refactor the MemoryBuiltin analysis:

 - provide more extensive set of functions to detect library allocation functions (e.g., malloc, calloc, strdup, etc)
 - provide an API to compute the size and offset of an object pointed by

Move a few clients (GVN, AA, instcombine, ...) to the new API.
This implementation is a lot more aggressive than each of the custom implementations being replaced.

Patch reviewed by Nick Lewycky and Chandler Carruth, thanks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158919 91177308-0d34-0410-b5e6-96231b3b80d8

11 files changed, 717 insertions, 290 deletions

diff --git a/include/llvm/Analysis/MemoryBuiltins.h b/include/llvm/Analysis/MemoryBuiltins.h
index e7dcbf3..ab4a3f3 100644
--- a/include/llvm/Analysis/MemoryBuiltins.h
+++ b/include/llvm/Analysis/MemoryBuiltins.h
@@ -15,6 +15,14 @@
 #ifndef LLVM_ANALYSIS_MEMORYBUILTINS_H
 #define LLVM_ANALYSIS_MEMORYBUILTINS_H
 
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/Operator.h"
+#include "llvm/Support/DataTypes.h"
+#include "llvm/Support/InstVisitor.h"
+#include "llvm/Support/IRBuilder.h"
+#include "llvm/Support/TargetFolder.h"
+
 namespace llvm {
 class CallInst;
 class PointerType;
@@ -22,24 +30,50 @@ class TargetData;
 class Type;
 class Value;
 
+
+/// \brief Tests if a value is a call to a library function that allocates or
+/// reallocates memory (either malloc, calloc, realloc, or strdup like).
+bool isAllocationFn(const Value *V, bool LookThroughBitCast = false);
+
+/// \brief Tests if a value is a call to a function that returns a NoAlias
+/// pointer (including malloc/calloc/strdup-like functions).
+bool isNoAliasFn(const Value *V, bool LookThroughBitCast = false);
+
+/// \brief Tests if a value is a call to a library function that allocates
+/// uninitialized memory (such as malloc).
+bool isMallocLikeFn(const Value *V, bool LookThroughBitCast = false);
+
+/// \brief Tests if a value is a call to a library function that allocates
+/// zero-filled memory (such as calloc).
+bool isCallocLikeFn(const Value *V, bool LookThroughBitCast = false);
+
+/// \brief Tests if a value is a call to a library function that allocates
+/// memory (either malloc, calloc, or strdup like).
+bool isAllocLikeFn(const Value *V, bool LookThroughBitCast = false);
+
+/// \brief Tests if a value is a call to a library function that reallocates
+/// memory (such as realloc).
+bool isReallocLikeFn(const Value *V, bool LookThroughBitCast = false);
+
+
 //===----------------------------------------------------------------------===//
 //  malloc Call Utility Functions.
 //
 
-/// isMalloc - Returns true if the value is either a malloc call or a bitcast of 
-/// the result of a malloc call
-bool isMalloc(const Value *I);
-
 /// extractMallocCall - Returns the corresponding CallInst if the instruction
 /// is a malloc call.  Since CallInst::CreateMalloc() only creates calls, we
 /// ignore InvokeInst here.
 const CallInst *extractMallocCall(const Value *I);
-CallInst *extractMallocCall(Value *I);
+static inline CallInst *extractMallocCall(Value *I) {
+  return const_cast<CallInst*>(extractMallocCall((const Value*)I));
+}
 
 /// extractMallocCallFromBitCast - Returns the corresponding CallInst if the
 /// instruction is a bitcast of the result of a malloc call.
 const CallInst *extractMallocCallFromBitCast(const Value *I);
-CallInst *extractMallocCallFromBitCast(Value *I);
+static inline CallInst *extractMallocCallFromBitCast(Value *I) {
+  return const_cast<CallInst*>(extractMallocCallFromBitCast((const Value*)I));
+}
 
 /// isArrayMalloc - Returns the corresponding CallInst if the instruction 
 /// is a call to malloc whose array size can be determined and the array size
@@ -67,7 +101,7 @@ Type *getMallocAllocatedType(const CallInst *CI);
 /// determined.
 Value *getMallocArraySize(CallInst *CI, const TargetData *TD,
                           bool LookThroughSExt = false);
-                          
+
 
 //===----------------------------------------------------------------------===//
 //  calloc Call Utility Functions.
@@ -76,7 +110,9 @@ Value *getMallocArraySize(CallInst *CI, const TargetData *TD,
 /// extractCallocCall - Returns the corresponding CallInst if the instruction
 /// is a calloc call.
 const CallInst *extractCallocCall(const Value *I);
-CallInst *extractCallocCall(Value *I);
+static inline CallInst *extractCallocCall(Value *I) {
+  return const_cast<CallInst*>(extractCallocCall((const Value*)I));
+}
 
 
 //===----------------------------------------------------------------------===//
@@ -90,6 +126,126 @@ static inline CallInst *isFreeCall(Value *I) {
   return const_cast<CallInst*>(isFreeCall((const Value*)I));
 }
 
