diff --git a/examples/BrainF/BrainF.cpp b/examples/BrainF/BrainF.cpp
index f17a950..a443ad4 100644
--- a/examples/BrainF/BrainF.cpp
+++ b/examples/BrainF/BrainF.cpp
@@ -81,8 +81,11 @@
   ConstantInt *val_mem = ConstantInt::get(C, APInt(32, memtotal));
   BasicBlock* BB = builder->GetInsertBlock();
   const Type* IntPtrTy = IntegerType::getInt32Ty(C);
-  ptr_arr = CallInst::CreateMalloc(BB, IntPtrTy, IntegerType::getInt8Ty(C),
-                                   val_mem, NULL, "arr");
+  const Type* Int8Ty = IntegerType::getInt8Ty(C);
+  Constant* allocsize = ConstantExpr::getSizeOf(Int8Ty);
+  allocsize = ConstantExpr::getTruncOrBitCast(allocsize, IntPtrTy);
+  ptr_arr = CallInst::CreateMalloc(BB, IntPtrTy, Int8Ty, allocsize, val_mem, 
+                                   NULL, "arr");
   BB->getInstList().push_back(cast<Instruction>(ptr_arr));
 
   //call void @llvm.memset.i32(i8 *%arr, i8 0, i32 %d, i32 1)
diff --git a/include/llvm/Analysis/MemoryBuiltins.h b/include/llvm/Analysis/MemoryBuiltins.h
index 5fd0bb0..fde7dc6 100644
--- a/include/llvm/Analysis/MemoryBuiltins.h
+++ b/include/llvm/Analysis/MemoryBuiltins.h
@@ -50,13 +50,17 @@
                               const TargetData* TD);
 
 /// getMallocType - Returns the PointerType resulting from the malloc call.
-/// This PointerType is the result type of the call's only bitcast use.
-/// If there is no unique bitcast use, then return NULL.
+/// The PointerType depends on the number of bitcast uses of the malloc call:
+///   0: PointerType is the malloc calls' return type.
+///   1: PointerType is the bitcast's result type.
+///  >1: Unique PointerType cannot be determined, return NULL.
 const PointerType* getMallocType(const CallInst* CI);
 
-/// getMallocAllocatedType - Returns the Type allocated by malloc call. This
-/// Type is the result type of the call's only bitcast use. If there is no
-/// unique bitcast use, then return NULL.
+/// getMallocAllocatedType - Returns the Type allocated by malloc call.
+/// The Type depends on the number of bitcast uses of the malloc call:
+///   0: PointerType is the malloc calls' return type.
+///   1: PointerType is the bitcast's result type.
+///  >1: Unique PointerType cannot be determined, return NULL.
 const Type* getMallocAllocatedType(const CallInst* CI);
 
 /// getMallocArraySize - Returns the array size of a malloc call.  If the 
diff --git a/include/llvm/Instructions.h b/include/llvm/Instructions.h
index 28854df..5b48e1a 100644
--- a/include/llvm/Instructions.h
+++ b/include/llvm/Instructions.h
@@ -899,11 +899,12 @@
   /// 3. Bitcast the result of the malloc call to the specified type.
   static Instruction *CreateMalloc(Instruction *InsertBefore,
                                    const Type *IntPtrTy, const Type *AllocTy,
-                                   Value *ArraySize = 0,
+                                   Value *AllocSize, Value *ArraySize = 0,
                                    const Twine &Name = "");
   static Instruction *CreateMalloc(BasicBlock *InsertAtEnd,
                                    const Type *IntPtrTy, const Type *AllocTy,
-                                   Value *ArraySize = 0, Function* MallocF = 0,
+                                   Value *AllocSize, Value *ArraySize = 0,
+                                   Function* MallocF = 0,
                                    const Twine &Name = "");
   /// CreateFree - Generate the IR for a call to the builtin free function.
   static void CreateFree(Value* Source, Instruction *InsertBefore);
diff --git a/lib/Analysis/MemoryBuiltins.cpp b/lib/Analysis/MemoryBuiltins.cpp
index e710350..f4eb793 100644
--- a/lib/Analysis/MemoryBuiltins.cpp
+++ b/lib/Analysis/MemoryBuiltins.cpp
@@ -17,6 +17,7 @@
 #include "llvm/Instructions.h"
 #include "llvm/Module.h"
 #include "llvm/Analysis/ConstantFolding.h"
+#include "llvm/Target/TargetData.h"
 using namespace llvm;
 
 //===----------------------------------------------------------------------===//
@@ -96,45 +97,47 @@
   if (!CI)
     return NULL;
 
-  // Type must be known to determine array size.
+  // The size of the malloc's result type must be known to determine array size.
   const Type *T = getMallocAllocatedType(CI);
-  if (!T)
+  if (!T || !T->isSized() || !TD)
     return NULL;
 
   Value *MallocArg = CI->getOperand(1);
+  const Type *ArgType = MallocArg->getType();
   ConstantExpr *CO = dyn_cast<ConstantExpr>(MallocArg);
   BinaryOperator *BO = dyn_cast<BinaryOperator>(MallocArg);
 
-  Constant *ElementSize = ConstantExpr::getSizeOf(T);
-  ElementSize = ConstantExpr::getTruncOrBitCast(ElementSize, 
-                                                MallocArg->getType());
-  Constant *FoldedElementSize =
-   ConstantFoldConstantExpression(cast<ConstantExpr>(ElementSize), Context, TD);
+  unsigned ElementSizeInt = TD->getTypeAllocSize(T);
+  if (const StructType *ST = dyn_cast<StructType>(T))
+    ElementSizeInt = TD->getStructLayout(ST)->getSizeInBytes();
+  Constant *ElementSize = ConstantInt::get(ArgType, ElementSizeInt);
 
   // First, check if CI is a non-array malloc.
-  if (CO && ((CO == ElementSize) ||
-             (FoldedElementSize && (CO == FoldedElementSize))))
+  if (CO && CO == ElementSize)
     // Match CreateMalloc's use of constant 1 array-size for non-array mallocs.
-    return ConstantInt::get(MallocArg->getType(), 1);
+    return ConstantInt::get(ArgType, 1);
 
   // Second, check if CI is an array malloc whose array size can be determined.
-  if (isConstantOne(ElementSize) || 
-      (FoldedElementSize && isConstantOne(FoldedElementSize)))
+  if (isConstantOne(ElementSize))
     return MallocArg;
 
