementPtrInst::Create(BufPtr, &Idx[0], &Idx[2], "JmpBuf",
                                      UnwindBlock);
   Idx[0] = new BitCastInst(Idx[0],
              Type::getInt8PtrTy(F.getContext()),
                            "tmp", UnwindBlock);
   Idx[1] = ConstantInt::get(Type::getInt32Ty(F.getContext()), 1);
-  CallInst::Create(LongJmpFn, Idx.begin(), Idx.end(), "", UnwindBlock);
+  CallInst::Create(LongJmpFn, &Idx[0], &Idx[2], "", UnwindBlock);
   new UnreachableInst(F.getContext(), UnwindBlock);
 
   // Set up the term block ("throw without a catch").
   new UnreachableInst(F.getContext(), TermBlock);
 
-  // Insert a new call to write(2, AbortMessage, AbortMessageLength);
-  writeAbortMessage(TermBlock->getTerminator());
-
   // Insert a call to abort()
   CallInst::Create(AbortFn, "",
                    TermBlock->getTerminator())->setTailCall();
diff --git a/lib/Transforms/Utils/PromoteMemoryToRegister.cpp b/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
index 13f0a28..c0de193 100644
--- a/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
+++ b/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
@@ -69,11 +69,12 @@ bool llvm::isAllocaPromotable(const AllocaInst *AI) {
 
   // Only allow direct and non-volatile loads and stores...
   for (Value::const_use_iterator UI = AI->use_begin(), UE = AI->use_end();
-       UI != UE; ++UI)     // Loop over all of the uses of the alloca
-    if (const LoadInst *LI = dyn_cast<LoadInst>(*UI)) {
+       UI != UE; ++UI) {   // Loop over all of the uses of the alloca
+    const User *U = *UI;
+    if (const LoadInst *LI = dyn_cast<LoadInst>(U)) {
       if (LI->isVolatile())
         return false;
-    } else if (const StoreInst *SI = dyn_cast<StoreInst>(*UI)) {
+    } else if (const StoreInst *SI = dyn_cast<StoreInst>(U)) {
       if (SI->getOperand(0) == AI)
         return false;   // Don't allow a store OF the AI, only INTO the AI.
       if (SI->isVolatile())
@@ -81,6 +82,7 @@ bool llvm::isAllocaPromotable(const AllocaInst *AI) {
     } else {
       return false;
     }
+  }
 
   return true;
 }
@@ -603,9 +605,8 @@ ComputeLiveInBlocks(AllocaInst *AI, AllocaInfo &Info,
   // To determine liveness, we must iterate through the predecessors of blocks
   // where the def is live.  Blocks are added to the worklist if we need to
   // check their predecessors.  Start with all the using blocks.
-  SmallVector<BasicBlock*, 64> LiveInBlockWorklist;
-  LiveInBlockWorklist.insert(LiveInBlockWorklist.end(), 
-                             Info.UsingBlocks.begin(), Info.UsingBlocks.end());
+  SmallVector<BasicBlock*, 64> LiveInBlockWorklist(Info.UsingBlocks.begin(),
+                                                   Info.UsingBlocks.end());
   
   // If any of the using blocks is also a definition block, check to see if the
   // definition occurs before or after the use.  If it happens before the use,
@@ -897,6 +898,9 @@ void PromoteMem2Reg::ConvertDebugDeclareToDebugValue(DbgDeclareInst *DDI,
   // Propagate any debug metadata from the store onto the dbg.value.
   if (MDNode *SIMD = SI->getMetadata("dbg"))
     DbgVal->setMetadata("dbg", SIMD);
+  // Otherwise propagate debug metadata from dbg.declare.
+  else if (MDNode *MD = DDI->getMetadata("dbg"))
+      DbgVal->setMetadata("dbg", MD);         
 }
 
 // QueuePhiNode - queues a phi-node to be added to a basic-block for a specific
diff --git a/lib/Transforms/Utils/SimplifyCFG.cpp b/lib/Transforms/Utils/SimplifyCFG.cpp
index 9f2209d..fd3ed3e 100644
--- a/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -1513,17 +1513,19 @@ static bool SimplifyCondBranchToCondBranch(BranchInst *PBI, BranchInst *BI) {
       // Okay, we're going to insert the PHI node.  Since PBI is not the only
       // predecessor, compute the PHI'd conditional value for all of the preds.
       // Any predecessor where the condition is not computable we keep symbolic.
-      for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
-        if ((PBI = dyn_cast<BranchInst>((*PI)->getTerminator())) &&
+      for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
+        BasicBlock *P = *PI;
+        if ((PBI = dyn_cast<BranchInst>(P->getTerminator())) &&
             PBI != BI && PBI->isConditional() &&
             PBI->getCondition() == BI->getCondition() &&
             PBI->getSuccessor(0) != PBI->getSuccessor(1)) {
           bool CondIsTrue = PBI->getSuccessor(0) == BB;
           NewPN->addIncoming(ConstantInt::get(Type::getInt1Ty(BB->getContext()), 
-                                              CondIsTrue), *PI);
+                                              CondIsTrue), P);
         } else {
-          NewPN->addIncoming(BI->getCondition(), *PI);
+          NewPN->addIncoming(BI->getCondition(), P);
         }
+      }
       