+  
+//===----------------------------------------------------------------------===//
+//  Utility functions to compute size of objects.
+//
+
+/// \brief Compute the size of the object pointed by Ptr. Returns true and the
+/// object size in Size if successful, and false otherwise.
+/// If RoundToAlign is true, then Size is rounded up to the aligment of allocas,
+/// byval arguments, and global variables.
+bool getObjectSize(const Value *Ptr, uint64_t &Size, const TargetData *TD,
+                   bool RoundToAlign = false);
+
+
+
+typedef std::pair<APInt, APInt> SizeOffsetType;
+
+/// \brief Evaluate the size and offset of an object ponted by a Value*
+/// statically. Fails if size or offset are not known at compile time.
+class ObjectSizeOffsetVisitor
+  : public InstVisitor<ObjectSizeOffsetVisitor, SizeOffsetType> {
+
+  const TargetData *TD;
+  bool RoundToAlign;
+  unsigned IntTyBits;
+  APInt Zero;
+
+  APInt align(APInt Size, uint64_t Align);
+
+  SizeOffsetType unknown() {
+    return std::make_pair(APInt(), APInt());
+  }
+
+public:
+  ObjectSizeOffsetVisitor(const TargetData *TD, LLVMContext &Context,
+                          bool RoundToAlign = false);
+
+  SizeOffsetType compute(Value *V);
+
+  bool knownSize(SizeOffsetType &SizeOffset) {
+    return SizeOffset.first.getBitWidth() > 1;
+  }
+
+  bool knownOffset(SizeOffsetType &SizeOffset) {
+    return SizeOffset.second.getBitWidth() > 1;
+  }
+
+  bool bothKnown(SizeOffsetType &SizeOffset) {
+    return knownSize(SizeOffset) && knownOffset(SizeOffset);
+  }
+
+  SizeOffsetType visitAllocaInst(AllocaInst &I);
+  SizeOffsetType visitArgument(Argument &A);
+  SizeOffsetType visitCallSite(CallSite CS);
+  SizeOffsetType visitConstantPointerNull(ConstantPointerNull&);
+  SizeOffsetType visitExtractValueInst(ExtractValueInst &I);
+  SizeOffsetType visitGEPOperator(GEPOperator &GEP);
+  SizeOffsetType visitGlobalVariable(GlobalVariable &GV);
+  SizeOffsetType visitIntToPtrInst(IntToPtrInst&);
+  SizeOffsetType visitLoadInst(LoadInst &I);
+  SizeOffsetType visitPHINode(PHINode&);
+  SizeOffsetType visitSelectInst(SelectInst &I);
+  SizeOffsetType visitUndefValue(UndefValue&);
+  SizeOffsetType visitInstruction(Instruction &I);
+};
+
+typedef std::pair<Value*, Value*> SizeOffsetEvalType;
+typedef IRBuilder<true, TargetFolder> BuilderTy;
+typedef DenseMap<const Value*, SizeOffsetEvalType> CacheMapTy;
+typedef SmallPtrSet<const Value*, 8> PtrSetTy;
+
+
+/// \brief Evaluate the size and offset of an object ponted by a Value*.
+/// May create code to compute the result at run-time.
+class ObjectSizeOffsetEvaluator
+  : public InstVisitor<ObjectSizeOffsetEvaluator, SizeOffsetEvalType> {
+
+  const TargetData *TD;
+  LLVMContext &Context;
+  BuilderTy Builder;
+  ObjectSizeOffsetVisitor Visitor;
+  IntegerType *IntTy;
+  Value *Zero;
+  CacheMapTy CacheMap;
+  PtrSetTy SeenVals;
+
+  SizeOffsetEvalType unknown() {
+    return std::make_pair((Value*)0, (Value*)0);
+  }
+  SizeOffsetEvalType compute_(Value *V);
+
+public:
+  ObjectSizeOffsetEvaluator(const TargetData *TD, LLVMContext &Context);
+  SizeOffsetEvalType compute(Value *V);
+
+  bool knownSize(SizeOffsetEvalType &SizeOffset) {
+    return SizeOffset.first;
+  }
+
+  bool knownOffset(SizeOffsetEvalType &SizeOffset) {
+    return SizeOffset.second;
+  }
+
+  bool anyKnown(SizeOffsetEvalType &SizeOffset) {
+    return knownSize(SizeOffset) || knownOffset(SizeOffset);
+  }
+
+  bool bothKnown(SizeOffsetEvalType &SizeOffset) {
+    return knownSize(SizeOffset) && knownOffset(SizeOffset);
+  }
+
+  SizeOffsetEvalType visitAllocaInst(AllocaInst &I);
+  SizeOffsetEvalType visitCallSite(CallSite CS);
+  SizeOffsetEvalType visitGEPOperator(GEPOperator &GEP);
+  SizeOffsetEvalType visitIntToPtrInst(IntToPtrInst&);
+  SizeOffsetEvalType visitLoadInst(LoadInst &I);
+  SizeOffsetEvalType visitPHINode(PHINode &PHI);
+  SizeOffsetEvalType visitSelectInst(SelectInst &I);
+  SizeOffsetEvalType visitInstruction(Instruction &I);
+};
+
 } // End llvm namespace
 
 #endif
diff --git a/lib/Analysis/BasicAliasAnalysis.cpp b/lib/Analysis/BasicAliasAnalysis.cpp
index 20ecfd2..1d028c2 100644
--- a/lib/Analysis/BasicAliasAnalysis.cpp
+++ b/lib/Analysis/BasicAliasAnalysis.cpp
@@ -86,47 +86,10 @@ static bool isEscapeSource(const Value *V) {
 /// UnknownSize if unknown.
 static uint64_t getObjectSize(const Value *V, const TargetData &TD,
                               bool RoundToAlign = false) {
-  Type *AccessTy;
-  unsigned Align;
-  if (const GlobalVariable *GV = dyn_cast<GlobalVariable>(V)) {
-    if (!GV->hasDefinitiveInitializer())
-      return AliasAnalysis::UnknownSize;
-    AccessTy = GV->getType()->getElementType();
-    Align = GV->getAlignment();
-  } else if (const AllocaInst *AI = dyn_cast<AllocaInst>(V)) {
-    if (!AI->isArrayAllocation())
-      AccessTy = AI->getType()->getElementType();
-    else
-      return AliasAnalysis::UnknownSize;
-    Align = AI->getAlignment();
-  } else if (const CallInst* CI = extractMallocCall(V)) {
-    if (!RoundToAlign && !isArrayMalloc(V, &TD))
-      // The size is the argument to the malloc call.
-      if (const ConstantInt* C = dyn_cast<ConstantInt>(CI->getArgOperand(0)))
-        return C->getZExtValue();
-    return AliasAnalysis::UnknownSize;
-  } else if (const Argument *A = dyn_cast<Argument>(V)) {
-    if (A->hasByValAttr()) {
-      AccessTy = cast<PointerType>(A->getType())->getElementType();
-      Align = A->getParamAlignment();
-    } else {
-      return AliasAnalysis::UnknownSize;
-    }
-  } else {
-    return AliasAnalysis::UnknownSize;
-  }
-
-  if (!AccessTy->isSized())
-    return AliasAnalysis::UnknownSize;
-
-  uint64_t Size = TD.getTypeAllocSize(AccessTy);
-  // If there is an explicitly specified alignment, and we need to
-  // take alignment into account, round up the size. (If the alignment
-  // is implicit, getTypeAllocSize is sufficient.)
-  if (RoundToAlign && Align)
-    Size = RoundUpToAlignment(Size, Align);
-
-  return Size;
+  uint64_t Size;
+  if (getObjectSize(V, Size, &TD, RoundToAlign))
+    return Size;
+  return AliasAnalysis::UnknownSize;
 }
 
 /// isObjectSmallerThan - Return true if we can prove that the object specified
diff --git a/lib/Analysis/IPA/GlobalsModRef.cpp b/lib/Analysis/IPA/GlobalsModRef.cpp
index c1d8e3e..22f6e96 100644
--- a/lib/Analysis/IPA/GlobalsModRef.cpp
+++ b/lib/Analysis/IPA/GlobalsModRef.cpp
@@ -329,15 +329,8 @@ bool GlobalsModRef::AnalyzeIndirectGlobalMemory(GlobalValue *GV) {
       // Check the value being stored.
       Value *Ptr = GetUnderlyingObject(SI->getOperand(0));
 
-      if (isMalloc(Ptr)) {
-        // Okay, easy case.
-      } else if (CallInst *CI = dyn_cast<CallInst>(Ptr)) {
-        Function *F = CI->getCalledFunction();
-        if (!F || !F->isDeclaration()) return false;     // Too hard to analyze.
-        if (F->getName() != "calloc") return false;   // Not calloc.
-      } else {
+      if (!isAllocLikeFn(Ptr))
         return false;  // Too hard to analyze.
-      }
 