+  if (ConstantInt *CInt = dyn_cast<ConstantInt>(MallocArg))
+    if (CInt->getZExtValue() % ElementSizeInt == 0)
+      return ConstantInt::get(ArgType, CInt->getZExtValue() / ElementSizeInt);
+
   if (!CO && !BO)
     return NULL;
 
   Value *Op0 = NULL;
   Value *Op1 = NULL;
   unsigned Opcode = 0;
-  if (CO && ((CO->getOpcode() == Instruction::Mul) || 
+  if (CO && ((CO->getOpcode() == Instruction::Mul) ||
              (CO->getOpcode() == Instruction::Shl))) {
     Op0 = CO->getOperand(0);
     Op1 = CO->getOperand(1);
     Opcode = CO->getOpcode();
   }
-  if (BO && ((BO->getOpcode() == Instruction::Mul) || 
+  if (BO && ((BO->getOpcode() == Instruction::Mul) ||
              (BO->getOpcode() == Instruction::Shl))) {
     Op0 = BO->getOperand(0);
     Op1 = BO->getOperand(1);
@@ -144,12 +147,10 @@
   // Determine array size if malloc's argument is the product of a mul or shl.
   if (Op0) {
     if (Opcode == Instruction::Mul) {
-      if ((Op1 == ElementSize) ||
-          (FoldedElementSize && (Op1 == FoldedElementSize)))
+      if (Op1 == ElementSize)
         // ArraySize * ElementSize
         return Op0;
-      if ((Op0 == ElementSize) ||
-          (FoldedElementSize && (Op0 == FoldedElementSize)))
+      if (Op0 == ElementSize)
         // ElementSize * ArraySize
         return Op1;
     }
@@ -161,11 +162,10 @@
       uint64_t BitToSet = Op1Int.getLimitedValue(Op1Int.getBitWidth() - 1);
       Value *Op1Pow = ConstantInt::get(Context, 
                                   APInt(Op1Int.getBitWidth(), 0).set(BitToSet));
-      if (Op0 == ElementSize || (FoldedElementSize && Op0 == FoldedElementSize))
+      if (Op0 == ElementSize)
         // ArraySize << log2(ElementSize)
         return Op1Pow;
-      if (Op1Pow == ElementSize ||
-          (FoldedElementSize && Op1Pow == FoldedElementSize))
+      if (Op1Pow == ElementSize)
         // ElementSize << log2(ArraySize)
         return Op0;
     }
@@ -205,35 +205,41 @@
 }
 
 /// getMallocType - Returns the PointerType resulting from the malloc call.
-/// This PointerType is the result type of the call's only bitcast use.
-/// If there is no unique bitcast use, then return NULL.
+/// The PointerType depends on the number of bitcast uses of the malloc call:
+///   0: PointerType is the calls' return type.
+///   1: PointerType is the bitcast's result type.
+///  >1: Unique PointerType cannot be determined, return NULL.
 const PointerType *llvm::getMallocType(const CallInst *CI) {
   assert(isMalloc(CI) && "GetMallocType and not malloc call");
   
-  const BitCastInst *BCI = NULL;
-  
+  const PointerType *MallocType = NULL;
+  unsigned NumOfBitCastUses = 0;
+
   // Determine if CallInst has a bitcast use.
   for (Value::use_const_iterator UI = CI->use_begin(), E = CI->use_end();
        UI != E; )
-    if ((BCI = dyn_cast<BitCastInst>(cast<Instruction>(*UI++))))
-      break;
+    if (const BitCastInst *BCI = dyn_cast<BitCastInst>(*UI++)) {
+      MallocType = cast<PointerType>(BCI->getDestTy());
+      NumOfBitCastUses++;
+    }
 
-  // Malloc call has 1 bitcast use and no other uses, so type is the bitcast's
-  // destination type.
-  if (BCI && CI->hasOneUse())
-    return cast<PointerType>(BCI->getDestTy());
+  // Malloc call has 1 bitcast use, so type is the bitcast's destination type.
+  if (NumOfBitCastUses == 1)
+    return MallocType;
 
   // Malloc call was not bitcast, so type is the malloc function's return type.
-  if (!BCI)
+  if (NumOfBitCastUses == 0)
     return cast<PointerType>(CI->getType());
 
   // Type could not be determined.
   return NULL;
 }
 
-/// getMallocAllocatedType - Returns the Type allocated by malloc call. This
-/// Type is the result type of the call's only bitcast use. If there is no
-/// unique bitcast use, then return NULL.
+/// getMallocAllocatedType - Returns the Type allocated by malloc call.
+/// The Type depends on the number of bitcast uses of the malloc call:
+///   0: PointerType is the malloc calls' return type.
+///   1: PointerType is the bitcast's result type.
+///  >1: Unique PointerType cannot be determined, return NULL.
 const Type *llvm::getMallocAllocatedType(const CallInst *CI) {
   const PointerType *PT = getMallocType(CI);
   return PT ? PT->getElementType() : NULL;
diff --git a/lib/AsmParser/LLParser.cpp b/lib/AsmParser/LLParser.cpp
index 0da0f4a..8e70170 100644
--- a/lib/AsmParser/LLParser.cpp
+++ b/lib/AsmParser/LLParser.cpp
@@ -3619,12 +3619,14 @@
   // Autoupgrade old malloc instruction to malloc call.
   // FIXME: Remove in LLVM 3.0.
   const Type *IntPtrTy = Type::getInt32Ty(Context);
+  Constant *AllocSize = ConstantExpr::getSizeOf(Ty);
+  AllocSize = ConstantExpr::getTruncOrBitCast(AllocSize, IntPtrTy);
   if (!MallocF)
     // Prototype malloc as "void *(int32)".
     // This function is renamed as "malloc" in ValidateEndOfModule().
     MallocF = cast<Function>(
        M->getOrInsertFunction("", Type::getInt8PtrTy(Context), IntPtrTy, NULL));
-  Inst = CallInst::CreateMalloc(BB, IntPtrTy, Ty, Size, MallocF);
+  Inst = CallInst::CreateMalloc(BB, IntPtrTy, Ty, AllocSize, Size, MallocF);
   return false;
 }
 