       BI->setCondition(NewPN);
       return true;
@@ -1697,10 +1699,11 @@ bool SimplifyCFGOpt::run(BasicBlock *BB) {
       SmallVector<BasicBlock*, 8> UncondBranchPreds;
       SmallVector<BranchInst*, 8> CondBranchPreds;
       for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
-        TerminatorInst *PTI = (*PI)->getTerminator();
+        BasicBlock *P = *PI;
+        TerminatorInst *PTI = P->getTerminator();
         if (BranchInst *BI = dyn_cast<BranchInst>(PTI)) {
           if (BI->isUnconditional())
-            UncondBranchPreds.push_back(*PI);
+            UncondBranchPreds.push_back(P);
           else
             CondBranchPreds.push_back(BI);
         }
diff --git a/lib/Transforms/Utils/ValueMapper.cpp b/lib/Transforms/Utils/ValueMapper.cpp
index 87ce631..3f6a90c 100644
--- a/lib/Transforms/Utils/ValueMapper.cpp
+++ b/lib/Transforms/Utils/ValueMapper.cpp
@@ -28,7 +28,7 @@ Value *llvm::MapValue(const Value *V, ValueToValueMapTy &VM) {
   // DenseMap.  This includes any recursive calls to MapValue.
 
   // Global values and non-function-local metadata do not need to be seeded into
-  // the ValueMap if they are using the identity mapping.
+  // the VM if they are using the identity mapping.
   if (isa<GlobalValue>(V) || isa<InlineAsm>(V) || isa<MDString>(V) ||
       (isa<MDNode>(V) && !cast<MDNode>(V)->isFunctionLocal()))
     return VMSlot = const_cast<Value*>(V);
@@ -45,7 +45,7 @@ Value *llvm::MapValue(const Value *V, ValueToValueMapTy &VM) {
   
   if (isa<ConstantInt>(C) || isa<ConstantFP>(C) ||
       isa<ConstantPointerNull>(C) || isa<ConstantAggregateZero>(C) ||
-      isa<UndefValue>(C) || isa<MDString>(C))
+      isa<UndefValue>(C))
     return VMSlot = C;           // Primitive constants map directly
   
   if (ConstantArray *CA = dyn_cast<ConstantArray>(C)) {
@@ -125,11 +125,11 @@ Value *llvm::MapValue(const Value *V, ValueToValueMapTy &VM) {
 }
 
 /// RemapInstruction - Convert the instruction operands from referencing the
-/// current values into those specified by ValueMap.
+/// current values into those specified by VMap.
 ///
-void llvm::RemapInstruction(Instruction *I, ValueToValueMapTy &ValueMap) {
+void llvm::RemapInstruction(Instruction *I, ValueToValueMapTy &VMap) {
   for (User::op_iterator op = I->op_begin(), E = I->op_end(); op != E; ++op) {
-    Value *V = MapValue(*op, ValueMap);
+    Value *V = MapValue(*op, VMap);
     assert(V && "Referenced value not in value map!");
     *op = V;
   }
diff --git a/lib/Transforms/Utils/ValueMapper.h b/lib/Transforms/Utils/ValueMapper.h
index d61c24c..f4ff643 100644
--- a/lib/Transforms/Utils/ValueMapper.h
+++ b/lib/Transforms/Utils/ValueMapper.h
@@ -15,12 +15,12 @@
 #ifndef VALUEMAPPER_H
 #define VALUEMAPPER_H
 
-#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/ValueMap.h"
 
 namespace llvm {
   class Value;
   class Instruction;
-  typedef DenseMap<const Value *, Value *> ValueToValueMapTy;
+  typedef ValueMap<const Value *, Value *> ValueToValueMapTy;
 
   Value *MapValue(const Value *V, ValueToValueMapTy &VM);
   void RemapInstruction(Instruction *I, ValueToValueMapTy &VM);