       // Analyze all uses of the allocation.  If any of them are used in a
       // non-simple way (e.g. stored to another global) bail out.
@@ -454,19 +447,18 @@ void GlobalsModRef::AnalyzeCallGraph(CallGraph &CG, Module &M) {
       for (inst_iterator II = inst_begin(SCC[i]->getFunction()),
              E = inst_end(SCC[i]->getFunction());
            II != E && FunctionEffect != ModRef; ++II)
-        if (isa<LoadInst>(*II)) {
+        if (LoadInst *LI = dyn_cast<LoadInst>(&*II)) {
           FunctionEffect |= Ref;
-          if (cast<LoadInst>(*II).isVolatile())
+          if (LI->isVolatile())
             // Volatile loads may have side-effects, so mark them as writing
             // memory (for example, a flag inside the processor).
             FunctionEffect |= Mod;
-        } else if (isa<StoreInst>(*II)) {
+        } else if (StoreInst *SI = dyn_cast<StoreInst>(&*II)) {
           FunctionEffect |= Mod;
-          if (cast<StoreInst>(*II).isVolatile())
+          if (SI->isVolatile())
             // Treat volatile stores as reading memory somewhere.
             FunctionEffect |= Ref;
-        } else if (isMalloc(&cast<Instruction>(*II)) ||
-                   isFreeCall(&cast<Instruction>(*II))) {
+        } else if (isAllocationFn(&*II) || isFreeCall(&*II)) {
           FunctionEffect |= ModRef;
         } else if (IntrinsicInst *Intrinsic = dyn_cast<IntrinsicInst>(&*II)) {
           // The callgraph doesn't include intrinsic calls.
diff --git a/lib/Analysis/MemoryBuiltins.cpp b/lib/Analysis/MemoryBuiltins.cpp
index 9695ae1..2a1afdc 100644
--- a/lib/Analysis/MemoryBuiltins.cpp
+++ b/lib/Analysis/MemoryBuiltins.cpp
@@ -12,80 +12,165 @@
 //
 //===----------------------------------------------------------------------===//
 
+#define DEBUG_TYPE "memory-builtins"
+#include "llvm/ADT/Statistic.h"
+#include "llvm/ADT/STLExtras.h"
 #include "llvm/Analysis/MemoryBuiltins.h"
-#include "llvm/Constants.h"
+#include "llvm/GlobalVariable.h"
 #include "llvm/Instructions.h"
+#include "llvm/Intrinsics.h"
+#include "llvm/Metadata.h"
 #include "llvm/Module.h"
 #include "llvm/Analysis/ValueTracking.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/MathExtras.h"
+#include "llvm/Support/raw_ostream.h"
 #include "llvm/Target/TargetData.h"
+#include "llvm/Transforms/Utils/Local.h"
 using namespace llvm;
 
-//===----------------------------------------------------------------------===//
-//  malloc Call Utility Functions.
-//
+enum AllocType {
+  MallocLike         = 1<<0, // allocates
+  CallocLike         = 1<<1, // allocates + bzero
+  ReallocLike        = 1<<2, // reallocates
+  StrDupLike         = 1<<3,
+  AllocLike          = MallocLike | CallocLike | StrDupLike,
+  AnyAlloc           = MallocLike | CallocLike | ReallocLike | StrDupLike
+};
+
+struct AllocFnsTy {
+  const char *Name;
+  AllocType AllocTy;
+  unsigned char NumParams;
+  // First and Second size parameters (or -1 if unused)
+  unsigned char FstParam, SndParam;
+};
+
+static const AllocFnsTy AllocationFnData[] = {
+  {"malloc",         MallocLike,  1, 0,  -1},
+  {"valloc",         MallocLike,  1, 0,  -1},
+  {"_Znwj",          MallocLike,  1, 0,  -1}, // operator new(unsigned int)
+  {"_Znwm",          MallocLike,  1, 0,  -1}, // operator new(unsigned long)
+  {"_Znaj",          MallocLike,  1, 0,  -1}, // operator new[](unsigned int)
+  {"_Znam",          MallocLike,  1, 0,  -1}, // operator new[](unsigned long)
+  {"posix_memalign", MallocLike,  3, 2,  -1},
+  {"calloc",         CallocLike,  2, 0,  1},
+  {"realloc",        ReallocLike, 2, 1,  -1},
+  {"reallocf",       ReallocLike, 2, 1,  -1},
+  {"strdup",         StrDupLike,  1, -1, -1},
+  {"strndup",        StrDupLike,  2, -1, -1}
+};
+
+
+static Function *getCalledFunction(const Value *V, bool LookThroughBitCast) {
+  if (LookThroughBitCast)
+    V = V->stripPointerCasts();
+  const CallInst *CI = dyn_cast<CallInst>(V);
+  if (!CI)
+    return 0;
 
-/// isMalloc - Returns true if the value is either a malloc call or a
-/// bitcast of the result of a malloc call.
-bool llvm::isMalloc(const Value *I) {
-  return extractMallocCall(I) || extractMallocCallFromBitCast(I);
+  Function *Callee = CI->getCalledFunction();
+  if (!Callee || !Callee->isDeclaration())
+    return 0;
+  return Callee;
 }
 
-static bool isMallocCall(const CallInst *CI) {
-  if (!CI)
-    return false;
+/// \brief Returns the allocation data for the given value if it is a call to a
+/// known allocation function, and NULL otherwise.
+static const AllocFnsTy *getAllocationData(const Value *V, AllocType AllocTy,
+                                           bool LookThroughBitCast = false) {
+  Function *Callee = getCalledFunction(V, LookThroughBitCast);
+  if (!Callee)
+    return 0;
 
-  Function *Callee = CI->getCalledFunction();
-  if (Callee == 0 || !Callee->isDeclaration())
-    return false;
-  if (Callee->getName() != "malloc" &&
-      Callee->getName() != "_Znwj" && // operator new(unsigned int)
-      Callee->getName() != "_Znwm" && // operator new(unsigned long)
-      Callee->getName() != "_Znaj" && // operator new[](unsigned int)
-      Callee->getName() != "_Znam")   // operator new[](unsigned long)
-    return false;
+  unsigned i = 0;
+  bool found = false;
+  for ( ; i < array_lengthof(AllocationFnData); ++i) {
+    if (Callee->getName() == AllocationFnData[i].Name) {
+      found = true;
+      break;
+    }
+  }
+  if (!found)
+    return 0;
 
-  // Check malloc prototype.
-  // FIXME: workaround for PR5130, this will be obsolete when a nobuiltin 
-  // attribute will exist.
+  const AllocFnsTy *FnData = &AllocationFnData[i];
+  if ((FnData->AllocTy & AllocTy) == 0)
+    return 0;
+
+  // Check function prototype.
+  // FIXME: Check the nobuiltin metadata?? (PR5130)
+  unsigned FstParam = FnData->FstParam;
+  unsigned SndParam = FnData->SndParam;
   FunctionType *FTy = Callee->getFunctionType();
-  return FTy->getReturnType() == Type::getInt8PtrTy(FTy->getContext()) &&
-         FTy->getNumParams() == 1 &&
-         (FTy->getParamType(0)->isIntegerTy(32) ||
-          FTy->getParamType(0)->isIntegerTy(64));
+
+  if (FTy->getReturnType() == Type::getInt8PtrTy(FTy->getContext()) &&
+      FTy->getNumParams() == FnData->NumParams &&
+      (FstParam == (unsigned char)-1 ||
+       (FTy->getParamType(FstParam)->isIntegerTy(32) ||
+        FTy->getParamType(FstParam)->isIntegerTy(64))) &&
+      (SndParam == (unsigned char)-1 ||
+       FTy->getParamType(SndParam)->isIntegerTy(32) ||
+       FTy->getParamType(SndParam)->isIntegerTy(64)))
+    return FnData;
+  return 0;
 }
 