diff --git a/lib/Bitcode/Reader/BitcodeReader.cpp b/lib/Bitcode/Reader/BitcodeReader.cpp
index 68527e3..9916388 100644
--- a/lib/Bitcode/Reader/BitcodeReader.cpp
+++ b/lib/Bitcode/Reader/BitcodeReader.cpp
@@ -2101,8 +2101,10 @@
       if (!Ty || !Size) return Error("Invalid MALLOC record");
       if (!CurBB) return Error("Invalid malloc instruction with no BB");
       const Type *Int32Ty = IntegerType::getInt32Ty(CurBB->getContext());
+      Constant *AllocSize = ConstantExpr::getSizeOf(Ty->getElementType());
+      AllocSize = ConstantExpr::getTruncOrBitCast(AllocSize, Int32Ty);
       I = CallInst::CreateMalloc(CurBB, Int32Ty, Ty->getElementType(),
-                                 Size, NULL);
+                                 AllocSize, Size, NULL);
       InstructionList.push_back(I);
       break;
     }
diff --git a/lib/Transforms/IPO/GlobalOpt.cpp b/lib/Transforms/IPO/GlobalOpt.cpp
index 5dab9ef..234d0ec 100644
--- a/lib/Transforms/IPO/GlobalOpt.cpp
+++ b/lib/Transforms/IPO/GlobalOpt.cpp
@@ -822,32 +822,42 @@
 /// malloc into a global, and any loads of GV as uses of the new global.
 static GlobalVariable *OptimizeGlobalAddressOfMalloc(GlobalVariable *GV,
                                                      CallInst *CI,
-                                                     BitCastInst *BCI,
+                                                     const Type *AllocTy,
                                                      Value* NElems,
                                                      LLVMContext &Context,
                                                      TargetData* TD) {
-  DEBUG(errs() << "PROMOTING MALLOC GLOBAL: " << *GV
-               << "  CALL = " << *CI << "  BCI = " << *BCI << '\n');
+  DEBUG(errs() << "PROMOTING GLOBAL: " << *GV << "  CALL = " << *CI << '\n');
 
   const Type *IntPtrTy = TD->getIntPtrType(Context);
   
+  // CI has either 0 or 1 bitcast uses (getMallocType() would otherwise have
+  // returned NULL and we would not be here).
+  BitCastInst *BCI = NULL;
+  for (Value::use_iterator UI = CI->use_begin(), E = CI->use_end(); UI != E; )
+    if ((BCI = dyn_cast<BitCastInst>(cast<Instruction>(*UI++))))
+      break;
+
   ConstantInt *NElements = cast<ConstantInt>(NElems);
   if (NElements->getZExtValue() != 1) {
     // If we have an array allocation, transform it to a single element
     // allocation to make the code below simpler.
-    Type *NewTy = ArrayType::get(getMallocAllocatedType(CI),
-                                 NElements->getZExtValue());
-    Value* NewM = CallInst::CreateMalloc(CI, IntPtrTy, NewTy);
-    Instruction* NewMI = cast<Instruction>(NewM);
+    Type *NewTy = ArrayType::get(AllocTy, NElements->getZExtValue());
+    unsigned TypeSize = TD->getTypeAllocSize(NewTy);
+    if (const StructType *ST = dyn_cast<StructType>(NewTy))
+      TypeSize = TD->getStructLayout(ST)->getSizeInBytes();
+    Instruction *NewCI = CallInst::CreateMalloc(CI, IntPtrTy, NewTy,
+                                         ConstantInt::get(IntPtrTy, TypeSize));
     Value* Indices[2];
     Indices[0] = Indices[1] = Constant::getNullValue(IntPtrTy);
-    Value *NewGEP = GetElementPtrInst::Create(NewMI, Indices, Indices + 2,
-                                              NewMI->getName()+".el0", CI);
-    BCI->replaceAllUsesWith(NewGEP);
-    BCI->eraseFromParent();
+    Value *NewGEP = GetElementPtrInst::Create(NewCI, Indices, Indices + 2,
+                                              NewCI->getName()+".el0", CI);
+    Value *Cast = new BitCastInst(NewGEP, CI->getType(), "el0", CI);
+    if (BCI) BCI->replaceAllUsesWith(NewGEP);
+    CI->replaceAllUsesWith(Cast);
+    if (BCI) BCI->eraseFromParent();
     CI->eraseFromParent();
-    BCI = cast<BitCastInst>(NewMI);
-    CI = extractMallocCallFromBitCast(NewMI);
+    BCI = dyn_cast<BitCastInst>(NewCI);
+    CI = BCI ? extractMallocCallFromBitCast(BCI) : cast<CallInst>(NewCI);
   }
 
   // Create the new global variable.  The contents of the malloc'd memory is
@@ -861,8 +871,9 @@
                                              GV,
                                              GV->isThreadLocal());
   
-  // Anything that used the malloc now uses the global directly.
-  BCI->replaceAllUsesWith(NewGV);
+  // Anything that used the malloc or its bitcast now uses the global directly.
+  if (BCI) BCI->replaceAllUsesWith(NewGV);
+  CI->replaceAllUsesWith(new BitCastInst(NewGV, CI->getType(), "newgv", CI));
 
   Constant *RepValue = NewGV;
   if (NewGV->getType() != GV->getType()->getElementType())
@@ -930,9 +941,9 @@
     GV->getParent()->getGlobalList().insert(GV, InitBool);
 
 
-  // Now the GV is dead, nuke it and the malloc.
+  // Now the GV is dead, nuke it and the malloc (both CI and BCI).
   GV->eraseFromParent();
-  BCI->eraseFromParent();
+  if (BCI) BCI->eraseFromParent();
   CI->eraseFromParent();
 
   // To further other optimizations, loop over all users of NewGV and try to
@@ -1273,13 +1284,10 @@
 
 /// PerformHeapAllocSRoA - CI is an allocation of an array of structures.  Break
 /// it up into multiple allocations of arrays of the fields.
-static GlobalVariable *PerformHeapAllocSRoA(GlobalVariable *GV,
-                                            CallInst *CI, BitCastInst* BCI,
-                                            Value* NElems,
-                                            LLVMContext &Context,
+static GlobalVariable *PerformHeapAllocSRoA(GlobalVariable *GV, CallInst *CI,
+                                            Value* NElems, LLVMContext &Context,
                                             TargetData *TD) {
-  DEBUG(errs() << "SROA HEAP ALLOC: " << *GV << "  MALLOC CALL = " << *CI 
-               << " BITCAST = " << *BCI << '\n');
+  DEBUG(errs() << "SROA HEAP ALLOC: " << *GV << "  MALLOC = " << *CI << '\n');
   const Type* MAT = getMallocAllocatedType(CI);
   const StructType *STy = cast<StructType>(MAT);
 
@@ -1287,8 +1295,8 @@
   // it into GV).  If there are other uses, change them to be uses of
   // the global to simplify later code.  This also deletes the store
   // into GV.
-  ReplaceUsesOfMallocWithGlobal(BCI, GV);
-  
+  ReplaceUsesOfMallocWithGlobal(CI, GV);
+
   // Okay, at this point, there are no users of the malloc.  Insert N
   // new mallocs at the same place as CI, and N globals.
   std::vector<Value*> FieldGlobals;
@@ -1306,11 +1314,16 @@
                          GV->isThreadLocal());
     FieldGlobals.push_back(NGV);
     
-    Value *NMI = CallInst::CreateMalloc(CI, TD->getIntPtrType(Context),
-                                        FieldTy, NElems,
-                                        BCI->getName() + ".f" + Twine(FieldNo));
+    unsigned TypeSize = TD->getTypeAllocSize(FieldTy);
+    if (const StructType* ST = dyn_cast<StructType>(FieldTy))
+      TypeSize = TD->getStructLayout(ST)->getSizeInBytes();
+    const Type* IntPtrTy = TD->getIntPtrType(Context);
+    Value *NMI = CallInst::CreateMalloc(CI, IntPtrTy, FieldTy,
+                                        ConstantInt::get(IntPtrTy, TypeSize),
+                                        NElems,
+                                        CI->getName() + ".f" + Twine(FieldNo));
     FieldMallocs.push_back(NMI);
-    new StoreInst(NMI, NGV, BCI);
+    new StoreInst(NMI, NGV, CI);
   }
   
   // The tricky aspect of this transformation is handling the case when malloc
@@ -1327,18 +1340,18 @@
   //    }
   Value *RunningOr = 0;
   for (unsigned i = 0, e = FieldMallocs.size(); i != e; ++i) {
-    Value *Cond = new ICmpInst(BCI, ICmpInst::ICMP_EQ, FieldMallocs[i],
-                              Constant::getNullValue(FieldMallocs[i]->getType()),
-                                  "isnull");
+    Value *Cond = new ICmpInst(CI, ICmpInst::ICMP_EQ, FieldMallocs[i],
+                             Constant::getNullValue(FieldMallocs[i]->getType()),
+                               "isnull");
     if (!RunningOr)
       RunningOr = Cond;   // First seteq
     else
-      RunningOr = BinaryOperator::CreateOr(RunningOr, Cond, "tmp", BCI);
+      RunningOr = BinaryOperator::CreateOr(RunningOr, Cond, "tmp", CI);
   }
 
   // Split the basic block at the old malloc.
-  BasicBlock *OrigBB = BCI->getParent();
-  BasicBlock *ContBB = OrigBB->splitBasicBlock(BCI, "malloc_cont");
+  BasicBlock *OrigBB = CI->getParent();
+  BasicBlock *ContBB = OrigBB->splitBasicBlock(CI, "malloc_cont");
   
   // Create the block to check the first condition.  Put all these blocks at the
   // end of the function as they are unlikely to be executed.
@@ -1374,9 +1387,8 @@
   }
   
   BranchInst::Create(ContBB, NullPtrBlock);
-  
-  // CI and BCI are no longer needed, remove them.
-  BCI->eraseFromParent();
+
+  // CI is no longer needed, remove it.
   CI->eraseFromParent();
 
   /// InsertedScalarizedLoads - As we process loads, if we can't immediately
@@ -1463,14 +1475,10 @@
 /// cast of malloc.
 static bool TryToOptimizeStoreOfMallocToGlobal(GlobalVariable *GV,
                                                CallInst *CI,
-                                               BitCastInst *BCI,
+                                               const Type *AllocTy,
                                                Module::global_iterator &GVI,
                                                TargetData *TD,
                                                LLVMContext &Context) {
-  // If we can't figure out the type being malloced, then we can't optimize.
-  const Type *AllocTy = getMallocAllocatedType(CI);
-  assert(AllocTy);
-
   // If this is a malloc of an abstract type, don't touch it.
   if (!AllocTy->isSized())
     return false;
@@ -1491,7 +1499,7 @@
   // for.
   {
     SmallPtrSet<PHINode*, 8> PHIs;
-    if (!ValueIsOnlyUsedLocallyOrStoredToOneGlobal(BCI, GV, PHIs))
+    if (!ValueIsOnlyUsedLocallyOrStoredToOneGlobal(CI, GV, PHIs))
       return false;
   }  
 
@@ -1499,16 +1507,16 @@
   // transform the program to use global memory instead of malloc'd memory.
   // This eliminates dynamic allocation, avoids an indirection accessing the
   // data, and exposes the resultant global to further GlobalOpt.
-  Value *NElems = getMallocArraySize(CI, Context, TD);
   // We cannot optimize the malloc if we cannot determine malloc array size.
-  if (NElems) {
+  if (Value *NElems = getMallocArraySize(CI, Context, TD)) {
     if (ConstantInt *NElements = dyn_cast<ConstantInt>(NElems))
       // Restrict this transformation to only working on small allocations
       // (2048 bytes currently), as we don't want to introduce a 16M global or
       // something.
       if (TD && 
           NElements->getZExtValue() * TD->getTypeAllocSize(AllocTy) < 2048) {
-        GVI = OptimizeGlobalAddressOfMalloc(GV, CI, BCI, NElems, Context, TD);
+        GVI = OptimizeGlobalAddressOfMalloc(GV, CI, AllocTy, NElems,
+                                            Context, TD);
         return true;
       }
   
@@ -1526,26 +1534,29 @@
       // This the structure has an unreasonable number of fields, leave it
       // alone.
       if (AllocSTy->getNumElements() <= 16 && AllocSTy->getNumElements() != 0 &&
-          AllGlobalLoadUsesSimpleEnoughForHeapSRA(GV, BCI)) {
+          AllGlobalLoadUsesSimpleEnoughForHeapSRA(GV, CI)) {
 
         // If this is a fixed size array, transform the Malloc to be an alloc of
         // structs.  malloc [100 x struct],1 -> malloc struct, 100
         if (const ArrayType *AT =
                               dyn_cast<ArrayType>(getMallocAllocatedType(CI))) {
-          Value* NumElements = ConstantInt::get(Type::getInt32Ty(Context),
-                                                AT->getNumElements());
-          Value* NewMI = CallInst::CreateMalloc(CI, TD->getIntPtrType(Context),
-                                                AllocSTy, NumElements,
-                                                BCI->getName());
-          Value *Cast = new BitCastInst(NewMI, getMallocType(CI), "tmp", CI);
-          BCI->replaceAllUsesWith(Cast);
-          BCI->eraseFromParent();
+          const Type *IntPtrTy = TD->getIntPtrType(Context);
+          unsigned TypeSize = TD->getStructLayout(AllocSTy)->getSizeInBytes();
+          Value *AllocSize = ConstantInt::get(IntPtrTy, TypeSize);
+          Value *NumElements = ConstantInt::get(IntPtrTy, AT->getNumElements());
+          Instruction *Malloc = CallInst::CreateMalloc(CI, IntPtrTy, AllocSTy,
+                                                       AllocSize, NumElements,
+                                                       CI->getName());
+          Instruction *Cast = new BitCastInst(Malloc, CI->getType(), "tmp", CI);
+          CI->replaceAllUsesWith(Cast);
           CI->eraseFromParent();
-          BCI = cast<BitCastInst>(NewMI);
-          CI = extractMallocCallFromBitCast(NewMI);
+          CI = dyn_cast<BitCastInst>(Malloc) ?
+               extractMallocCallFromBitCast(Malloc):
+               cast<CallInst>(Malloc);
         }
       