-/// extractMallocCall - Returns the corresponding CallInst if the instruction
-/// is a malloc call.  Since CallInst::CreateMalloc() only creates calls, we
-/// ignore InvokeInst here.
-const CallInst *llvm::extractMallocCall(const Value *I) {
-  const CallInst *CI = dyn_cast<CallInst>(I);
-  return (isMallocCall(CI)) ? CI : NULL;
+static bool hasNoAliasAttr(const Value *V, bool LookThroughBitCast) {
+  Function *Callee = getCalledFunction(V, LookThroughBitCast);
+  return Callee && Callee->hasFnAttr(Attribute::NoAlias);
 }
 
-CallInst *llvm::extractMallocCall(Value *I) {
-  CallInst *CI = dyn_cast<CallInst>(I);
-  return (isMallocCall(CI)) ? CI : NULL;
+
+/// \brief Tests if a value is a call to a library function that allocates or
+/// reallocates memory (either malloc, calloc, realloc, or strdup like).
+bool llvm::isAllocationFn(const Value *V, bool LookThroughBitCast) {
+  return getAllocationData(V, AnyAlloc, LookThroughBitCast);
 }
 
-static bool isBitCastOfMallocCall(const BitCastInst *BCI) {
-  if (!BCI)
-    return false;
-    
-  return isMallocCall(dyn_cast<CallInst>(BCI->getOperand(0)));
+/// \brief Tests if a value is a call to a function that returns a NoAlias
+/// pointer (including malloc/calloc/strdup-like functions).
+bool llvm::isNoAliasFn(const Value *V, bool LookThroughBitCast) {
+  return isAllocLikeFn(V, LookThroughBitCast) ||
+         hasNoAliasAttr(V, LookThroughBitCast);
 }
 
-/// extractMallocCallFromBitCast - Returns the corresponding CallInst if the
-/// instruction is a bitcast of the result of a malloc call.
-CallInst *llvm::extractMallocCallFromBitCast(Value *I) {
-  BitCastInst *BCI = dyn_cast<BitCastInst>(I);
-  return (isBitCastOfMallocCall(BCI)) ? cast<CallInst>(BCI->getOperand(0))
-                                      : NULL;
+/// \brief Tests if a value is a call to a library function that allocates
+/// uninitialized memory (such as malloc).
+bool llvm::isMallocLikeFn(const Value *V, bool LookThroughBitCast) {
+  return getAllocationData(V, MallocLike, LookThroughBitCast);
+}
+
+/// \brief Tests if a value is a call to a library function that allocates
+/// zero-filled memory (such as calloc).
+bool llvm::isCallocLikeFn(const Value *V, bool LookThroughBitCast) {
+  return getAllocationData(V, CallocLike, LookThroughBitCast);
+}
+
+/// \brief Tests if a value is a call to a library function that allocates
+/// memory (either malloc, calloc, or strdup like).
+bool llvm::isAllocLikeFn(const Value *V, bool LookThroughBitCast) {
+  return getAllocationData(V, AllocLike, LookThroughBitCast);
+}
+
+/// \brief Tests if a value is a call to a library function that reallocates
+/// memory (such as realloc).
+bool llvm::isReallocLikeFn(const Value *V, bool LookThroughBitCast) {
+  return getAllocationData(V, ReallocLike, LookThroughBitCast);
+}
+
+/// extractMallocCall - Returns the corresponding CallInst if the instruction
+/// is a malloc call.  Since CallInst::CreateMalloc() only creates calls, we
+/// ignore InvokeInst here.
+const CallInst *llvm::extractMallocCall(const Value *I) {
+  return isMallocLikeFn(I) ? cast<CallInst>(I) : 0;
 }
 
+/// extractMallocCallFromBitCast - Returns the corresponding CallInst if the
+/// instruction is a bitcast of the result of a malloc call.
 const CallInst *llvm::extractMallocCallFromBitCast(const Value *I) {
   const BitCastInst *BCI = dyn_cast<BitCastInst>(I);
-  return (isBitCastOfMallocCall(BCI)) ? cast<CallInst>(BCI->getOperand(0))
-                                      : NULL;
+  return BCI ? extractMallocCall(BCI->getOperand(0)) : 0;
 }
 
 static Value *computeArraySize(const CallInst *CI, const TargetData *TD,
@@ -134,7 +219,7 @@ const CallInst *llvm::isArrayMalloc(const Value *I, const TargetData *TD) {
 ///   1: PointerType is the bitcast's result type.
 ///  >1: Unique PointerType cannot be determined, return NULL.
 PointerType *llvm::getMallocType(const CallInst *CI) {
-  assert(isMalloc(CI) && "getMallocType and not malloc call");
+  assert(isMallocLikeFn(CI) && "getMallocType and not malloc call");
   
   PointerType *MallocType = NULL;
   unsigned NumOfBitCastUses = 0;
@@ -176,53 +261,17 @@ Type *llvm::getMallocAllocatedType(const CallInst *CI) {
 /// determined.
 Value *llvm::getMallocArraySize(CallInst *CI, const TargetData *TD,
                                 bool LookThroughSExt) {
-  assert(isMalloc(CI) && "getMallocArraySize and not malloc call");
+  assert(isMallocLikeFn(CI) && "getMallocArraySize and not malloc call");
   return computeArraySize(CI, TD, LookThroughSExt);
 }
 
 
-//===----------------------------------------------------------------------===//
-//  calloc Call Utility Functions.
-//
-
-static bool isCallocCall(const CallInst *CI) {
-  if (!CI)
-    return false;
-  
-  Function *Callee = CI->getCalledFunction();
-  if (Callee == 0 || !Callee->isDeclaration())
-    return false;
-  if (Callee->getName() != "calloc")
-    return false;
-  
-  // Check malloc prototype.
-  // FIXME: workaround for PR5130, this will be obsolete when a nobuiltin 
-  // attribute exists.
-  FunctionType *FTy = Callee->getFunctionType();
-  return FTy->getReturnType() == Type::getInt8PtrTy(FTy->getContext()) &&
-  FTy->getNumParams() == 2 &&
-  ((FTy->getParamType(0)->isIntegerTy(32) &&
-    FTy->getParamType(1)->isIntegerTy(32)) ||
-   (FTy->getParamType(0)->isIntegerTy(64) &&
-    FTy->getParamType(1)->isIntegerTy(64)));
-}
-
 /// extractCallocCall - Returns the corresponding CallInst if the instruction
 /// is a calloc call.
 const CallInst *llvm::extractCallocCall(const Value *I) {
-  const CallInst *CI = dyn_cast<CallInst>(I);
-  return isCallocCall(CI) ? CI : 0;
+  return isCallocLikeFn(I) ? cast<CallInst>(I) : 0;
 }
 
-CallInst *llvm::extractCallocCall(Value *I) {
-  CallInst *CI = dyn_cast<CallInst>(I);
-  return isCallocCall(CI) ? CI : 0;
-}
-
-
-//===----------------------------------------------------------------------===//
-//  free Call Utility Functions.
-//
 
 /// isFreeCall - Returns non-null if the value is a call to the builtin free()
 const CallInst *llvm::isFreeCall(const Value *I) {
@@ -251,3 +300,388 @@ const CallInst *llvm::isFreeCall(const Value *I) {
 
   return CI;
 }
+
+
+
+//===----------------------------------------------------------------------===//
+//  Utility functions to compute size of objects.
+//
+
+
+/// \brief Compute the size of the object pointed by Ptr. Returns true and the
+/// object size in Size if successful, and false otherwise.
+/// If RoundToAlign is true, then Size is rounded up to the aligment of allocas,
+/// byval arguments, and global variables.
+bool llvm::getObjectSize(const Value *Ptr, uint64_t &Size, const TargetData *TD,
+                         bool RoundToAlign) {
+  if (!TD)
+    return false;
+
+  ObjectSizeOffsetVisitor Visitor(TD, Ptr->getContext(), RoundToAlign);
+  SizeOffsetType Data = Visitor.compute(const_cast<Value*>(Ptr));
+  if (!Visitor.bothKnown(Data))
+    return false;
+
+  APInt ObjSize = Data.first, Offset = Data.second;
+  // check for overflow
+  if (Offset.slt(0) || ObjSize.ult(Offset))
+    Size = 0;
+  else
+    Size = (ObjSize - Offset).getZExtValue();
+  return true;
+}
+
+
+STATISTIC(ObjectVisitorArgument,
+          "Number of arguments with unsolved size and offset");
+STATISTIC(ObjectVisitorLoad,
+          "Number of load instructions with unsolved size and offset");
+
+
+APInt ObjectSizeOffsetVisitor::align(APInt Size, uint64_t Align) {
+  if (RoundToAlign && Align)
+    return APInt(IntTyBits, RoundUpToAlignment(Size.getZExtValue(), Align));
+  return Size;
+}
+
+ObjectSizeOffsetVisitor::ObjectSizeOffsetVisitor(const TargetData *TD,
+                                                 LLVMContext &Context,
+                                                 bool RoundToAlign)
+: TD(TD), RoundToAlign(RoundToAlign) {
+  IntegerType *IntTy = TD->getIntPtrType(Context);
+  IntTyBits = IntTy->getBitWidth();
+  Zero = APInt::getNullValue(IntTyBits);
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::compute(Value *V) {
+  V = V->stripPointerCasts();
+
+  if (GEPOperator *GEP = dyn_cast<GEPOperator>(V))
+    return visitGEPOperator(*GEP);
+  if (Instruction *I = dyn_cast<Instruction>(V))
+    return visit(*I);
+  if (Argument *A = dyn_cast<Argument>(V))
+    return visitArgument(*A);
+  if (ConstantPointerNull *P = dyn_cast<ConstantPointerNull>(V))
+    return visitConstantPointerNull(*P);
+  if (GlobalVariable *GV = dyn_cast<GlobalVariable>(V))
+    return visitGlobalVariable(*GV);
+  if (UndefValue *UV = dyn_cast<UndefValue>(V))
+    return visitUndefValue(*UV);
+  if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
+    if (CE->getOpcode() == Instruction::IntToPtr)
+      return unknown(); // clueless
+
+  DEBUG(dbgs() << "ObjectSizeOffsetVisitor::compute() unhandled value: " << *V
+        << '\n');
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitAllocaInst(AllocaInst &I) {
+  if (!I.getAllocatedType()->isSized())
+    return unknown();
+
+  APInt Size(IntTyBits, TD->getTypeAllocSize(I.getAllocatedType()));
+  if (!I.isArrayAllocation())
+    return std::make_pair(align(Size, I.getAlignment()), Zero);
+
+  Value *ArraySize = I.getArraySize();
+  if (const ConstantInt *C = dyn_cast<ConstantInt>(ArraySize)) {
+    Size *= C->getValue().zextOrSelf(IntTyBits);
+    return std::make_pair(align(Size, I.getAlignment()), Zero);
+  }
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitArgument(Argument &A) {
+  // no interprocedural analysis is done at the moment
+  if (!A.hasByValAttr()) {
+    ++ObjectVisitorArgument;
+    return unknown();
+  }
+  PointerType *PT = cast<PointerType>(A.getType());
+  APInt Size(IntTyBits, TD->getTypeAllocSize(PT->getElementType()));
+  return std::make_pair(align(Size, A.getParamAlignment()), Zero);
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitCallSite(CallSite CS) {
+  const AllocFnsTy *FnData = getAllocationData(CS.getInstruction(), AnyAlloc);
+  if (!FnData)
+    return unknown();
+
+  // handle strdup-like functions separately
+  if (FnData->AllocTy == StrDupLike) {
+    // TODO
+    return unknown();
+  }
+
+  ConstantInt *Arg = dyn_cast<ConstantInt>(CS.getArgument(FnData->FstParam));
+  if (!Arg)
+    return unknown();
+
+  APInt Size = Arg->getValue();
+  // size determined by just 1 parameter
+  if (FnData->SndParam == (unsigned char)-1)
+    return std::make_pair(Size, Zero);
+
+  Arg = dyn_cast<ConstantInt>(CS.getArgument(FnData->SndParam));
+  if (!Arg)
+    return unknown();
+
+  Size *= Arg->getValue();
+  return std::make_pair(Size, Zero);
+
+  // TODO: handle more standard functions (+ wchar cousins):
+  // - strdup / strndup
+  // - strcpy / strncpy
+  // - strcat / strncat
+  // - memcpy / memmove
+  // - strcat / strncat
+  // - memset
+}
+
+SizeOffsetType
+ObjectSizeOffsetVisitor::visitConstantPointerNull(ConstantPointerNull&) {
+  return std::make_pair(Zero, Zero);
+}
+
+SizeOffsetType
+ObjectSizeOffsetVisitor::visitExtractValueInst(ExtractValueInst&) {
+  // Easy cases were already folded by previous passes.
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitGEPOperator(GEPOperator &GEP) {
+  SizeOffsetType PtrData = compute(GEP.getPointerOperand());
+  if (!bothKnown(PtrData) || !GEP.hasAllConstantIndices())
+    return unknown();
+
+  SmallVector<Value*, 8> Ops(GEP.idx_begin(), GEP.idx_end());
+  APInt Offset(IntTyBits,TD->getIndexedOffset(GEP.getPointerOperandType(),Ops));
+  return std::make_pair(PtrData.first, PtrData.second + Offset);
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitGlobalVariable(GlobalVariable &GV){
+  if (!GV.hasDefinitiveInitializer())
+    return unknown();
+
+  APInt Size(IntTyBits, TD->getTypeAllocSize(GV.getType()->getElementType()));
+  return std::make_pair(align(Size, GV.getAlignment()), Zero);
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitIntToPtrInst(IntToPtrInst&) {
+  // clueless
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitLoadInst(LoadInst&) {
+  ++ObjectVisitorLoad;
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitPHINode(PHINode&) {
+  // too complex to analyze statically.
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitSelectInst(SelectInst &I) {
+  SizeOffsetType TrueSide  = compute(I.getTrueValue());
+  SizeOffsetType FalseSide = compute(I.getFalseValue());
+  if (bothKnown(TrueSide) && bothKnown(FalseSide) && TrueSide == FalseSide)
+    return TrueSide;
+  return unknown();
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitUndefValue(UndefValue&) {
+  return std::make_pair(Zero, Zero);
+}
+
+SizeOffsetType ObjectSizeOffsetVisitor::visitInstruction(Instruction &I) {
+  DEBUG(dbgs() << "ObjectSizeOffsetVisitor unknown instruction:" << I << '\n');
+  return unknown();
+}
+
+
+ObjectSizeOffsetEvaluator::ObjectSizeOffsetEvaluator(const TargetData *TD,
+                                                     LLVMContext &Context)
+: TD(TD), Context(Context), Builder(Context, TargetFolder(TD)),