-        GVI = PerformHeapAllocSRoA(GV, CI, BCI, NElems, Context, TD);
+        GVI = PerformHeapAllocSRoA(GV, CI, getMallocArraySize(CI, Context, TD), 
+                                   Context, TD);
         return true;
       }
     }
@@ -1577,15 +1588,10 @@
       if (OptimizeAwayTrappingUsesOfLoads(GV, SOVC, Context))
         return true;
     } else if (CallInst *CI = extractMallocCall(StoredOnceVal)) {
-      if (getMallocAllocatedType(CI)) {
-        BitCastInst* BCI = NULL;
-        for (Value::use_iterator UI = CI->use_begin(), E = CI->use_end();
-             UI != E; )
-          BCI = dyn_cast<BitCastInst>(cast<Instruction>(*UI++));
-        if (BCI &&
-            TryToOptimizeStoreOfMallocToGlobal(GV, CI, BCI, GVI, TD, Context))
-          return true;
-      }
+      const Type* MallocType = getMallocAllocatedType(CI);
+      if (MallocType && TryToOptimizeStoreOfMallocToGlobal(GV, CI, MallocType, 
+                                                           GVI, TD, Context))
+        return true;
     }
   }
 
diff --git a/lib/VMCore/Core.cpp b/lib/VMCore/Core.cpp
index 9a49d42..1a34180 100644
--- a/lib/VMCore/Core.cpp
+++ b/lib/VMCore/Core.cpp
@@ -1699,18 +1699,24 @@
 
 LLVMValueRef LLVMBuildMalloc(LLVMBuilderRef B, LLVMTypeRef Ty,
                              const char *Name) {
-  const Type* IntPtrT = Type::getInt32Ty(unwrap(B)->GetInsertBlock()->getContext());
-  return wrap(unwrap(B)->Insert(CallInst::CreateMalloc(
-      unwrap(B)->GetInsertBlock(), IntPtrT, unwrap(Ty), 0, 0, ""),
-      Twine(Name)));
+  const Type* ITy = Type::getInt32Ty(unwrap(B)->GetInsertBlock()->getContext());
+  Constant* AllocSize = ConstantExpr::getSizeOf(unwrap(Ty));
+  AllocSize = ConstantExpr::getTruncOrBitCast(AllocSize, ITy);
+  Instruction* Malloc = CallInst::CreateMalloc(unwrap(B)->GetInsertBlock(), 
+                                               ITy, unwrap(Ty), AllocSize, 
+                                               0, 0, "");
+  return wrap(unwrap(B)->Insert(Malloc, Twine(Name)));
 }
 
 LLVMValueRef LLVMBuildArrayMalloc(LLVMBuilderRef B, LLVMTypeRef Ty,
                                   LLVMValueRef Val, const char *Name) {
-  const Type* IntPtrT = Type::getInt32Ty(unwrap(B)->GetInsertBlock()->getContext());
-  return wrap(unwrap(B)->Insert(CallInst::CreateMalloc(
-      unwrap(B)->GetInsertBlock(), IntPtrT, unwrap(Ty), unwrap(Val), 0, ""),
-      Twine(Name)));
+  const Type* ITy = Type::getInt32Ty(unwrap(B)->GetInsertBlock()->getContext());
+  Constant* AllocSize = ConstantExpr::getSizeOf(unwrap(Ty));
+  AllocSize = ConstantExpr::getTruncOrBitCast(AllocSize, ITy);
+  Instruction* Malloc = CallInst::CreateMalloc(unwrap(B)->GetInsertBlock(), 
+                                               ITy, unwrap(Ty), AllocSize, 
+                                               unwrap(Val), 0, "");
+  return wrap(unwrap(B)->Insert(Malloc, Twine(Name)));
 }
 
 LLVMValueRef LLVMBuildAlloca(LLVMBuilderRef B, LLVMTypeRef Ty,
diff --git a/lib/VMCore/Instructions.cpp b/lib/VMCore/Instructions.cpp
index 52d8735..279bc73 100644
--- a/lib/VMCore/Instructions.cpp
+++ b/lib/VMCore/Instructions.cpp
@@ -24,6 +24,7 @@
 #include "llvm/Support/CallSite.h"
 #include "llvm/Support/ConstantRange.h"
 #include "llvm/Support/MathExtras.h"
+#include "llvm/Target/TargetData.h"
 
 using namespace llvm;
 
@@ -448,22 +449,11 @@
   return isa<ConstantInt>(val) && cast<ConstantInt>(val)->isOne();
 }
 
-static Value *checkArraySize(Value *Amt, const Type *IntPtrTy) {
-  if (!Amt)
-    Amt = ConstantInt::get(IntPtrTy, 1);
-  else {
-    assert(!isa<BasicBlock>(Amt) &&
-           "Passed basic block into malloc size parameter! Use other ctor");
-    assert(Amt->getType() == IntPtrTy &&
-           "Malloc array size is not an intptr!");
-  }
-  return Amt;
-}
-
 static Instruction *createMalloc(Instruction *InsertBefore,
                                  BasicBlock *InsertAtEnd, const Type *IntPtrTy,
-                                 const Type *AllocTy, Value *ArraySize,
-                                 Function *MallocF, const Twine &NameStr) {
+                                 const Type *AllocTy, Value *AllocSize, 
+                                 Value *ArraySize, Function *MallocF,
+                                 const Twine &Name) {
   assert(((!InsertBefore && InsertAtEnd) || (InsertBefore && !InsertAtEnd)) &&
          "createMalloc needs either InsertBefore or InsertAtEnd");
 
@@ -471,10 +461,14 @@
   //       bitcast (i8* malloc(typeSize)) to type*
   // malloc(type, arraySize) becomes:
   //       bitcast (i8 *malloc(typeSize*arraySize)) to type*
-  Value *AllocSize = ConstantExpr::getSizeOf(AllocTy);
-  AllocSize = ConstantExpr::getTruncOrBitCast(cast<Constant>(AllocSize),
-                                              IntPtrTy);
-  ArraySize = checkArraySize(ArraySize, IntPtrTy);
+  if (!ArraySize)
+    ArraySize = ConstantInt::get(IntPtrTy, 1);
+  else if (ArraySize->getType() != IntPtrTy) {
+    if (InsertBefore)
+      ArraySize = CastInst::CreateIntegerCast(ArraySize, IntPtrTy, false, "", InsertBefore);
+    else
+      ArraySize = CastInst::CreateIntegerCast(ArraySize, IntPtrTy, false, "", InsertAtEnd);
+  }
 
   if (!IsConstantOne(ArraySize)) {
     if (IsConstantOne(AllocSize)) {
@@ -513,14 +507,14 @@
     Result = MCall;
     if (Result->getType() != AllocPtrType)
       // Create a cast instruction to convert to the right type...
-      Result = new BitCastInst(MCall, AllocPtrType, NameStr, InsertBefore);
+      Result = new BitCastInst(MCall, AllocPtrType, Name, InsertBefore);
   } else {
     MCall = CallInst::Create(MallocF, AllocSize, "malloccall");
     Result = MCall;
     if (Result->getType() != AllocPtrType) {
       InsertAtEnd->getInstList().push_back(MCall);
       // Create a cast instruction to convert to the right type...
-      Result = new BitCastInst(MCall, AllocPtrType, NameStr);
+      Result = new BitCastInst(MCall, AllocPtrType, Name);
     }
   }
   MCall->setTailCall();
@@ -538,8 +532,9 @@
 /// 3. Bitcast the result of the malloc call to the specified type.
 Instruction *CallInst::CreateMalloc(Instruction *InsertBefore,
                                     const Type *IntPtrTy, const Type *AllocTy,
-                                    Value *ArraySize, const Twine &Name) {
-  return createMalloc(InsertBefore, NULL, IntPtrTy, AllocTy, 
+                                    Value *AllocSize, Value *ArraySize,
+                                    const Twine &Name) {
+  return createMalloc(InsertBefore, NULL, IntPtrTy, AllocTy, AllocSize,
                       ArraySize, NULL, Name);
 }
 
@@ -553,9 +548,9 @@
 /// responsibility of the caller.
 Instruction *CallInst::CreateMalloc(BasicBlock *InsertAtEnd,
                                     const Type *IntPtrTy, const Type *AllocTy,
-                                    Value *ArraySize, Function* MallocF,
-                                    const Twine &Name) {
-  return createMalloc(NULL, InsertAtEnd, IntPtrTy, AllocTy,
+                                    Value *AllocSize, Value *ArraySize, 
+                                    Function *MallocF, const Twine &Name) {
+  return createMalloc(NULL, InsertAtEnd, IntPtrTy, AllocTy, AllocSize,
                       ArraySize, MallocF, Name);
 }
 
diff --git a/test/Analysis/PointerTracking/sizes.ll b/test/Analysis/PointerTracking/sizes.ll
index c0b0606..267c3b8 100644
--- a/test/Analysis/PointerTracking/sizes.ll
+++ b/test/Analysis/PointerTracking/sizes.ll
@@ -31,6 +31,7 @@
 }
 
 declare i32 @bar(i8*)
+declare i32 @bar2(i64*)
 
 define i32 @foo1(i32 %n) nounwind {
 entry:
@@ -60,11 +61,16 @@
 	ret i32 %add16
 }
 
-define i32 @foo2(i32 %n) nounwind {
+define i32 @foo2(i64 %n) nounwind {
 entry:
-	%call = malloc i8, i32 %n		; <i8*> [#uses=1]
+	%call = tail call i8* @malloc(i64 %n)  ; <i8*> [#uses=1]
 ; CHECK: %call =
 ; CHECK: ==> %n elements, %n bytes allocated
+	%mallocsize = mul i64 %n, 8                     ; <i64> [#uses=1]
+	%malloccall = tail call i8* @malloc(i64 %mallocsize) ; <i8*> [#uses=1]
+	%call3 = bitcast i8* %malloccall to i64*        ; <i64*> [#uses=1]
+; CHECK: %malloccall =
+; CHECK: ==> (8 * %n) elements, (8 * %n) bytes allocated
 	%call2 = tail call i8* @calloc(i64 2, i64 4) nounwind		; <i8*> [#uses=1]
 ; CHECK: %call2 =
 ; CHECK: ==> 8 elements, 8 bytes allocated
@@ -72,13 +78,17 @@
 ; CHECK: %call4 =
 ; CHECK: ==> 16 elements, 16 bytes allocated
 	%call6 = tail call i32 @bar(i8* %call) nounwind		; <i32> [#uses=1]
+	%call7 = tail call i32 @bar2(i64* %call3) nounwind ; <i32> [#uses=1]
 	%call8 = tail call i32 @bar(i8* %call2) nounwind		; <i32> [#uses=1]
 	%call10 = tail call i32 @bar(i8* %call4) nounwind		; <i32> [#uses=1]
-	%add = add i32 %call8, %call6		; <i32> [#uses=1]
-	%add11 = add i32 %add, %call10		; <i32> [#uses=1]
+	%add = add i32 %call8, %call6                   ; <i32> [#uses=1]
+	%add10 = add i32 %add, %call7                   ; <i32> [#uses=1]
+	%add11 = add i32 %add10, %call10                ; <i32> [#uses=1]
 	ret i32 %add11
 }
 