+Visitor(TD, Context) {
+  IntTy = TD->getIntPtrType(Context);
+  Zero = ConstantInt::get(IntTy, 0);
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::compute(Value *V) {
+  SizeOffsetEvalType Result = compute_(V);
+
+  if (!bothKnown(Result)) {
+    // erase everything that was computed in this iteration from the cache, so
+    // that no dangling references are left behind. We could be a bit smarter if
+    // we kept a dependency graph. It's probably not worth the complexity.
+    for (PtrSetTy::iterator I=SeenVals.begin(), E=SeenVals.end(); I != E; ++I) {
+      CacheMapTy::iterator CacheIt = CacheMap.find(*I);
+      // non-computable results can be safely cached
+      if (CacheIt != CacheMap.end() && anyKnown(CacheIt->second))
+        CacheMap.erase(CacheIt);
+    }
+  }
+
+  SeenVals.clear();
+  return Result;
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::compute_(Value *V) {
+  SizeOffsetType Const = Visitor.compute(V);
+  if (Visitor.bothKnown(Const))
+    return std::make_pair(ConstantInt::get(Context, Const.first),
+                          ConstantInt::get(Context, Const.second));
+
+  V = V->stripPointerCasts();
+
+  // check cache
+  CacheMapTy::iterator CacheIt = CacheMap.find(V);
+  if (CacheIt != CacheMap.end())
+    return CacheIt->second;
+
+  // always generate code immediately before the instruction being
+  // processed, so that the generated code dominates the same BBs
+  Instruction *PrevInsertPoint = Builder.GetInsertPoint();
+  if (Instruction *I = dyn_cast<Instruction>(V))
+    Builder.SetInsertPoint(I);
+
+  // record the pointers that were handled in this run, so that they can be
+  // cleaned later if something fails
+  SeenVals.insert(V);
+
+  // now compute the size and offset
+  SizeOffsetEvalType Result;
+  if (GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
+    Result = visitGEPOperator(*GEP);
+  } else if (Instruction *I = dyn_cast<Instruction>(V)) {
+    Result = visit(*I);
+  } else if (isa<Argument>(V) ||
+             (isa<ConstantExpr>(V) &&
+              cast<ConstantExpr>(V)->getOpcode() == Instruction::IntToPtr) ||
+             isa<GlobalVariable>(V)) {
+    // ignore values where we cannot do more than what ObjectSizeVisitor can
+    Result = unknown();
+  } else {
+    DEBUG(dbgs() << "ObjectSizeOffsetEvaluator::compute() unhandled value: "
+          << *V << '\n');
+    Result = unknown();
+  }
+
+  if (PrevInsertPoint)
+    Builder.SetInsertPoint(PrevInsertPoint);
+
+  // Don't reuse CacheIt since it may be invalid at this point.
+  CacheMap[V] = Result;
+  return Result;
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitAllocaInst(AllocaInst &I) {
+  if (!I.getAllocatedType()->isSized())
+    return unknown();
+
+  // must be a VLA
+  assert(I.isArrayAllocation());
+  Value *ArraySize = I.getArraySize();
+  Value *Size = ConstantInt::get(ArraySize->getType(),
+                                 TD->getTypeAllocSize(I.getAllocatedType()));
+  Size = Builder.CreateMul(Size, ArraySize);
+  return std::make_pair(Size, Zero);
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitCallSite(CallSite CS) {
+  const AllocFnsTy *FnData = getAllocationData(CS.getInstruction(), AnyAlloc);
+  if (!FnData)
+    return unknown();
+
+  // handle strdup-like functions separately
+  if (FnData->AllocTy == StrDupLike) {
+    // TODO
+    return unknown();
+  }
+
+  Value *FirstArg  = CS.getArgument(FnData->FstParam);
+  if (FnData->SndParam == (unsigned char)-1)
+    return std::make_pair(FirstArg, Zero);
+
+  Value *SecondArg = CS.getArgument(FnData->SndParam);
+  Value *Size = Builder.CreateMul(FirstArg, SecondArg);
+  return std::make_pair(Size, Zero);
+
+  // TODO: handle more standard functions (+ wchar cousins):
+  // - strdup / strndup
+  // - strcpy / strncpy
+  // - strcat / strncat
+  // - memcpy / memmove
+  // - strcat / strncat
+  // - memset
+}
+
+SizeOffsetEvalType
+ObjectSizeOffsetEvaluator::visitGEPOperator(GEPOperator &GEP) {
+  SizeOffsetEvalType PtrData = compute_(GEP.getPointerOperand());
+  if (!bothKnown(PtrData))
+    return unknown();
+
+  Value *Offset = EmitGEPOffset(&Builder, *TD, &GEP);
+  Offset = Builder.CreateAdd(PtrData.second, Offset);
+  return std::make_pair(PtrData.first, Offset);
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitIntToPtrInst(IntToPtrInst&) {
+  // clueless
+  return unknown();
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitLoadInst(LoadInst&) {
+  return unknown();
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitPHINode(PHINode &PHI) {
+  // create 2 PHIs: one for size and another for offset
+  PHINode *SizePHI   = Builder.CreatePHI(IntTy, PHI.getNumIncomingValues());
+  PHINode *OffsetPHI = Builder.CreatePHI(IntTy, PHI.getNumIncomingValues());
+
+  // insert right away in the cache to handle recursive PHIs
+  CacheMap[&PHI] = std::make_pair(SizePHI, OffsetPHI);
+
+  // compute offset/size for each PHI incoming pointer
+  for (unsigned i = 0, e = PHI.getNumIncomingValues(); i != e; ++i) {
+    Builder.SetInsertPoint(PHI.getIncomingBlock(i)->getFirstInsertionPt());
+    SizeOffsetEvalType EdgeData = compute_(PHI.getIncomingValue(i));
+
+    if (!bothKnown(EdgeData)) {
+      OffsetPHI->replaceAllUsesWith(UndefValue::get(IntTy));
+      OffsetPHI->eraseFromParent();
+      SizePHI->replaceAllUsesWith(UndefValue::get(IntTy));
+      SizePHI->eraseFromParent();
+      return unknown();
+    }
+    SizePHI->addIncoming(EdgeData.first, PHI.getIncomingBlock(i));
+    OffsetPHI->addIncoming(EdgeData.second, PHI.getIncomingBlock(i));
+  }
+  return std::make_pair(SizePHI, OffsetPHI);
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitSelectInst(SelectInst &I) {
+  SizeOffsetEvalType TrueSide  = compute_(I.getTrueValue());
+  SizeOffsetEvalType FalseSide = compute_(I.getFalseValue());
+
+  if (!bothKnown(TrueSide) || !bothKnown(FalseSide))
+    return unknown();
+  if (TrueSide == FalseSide)
+    return TrueSide;
+
+  Value *Size = Builder.CreateSelect(I.getCondition(), TrueSide.first,
+                                     FalseSide.first);
+  Value *Offset = Builder.CreateSelect(I.getCondition(), TrueSide.second,
+                                       FalseSide.second);
+  return std::make_pair(Size, Offset);
+}
+
+SizeOffsetEvalType ObjectSizeOffsetEvaluator::visitInstruction(Instruction &I) {
+  DEBUG(dbgs() << "ObjectSizeOffsetEvaluator unknown instruction:" << I <<'\n');
+  return unknown();
+}
diff --git a/lib/Analysis/MemoryDependenceAnalysis.cpp b/lib/Analysis/MemoryDependenceAnalysis.cpp
index 21ab7a8..7fb154d 100644
--- a/lib/Analysis/MemoryDependenceAnalysis.cpp
+++ b/lib/Analysis/MemoryDependenceAnalysis.cpp
@@ -474,8 +474,7 @@ getPointerDependencyFrom(const AliasAnalysis::Location &MemLoc, bool isLoad,
     // a subsequent bitcast of the malloc call result.  There can be stores to
     // the malloced memory between the malloc call and its bitcast uses, and we
     // need to continue scanning until the malloc call.
-    if (isa<AllocaInst>(Inst) ||
-        (isa<CallInst>(Inst) && extractMallocCall(Inst))) {
+    if (isa<AllocaInst>(Inst) || isNoAliasFn(Inst)) {
       const Value *AccessPtr = GetUnderlyingObject(MemLoc.Ptr, TD);
       
       if (AccessPtr == Inst || AA->isMustAlias(Inst, AccessPtr))
diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 19776b1..ed38829 100644
--- a/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -172,7 +172,7 @@ Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
 Instruction *InstCombiner::visitCallInst(CallInst &CI) {
   if (isFreeCall(&CI))
     return visitFree(CI);
-  if (extractMallocCall(&CI) || extractCallocCall(&CI))
+  if (isAllocLikeFn(&CI))
     return visitMalloc(CI);
 
   // If the caller function is nounwind, mark the call as nounwind, even if the
@@ -246,84 +246,10 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
   switch (II->getIntrinsicID()) {
   default: break;
   case Intrinsic::objectsize: {
-    // We need target data for just about everything so depend on it.
-    if (!TD) return 0;
-
-    Type *ReturnTy = CI.getType();
-    uint64_t DontKnow = II->getArgOperand(1) == Builder->getTrue() ? 0 : -1ULL;
-
-    // Get to the real allocated thing and offset as fast as possible.
-    Value *Op1 = II->getArgOperand(0)->stripPointerCasts();
-
-    uint64_t Offset = 0;
-    uint64_t Size = -1ULL;
-
-    // Try to look through constant GEPs.
-    if (GEPOperator *GEP = dyn_cast<GEPOperator>(Op1)) {
-      if (!GEP->hasAllConstantIndices()) return 0;
-
-      // Get the current byte offset into the thing. Use the original
-      // operand in case we're looking through a bitcast.
-      SmallVector<Value*, 8> Ops(GEP->idx_begin(), GEP->idx_end());
-      if (!GEP->getPointerOperandType()->isPointerTy())
-        return 0;
-      Offset = TD->getIndexedOffset(GEP->getPointerOperandType(), Ops);
-
-      Op1 = GEP->getPointerOperand()->stripPointerCasts();
-
-      // Make sure we're not a constant offset from an external
-      // global.
-      if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op1))
-        if (!GV->hasDefinitiveInitializer()) return 0;
-    }
-
-    // If we've stripped down to a single global variable that we
-    // can know the size of then just return that.
-    if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Op1)) {
-      if (GV->hasDefinitiveInitializer()) {
-        Constant *C = GV->getInitializer();
-        Size = TD->getTypeAllocSize(C->getType());
-      } else {
-        // Can't determine size of the GV.
-        Constant *RetVal = ConstantInt::get(ReturnTy, DontKnow);
-        return ReplaceInstUsesWith(CI, RetVal);
-      }
-    } else if (AllocaInst *AI = dyn_cast<AllocaInst>(Op1)) {
-      // Get alloca size.
-      if (AI->getAllocatedType()->isSized()) {
-        Size = TD->getTypeAllocSize(AI->getAllocatedType());
-        if (AI->isArrayAllocation()) {
-          const ConstantInt *C = dyn_cast<ConstantInt>(AI->getArraySize());
-          if (!C) return 0;
-          Size *= C->getZExtValue();
-        }
-      }
-    } else if (CallInst *MI = extractMallocCall(Op1)) {
-      // Get allocation size.
-      Value *Arg = MI->getArgOperand(0);
-      if (ConstantInt *CI = dyn_cast<ConstantInt>(Arg))
-          Size = CI->getZExtValue();
-
-    } else if (CallInst *MI = extractCallocCall(Op1)) {
-      // Get allocation size.
-      Value *Arg1 = MI->getArgOperand(0);
-      Value *Arg2 = MI->getArgOperand(1);
-      if (ConstantInt *CI1 = dyn_cast<ConstantInt>(Arg1))
-        if (ConstantInt *CI2 = dyn_cast<ConstantInt>(Arg2))
-          Size = (CI1->getValue() * CI2->getValue()).getZExtValue();
-    }
-
-    // Do not return "I don't know" here. Later optimization passes could
-    // make it possible to evaluate objectsize to a constant.
-    if (Size == -1ULL)
-      return 0;
-
-    if (Size < Offset) {
-      // Out of bound reference? Negative index normalized to large
-      // index? Just return "I don't know".
-      return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, DontKnow));
-    }
-    return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, Size-Offset));
+    uint64_t Size;
+    if (getObjectSize(II->getArgOperand(0), Size, TD))
+      return ReplaceInstUsesWith(CI, ConstantInt::get(CI.getType(), Size));
+    return 0;
   }
   case Intrinsic::bswap:
     // bswap(bswap(x)) -> x
@@ -768,7 +694,7 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
     TerminatorInst *TI = II->getParent()->getTerminator();
     bool CannotRemove = false;
     for (++BI; &*BI != TI; ++BI) {
-      if (isa<AllocaInst>(BI) || isMalloc(BI)) {
+      if (isa<AllocaInst>(BI)) {
         CannotRemove = true;
         break;
       }
diff --git a/lib/Transforms/InstCombine/InstructionCombining.cpp b/lib/Transforms/InstCombine/InstructionCombining.cpp
index 2548ac1..2653fc3 100644
--- a/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1068,7 +1068,7 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
         // If the bitcast is of an allocation, and the allocation will be
         // converted to match the type of the cast, don't touch this.
         if (isa<AllocaInst>(BCI->getOperand(0)) ||
-            isMalloc(BCI->getOperand(0))) {
+            isAllocationFn(BCI->getOperand(0))) {
           // See if the bitcast simplifies, if so, don't nuke this GEP yet.
           if (Instruction *I = visitBitCast(*BCI)) {
             if (I != BCI) {
diff --git a/lib/Transforms/Scalar/DeadStoreElimination.cpp b/lib/Transforms/Scalar/DeadStoreElimination.cpp
index 0eb22a0..c8448fa 100644
--- a/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -275,39 +275,9 @@ static Value *getStoredPointerOperand(Instruction *I) {
 }
 