+declare noalias i8* @malloc(i64) nounwind
+
 declare noalias i8* @calloc(i64, i64) nounwind
 
 declare noalias i8* @realloc(i8* nocapture, i64) nounwind
diff --git a/test/Transforms/GlobalOpt/2009-06-01-RecursivePHI.ll b/test/Transforms/GlobalOpt/2009-06-01-RecursivePHI.ll
index abd3109..d3c3ff5 100644
--- a/test/Transforms/GlobalOpt/2009-06-01-RecursivePHI.ll
+++ b/test/Transforms/GlobalOpt/2009-06-01-RecursivePHI.ll
@@ -1,4 +1,5 @@
 ; RUN: opt < %s -globalopt
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 	%struct.s_annealing_sched = type { i32, float, float, float, float }
 	%struct.s_bb = type { i32, i32, i32, i32 }
@@ -96,7 +97,9 @@
 	unreachable
 
 bb1.i38:		; preds = %bb
-	%0 = malloc %struct.s_net, i32 undef		; <%struct.s_net*> [#uses=1]
+	%mallocsize = mul i64 28, undef                  ; <i64> [#uses=1]
+	%malloccall = tail call i8* @malloc(i64 %mallocsize)      ; <i8*> [#uses=1]
+	%0 = bitcast i8* %malloccall to %struct.s_net*  ; <%struct.s_net*> [#uses=1]
 	br i1 undef, label %bb.i1.i39, label %my_malloc.exit2.i
 
 bb.i1.i39:		; preds = %bb1.i38
@@ -115,3 +118,5 @@
 bb7:		; preds = %bb6.preheader
 	unreachable
 }
+
+declare noalias i8* @malloc(i64)
diff --git a/test/Transforms/GlobalOpt/heap-sra-1.ll b/test/Transforms/GlobalOpt/heap-sra-1.ll
index 6df559e..9d5148f 100644
--- a/test/Transforms/GlobalOpt/heap-sra-1.ll
+++ b/test/Transforms/GlobalOpt/heap-sra-1.ll
@@ -1,18 +1,22 @@
-; RUN: opt < %s -globalopt -S | grep {@X.f0}
-; RUN: opt < %s -globalopt -S | grep {@X.f1}
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i386-apple-darwin7"
+; RUN: opt < %s -globalopt -S | FileCheck %s
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 	%struct.foo = type { i32, i32 }
 @X = internal global %struct.foo* null
+; CHECK: @X.f0
+; CHECK: @X.f1
 
-define void @bar(i32 %Size) nounwind noinline {
+define void @bar(i64 %Size) nounwind noinline {
 entry:
-	%.sub = malloc %struct.foo, i32 %Size	
+  %mallocsize = mul i64 %Size, 8                  ; <i64> [#uses=1]
+  %malloccall = tail call i8* @malloc(i64 %mallocsize) ; <i8*> [#uses=1]
+  %.sub = bitcast i8* %malloccall to %struct.foo* ; <%struct.foo*> [#uses=1]
 	store %struct.foo* %.sub, %struct.foo** @X, align 4
 	ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @baz() nounwind readonly noinline {
 bb1.thread:
 	%0 = load %struct.foo** @X, align 4		
diff --git a/test/Transforms/GlobalOpt/heap-sra-2.ll b/test/Transforms/GlobalOpt/heap-sra-2.ll
index 5a3c3cd..fa8c362 100644
--- a/test/Transforms/GlobalOpt/heap-sra-2.ll
+++ b/test/Transforms/GlobalOpt/heap-sra-2.ll
@@ -1,20 +1,22 @@
-; RUN: opt < %s -globalopt -S | grep {@X.f0}
-; RUN: opt < %s -globalopt -S | grep {@X.f1}
+; RUN: opt < %s -globalopt -S | FileCheck %s
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i386-apple-darwin7"
 	%struct.foo = type { i32, i32 }
 @X = internal global %struct.foo* null		; <%struct.foo**> [#uses=2]
+; CHECK: @X.f0
+; CHECK: @X.f1
 
 define void @bar(i32 %Size) nounwind noinline {
 entry:
-	%0 = malloc [1000000 x %struct.foo]
-        ;%.sub = bitcast [1000000 x %struct.foo]* %0 to %struct.foo*
+	%malloccall = tail call i8* @malloc(i64 8000000) ; <i8*> [#uses=1]
+	%0 = bitcast i8* %malloccall to [1000000 x %struct.foo]* ; <[1000000 x %struct.foo]*> [#uses=1]
 	%.sub = getelementptr [1000000 x %struct.foo]* %0, i32 0, i32 0		; <%struct.foo*> [#uses=1]
 	store %struct.foo* %.sub, %struct.foo** @X, align 4
 	ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @baz() nounwind readonly noinline {
 bb1.thread:
 	%0 = load %struct.foo** @X, align 4		; <%struct.foo*> [#uses=1]
diff --git a/test/Transforms/GlobalOpt/heap-sra-3.ll b/test/Transforms/GlobalOpt/heap-sra-3.ll
index 1496485..cbbcdfc 100644
--- a/test/Transforms/GlobalOpt/heap-sra-3.ll
+++ b/test/Transforms/GlobalOpt/heap-sra-3.ll
@@ -1,24 +1,22 @@
 ; RUN: opt < %s -globalopt -S | FileCheck %s
-
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i386-apple-darwin10"
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 	%struct.foo = type { i32, i32 }
 @X = internal global %struct.foo* null
 ; CHECK: @X.f0
 ; CHECK: @X.f1
 
-define void @bar(i32 %Size) nounwind noinline {
+define void @bar(i64 %Size) nounwind noinline {
 entry:
-  %mallocsize = mul i32 ptrtoint (%struct.foo* getelementptr (%struct.foo* null, i32 1) to i32), %Size, ; <i32> [#uses=1]
-; CHECK: mul i32 %Size
-  %malloccall = tail call i8* @malloc(i32 %mallocsize) ; <i8*> [#uses=1]
+  %mallocsize = mul i64 8, %Size, ; <i64> [#uses=1]
+; CHECK: mul i64 %Size, 4
+  %malloccall = tail call i8* @malloc(i64 %mallocsize) ; <i8*> [#uses=1]
   %.sub = bitcast i8* %malloccall to %struct.foo* ; <%struct.foo*> [#uses=1]
 	store %struct.foo* %.sub, %struct.foo** @X, align 4
 	ret void
 }
 
-declare noalias i8* @malloc(i32)
+declare noalias i8* @malloc(i64)
 
 define i32 @baz() nounwind readonly noinline {
 bb1.thread:
diff --git a/test/Transforms/GlobalOpt/heap-sra-4.ll b/test/Transforms/GlobalOpt/heap-sra-4.ll
index ae97ef1..d5a5828 100644
--- a/test/Transforms/GlobalOpt/heap-sra-4.ll
+++ b/test/Transforms/GlobalOpt/heap-sra-4.ll
@@ -1,24 +1,22 @@
 ; RUN: opt < %s -globalopt -S | FileCheck %s
-
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i386-apple-darwin7"
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 	%struct.foo = type { i32, i32 }
 @X = internal global %struct.foo* null
 ; CHECK: @X.f0
 ; CHECK: @X.f1
 
-define void @bar(i32 %Size) nounwind noinline {
+define void @bar(i64 %Size) nounwind noinline {
 entry:
-  %mallocsize = shl i32 ptrtoint (%struct.foo* getelementptr (%struct.foo* null, i32 1) to i32), 9, ; <i32> [#uses=1]
-  %malloccall = tail call i8* @malloc(i32 %mallocsize) ; <i8*> [#uses=1]
-; CHECK: @malloc(i32 mul (i32 512
+  %mallocsize = shl i64 %Size, 3                  ; <i64> [#uses=1]
+  %malloccall = tail call i8* @malloc(i64 %mallocsize) ; <i8*> [#uses=1]
+; CHECK: mul i64 %Size, 4
   %.sub = bitcast i8* %malloccall to %struct.foo* ; <%struct.foo*> [#uses=1]
 	store %struct.foo* %.sub, %struct.foo** @X, align 4
 	ret void
 }
 
-declare noalias i8* @malloc(i32)
+declare noalias i8* @malloc(i64)
 
 define i32 @baz() nounwind readonly noinline {
 bb1.thread:
diff --git a/test/Transforms/GlobalOpt/heap-sra-phi.ll b/test/Transforms/GlobalOpt/heap-sra-phi.ll
index 2eba944..6188e5a 100644
--- a/test/Transforms/GlobalOpt/heap-sra-phi.ll
+++ b/test/Transforms/GlobalOpt/heap-sra-phi.ll
@@ -1,19 +1,21 @@
 ; RUN: opt < %s -globalopt -S | grep {tmp.f1 = phi i32. }
 ; RUN: opt < %s -globalopt -S | grep {tmp.f0 = phi i32. }
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i386-apple-darwin7"
 	%struct.foo = type { i32, i32 }
 @X = internal global %struct.foo* null		; <%struct.foo**> [#uses=2]
 
 define void @bar(i32 %Size) nounwind noinline {
 entry:
-	%tmp = malloc [1000000 x %struct.foo]		; <[1000000 x %struct.foo]*> [#uses=1]
+	%malloccall = tail call i8* @malloc(i64 8000000) ; <i8*> [#uses=1]
+	%tmp = bitcast i8* %malloccall to [1000000 x %struct.foo]* ; <[1000000 x %struct.foo]*> [#uses=1]
 	%.sub = getelementptr [1000000 x %struct.foo]* %tmp, i32 0, i32 0		; <%struct.foo*> [#uses=1]
 	store %struct.foo* %.sub, %struct.foo** @X, align 4
 	ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @baz() nounwind readonly noinline {
 bb1.thread:
 	%tmpLD1 = load %struct.foo** @X, align 4		; <%struct.foo*> [#uses=1]
diff --git a/test/Transforms/GlobalOpt/malloc-promote-1.ll b/test/Transforms/GlobalOpt/malloc-promote-1.ll
index fd510e3..51ccbbd 100644
--- a/test/Transforms/GlobalOpt/malloc-promote-1.ll
+++ b/test/Transforms/GlobalOpt/malloc-promote-1.ll
@@ -1,19 +1,24 @@
-; RUN: opt < %s -globalopt -S | not grep global
+; RUN: opt < %s -globalopt -S | FileCheck %s
 target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 @G = internal global i32* null          ; <i32**> [#uses=3]
+; CHECK-NOT: global
 
 define void @init() {
-        %P = malloc i32         ; <i32*> [#uses=1]
+        %malloccall = tail call i8* @malloc(i64 4)      ; <i8*> [#uses=1]
+        %P = bitcast i8* %malloccall to i32*            ; <i32*> [#uses=1]
         store i32* %P, i32** @G
         %GV = load i32** @G             ; <i32*> [#uses=1]
         store i32 0, i32* %GV
         ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @get() {
         %GV = load i32** @G             ; <i32*> [#uses=1]
         %V = load i32* %GV              ; <i32> [#uses=1]
         ret i32 %V
+; CHECK: ret i32 0
 }
 
diff --git a/test/Transforms/GlobalOpt/malloc-promote-2.ll b/test/Transforms/GlobalOpt/malloc-promote-2.ll
index d3d2252..f989b79 100644
--- a/test/Transforms/GlobalOpt/malloc-promote-2.ll
+++ b/test/Transforms/GlobalOpt/malloc-promote-2.ll
@@ -1,11 +1,11 @@
 ; RUN: opt < %s -globalopt -globaldce -S | not grep malloc
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i686-apple-darwin8"
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 @G = internal global i32* null          ; <i32**> [#uses=3]
 
 define void @init() {
-        %P = malloc i32, i32 100                ; <i32*> [#uses=1]
+        %malloccall = tail call i8* @malloc(i64 mul (i64 100, i64 4)) ; <i8*> [#uses=1]
+        %P = bitcast i8* %malloccall to i32*            ; <i32*> [#uses=1]
         store i32* %P, i32** @G
         %GV = load i32** @G             ; <i32*> [#uses=1]
         %GVe = getelementptr i32* %GV, i32 40           ; <i32*> [#uses=1]
@@ -13,6 +13,8 @@
         ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @get() {
         %GV = load i32** @G             ; <i32*> [#uses=1]
         %GVe = getelementptr i32* %GV, i32 40           ; <i32*> [#uses=1]
diff --git a/test/Transforms/GlobalOpt/malloc-promote-3.ll b/test/Transforms/GlobalOpt/malloc-promote-3.ll
index a920b61..57f937d 100644
--- a/test/Transforms/GlobalOpt/malloc-promote-3.ll
+++ b/test/Transforms/GlobalOpt/malloc-promote-3.ll
@@ -1,11 +1,11 @@
 ; RUN: opt < %s -globalopt -globaldce -S | not grep malloc
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i686-apple-darwin8"
+target datalayout = "E-p:64:64:64-a0:0:8-f32:32:32-f64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-v64:64:64-v128:128:128"
 
 @G = internal global i32* null          ; <i32**> [#uses=4]
 
 define void @init() {
-        %P = malloc i32, i32 100                ; <i32*> [#uses=1]
+        %malloccall = tail call i8* @malloc(i64 mul (i64 100, i64 4)) ; <i8*> [#uses=1]
+        %P = bitcast i8* %malloccall to i32*            ; <i32*> [#uses=1]
         store i32* %P, i32** @G
         %GV = load i32** @G             ; <i32*> [#uses=1]
         %GVe = getelementptr i32* %GV, i32 40           ; <i32*> [#uses=1]
@@ -13,6 +13,8 @@
         ret void
 }
 
+declare noalias i8* @malloc(i64)
+
 define i32 @get() {
         %GV = load i32** @G             ; <i32*> [#uses=1]
         %GVe = getelementptr i32* %GV, i32 40           ; <i32*> [#uses=1]