 static uint64_t getPointerSize(const Value *V, AliasAnalysis &AA) {
-  const TargetData *TD = AA.getTargetData();
-
-  if (const CallInst *CI = extractMallocCall(V)) {
-    if (const ConstantInt *C = dyn_cast<ConstantInt>(CI->getArgOperand(0)))
-      return C->getZExtValue();
-  }
-
-  if (const CallInst *CI = extractCallocCall(V)) {
-    if (const ConstantInt *C1 = dyn_cast<ConstantInt>(CI->getArgOperand(0)))
-      if (const ConstantInt *C2 = dyn_cast<ConstantInt>(CI->getArgOperand(1)))
-       return (C1->getValue() * C2->getValue()).getZExtValue();
-  }
-
-  if (TD == 0)
-    return AliasAnalysis::UnknownSize;
-
-  if (const AllocaInst *A = dyn_cast<AllocaInst>(V)) {
-    // Get size information for the alloca
-    if (const ConstantInt *C = dyn_cast<ConstantInt>(A->getArraySize()))
-      return C->getZExtValue() * TD->getTypeAllocSize(A->getAllocatedType());
-  }
-
-  if (const Argument *A = dyn_cast<Argument>(V)) {
-    if (A->hasByValAttr())
-      if (PointerType *PT = dyn_cast<PointerType>(A->getType()))
-        return TD->getTypeAllocSize(PT->getElementType());
-  }
-
-  if (const GlobalVariable *GV = dyn_cast<GlobalVariable>(V)) {
-    if (!GV->mayBeOverridden())
-      return TD->getTypeAllocSize(GV->getType()->getElementType());
-  }
-
+  uint64_t Size;
+  if (getObjectSize(V, Size, AA.getTargetData()))
+    return Size;
   return AliasAnalysis::UnknownSize;
 }
 
@@ -705,16 +675,13 @@ bool DSE::handleEndBlock(BasicBlock &BB) {
   // Find all of the alloca'd pointers in the entry block.
   BasicBlock *Entry = BB.getParent()->begin();
   for (BasicBlock::iterator I = Entry->begin(), E = Entry->end(); I != E; ++I) {
-    if (AllocaInst *AI = dyn_cast<AllocaInst>(I))
-      DeadStackObjects.insert(AI);
+    if (isa<AllocaInst>(I))
+      DeadStackObjects.insert(I);
 
     // Okay, so these are dead heap objects, but if the pointer never escapes
     // then it's leaked by this function anyways.
-    CallInst *CI = extractMallocCall(I);
-    if (!CI)
-      CI = extractCallocCall(I);
-    if (CI && !PointerMayBeCaptured(CI, true, true))
-      DeadStackObjects.insert(CI);
+    else if (isAllocLikeFn(I) && !PointerMayBeCaptured(I, true, true))
+      DeadStackObjects.insert(I);
   }
 
   // Treat byval arguments the same, stores to them are dead at the end of the
@@ -773,18 +740,8 @@ bool DSE::handleEndBlock(BasicBlock &BB) {
       continue;
     }
 
-    if (AllocaInst *A = dyn_cast<AllocaInst>(BBI)) {
-      DeadStackObjects.remove(A);
-      continue;
-    }
-
-    if (CallInst *CI = extractMallocCall(BBI)) {
-      DeadStackObjects.remove(CI);
-      continue;
-    }
-
-    if (CallInst *CI = extractCallocCall(BBI)) {
-      DeadStackObjects.remove(CI);
+    if (isa<AllocaInst>(BBI) || isAllocLikeFn(BBI)) {
+      DeadStackObjects.remove(BBI);
       continue;
     }
 
diff --git a/lib/Transforms/Scalar/GVN.cpp b/lib/Transforms/Scalar/GVN.cpp
index 52f31b1..1b886fd 100644
--- a/lib/Transforms/Scalar/GVN.cpp
+++ b/lib/Transforms/Scalar/GVN.cpp
@@ -1436,7 +1436,7 @@ bool GVN::processNonLocalLoad(LoadInst *LI) {
     Instruction *DepInst = DepInfo.getInst();
 
     // Loading the allocation -> undef.
-    if (isa<AllocaInst>(DepInst) || isMalloc(DepInst) ||
+    if (isa<AllocaInst>(DepInst) || isMallocLikeFn(DepInst) ||
         // Loading immediately after lifetime begin -> undef.
         isLifetimeStart(DepInst)) {
       ValuesPerBlock.push_back(AvailableValueInBlock::get(DepBB,
@@ -1951,7 +1951,7 @@ bool GVN::processLoad(LoadInst *L) {
   // If this load really doesn't depend on anything, then we must be loading an
   // undef value.  This can happen when loading for a fresh allocation with no
   // intervening stores, for example.
-  if (isa<AllocaInst>(DepInst) || isMalloc(DepInst)) {
+  if (isa<AllocaInst>(DepInst) || isMallocLikeFn(DepInst)) {
     L->replaceAllUsesWith(UndefValue::get(L->getType()));
     markInstructionForDeletion(L);
     ++NumGVNLoad;
diff --git a/lib/Transforms/Utils/Local.cpp b/lib/Transforms/Utils/Local.cpp
index b08f8e2..8e59274 100644
--- a/lib/Transforms/Utils/Local.cpp
+++ b/lib/Transforms/Utils/Local.cpp
@@ -266,7 +266,7 @@ bool llvm::isInstructionTriviallyDead(Instruction *I) {
       return isa<UndefValue>(II->getArgOperand(1));
   }
 
-  if (extractMallocCall(I) || extractCallocCall(I)) return true;
+  if (isAllocLikeFn(I)) return true;
 
   if (CallInst *CI = isFreeCall(I))
     if (Constant *C = dyn_cast<Constant>(CI->getArgOperand(0)))
diff --git a/test/Transforms/InstCombine/objsize.ll b/test/Transforms/InstCombine/objsize.ll
index 9bf215c..2d2d6c2 100644
--- a/test/Transforms/InstCombine/objsize.ll
+++ b/test/Transforms/InstCombine/objsize.ll
@@ -42,7 +42,7 @@ define i32 @f() nounwind {
 
 define i1 @baz() nounwind {
 ; CHECK: @baz
-; CHECK-NEXT: ret i1 true
+; CHECK-NEXT: objectsize
   %1 = tail call i32 @llvm.objectsize.i32(i8* getelementptr inbounds ([0 x i8]* @window, i32 0, i32 0), i1 false)
   %2 = icmp eq i32 %1, -1
   ret i1 %2