Merge with LLVM upstream r146714 (Dec 16th 2011)

Change-Id: Ied458adb08bf9a69250cbcee9b14b44d17e8701a

13 files changed, 167 insertions, 640 deletions

diff --git a/lib/VMCore/AsmWriter.cpp b/lib/VMCore/AsmWriter.cpp
index d7863f5..4fb5fd3 100644
--- a/lib/VMCore/AsmWriter.cpp
+++ b/lib/VMCore/AsmWriter.cpp
@@ -2110,3 +2110,6 @@ void Type::dump() const { print(dbgs()); }
 
 // Module::dump() - Allow printing of Modules from the debugger.
 void Module::dump() const { print(dbgs(), 0); }
+
+// NamedMDNode::dump() - Allow printing of NamedMDNodes from the debugger.
+void NamedMDNode::dump() const { print(dbgs(), 0); }
diff --git a/lib/VMCore/AutoUpgrade.cpp b/lib/VMCore/AutoUpgrade.cpp
index b849d3e..59424f9 100644
--- a/lib/VMCore/AutoUpgrade.cpp
+++ b/lib/VMCore/AutoUpgrade.cpp
@@ -38,105 +38,21 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
     return false;
   Name = Name.substr(5); // Strip off "llvm."
 
-  FunctionType *FTy = F->getFunctionType();
-  Module *M = F->getParent();
-  
   switch (Name[0]) {
   default: break;
-  case 'a':
-    if (Name.startswith("atomic.cmp.swap") ||
-        Name.startswith("atomic.swap") ||
-        Name.startswith("atomic.load.add") ||
-        Name.startswith("atomic.load.sub") ||
-        Name.startswith("atomic.load.and") ||
-        Name.startswith("atomic.load.nand") ||
-        Name.startswith("atomic.load.or") ||
-        Name.startswith("atomic.load.xor") ||
-        Name.startswith("atomic.load.max") ||
-        Name.startswith("atomic.load.min") ||
-        Name.startswith("atomic.load.umax") ||
-        Name.startswith("atomic.load.umin"))
-      return true;
-  case 'i':
-    //  This upgrades the old llvm.init.trampoline to the new
-    //  llvm.init.trampoline and llvm.adjust.trampoline pair.
-    if (Name == "init.trampoline") {
-      // The new llvm.init.trampoline returns nothing.
-      if (FTy->getReturnType()->isVoidTy())
-        break;
-
-      assert(FTy->getNumParams() == 3 && "old init.trampoline takes 3 args!");
-
-      // Change the name of the old intrinsic so that we can play with its type.
-      std::string NameTmp = F->getName();
-      F->setName("");
-      NewFn = cast<Function>(M->getOrInsertFunction(
-                               NameTmp,
-                               Type::getVoidTy(M->getContext()),
-                               FTy->getParamType(0), FTy->getParamType(1),
-                               FTy->getParamType(2), (Type *)0));
+  case 'c': {
+    if (Name.startswith("ctlz.") && F->arg_size() == 1) {
+      F->setName(Name + ".old");
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::ctlz,
+                                        F->arg_begin()->getType());
       return true;
     }
-  case 'm':
-    if (Name == "memory.barrier")
-      return true;
-  case 'p':
-    //  This upgrades the llvm.prefetch intrinsic to accept one more parameter,
-    //  which is a instruction / data cache identifier. The old version only
-    //  implicitly accepted the data version.
-    if (Name == "prefetch") {
-      // Don't do anything if it has the correct number of arguments already
-      if (FTy->getNumParams() == 4)
-        break;
-
-      assert(FTy->getNumParams() == 3 && "old prefetch takes 3 args!");
-      //  We first need to change the name of the old (bad) intrinsic, because
-      //  its type is incorrect, but we cannot overload that name. We
-      //  arbitrarily unique it here allowing us to construct a correctly named
-      //  and typed function below.
-      std::string NameTmp = F->getName();
-      F->setName("");
-      NewFn = cast<Function>(M->getOrInsertFunction(NameTmp,
-                                                    FTy->getReturnType(),
-                                                    FTy->getParamType(0),
-                                                    FTy->getParamType(1),
-                                                    FTy->getParamType(2),
-                                                    FTy->getParamType(2),
-                                                    (Type*)0));
+    if (Name.startswith("cttz.") && F->arg_size() == 1) {
+      F->setName(Name + ".old");
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::cttz,
+                                        F->arg_begin()->getType());
       return true;
     }
-
-    break;
-  case 'x': {
-    const char *NewFnName = NULL;
-    // This fixes the poorly named crc32 intrinsics.
-    if (Name == "x86.sse42.crc32.8")
-      NewFnName = "llvm.x86.sse42.crc32.32.8";
-    else if (Name == "x86.sse42.crc32.16")
-      NewFnName = "llvm.x86.sse42.crc32.32.16";
-    else if (Name == "x86.sse42.crc32.32")
-      NewFnName = "llvm.x86.sse42.crc32.32.32";
-    else if (Name == "x86.sse42.crc64.8")
-      NewFnName = "llvm.x86.sse42.crc32.64.8";
-    else if (Name == "x86.sse42.crc64.64")
-      NewFnName = "llvm.x86.sse42.crc32.64.64";
-    
-    if (NewFnName) {
-      F->setName(NewFnName);
-      NewFn = F;
-      return true;
-    }
-
-    // Calls to these instructions are transformed into unaligned loads.
-    if (Name == "x86.sse.loadu.ps" || Name == "x86.sse2.loadu.dq" ||
-        Name == "x86.sse2.loadu.pd")
-      return true;
-      
-    // Calls to these instructions are transformed into nontemporal stores.
-    if (Name == "x86.sse.movnt.ps"  || Name == "x86.sse2.movnt.dq" ||
-        Name == "x86.sse2.movnt.pd" || Name == "x86.sse2.movnt.i")
-      return true;
-
     break;
   }
   }
@@ -169,190 +85,27 @@ bool llvm::UpgradeGlobalVariable(GlobalVariable *GV) {
 // upgraded intrinsic. All argument and return casting must be provided in 
 // order to seamlessly integrate with existing context.
 void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
-  Function *F = CI->getCalledFunction();
-  LLVMContext &C = CI->getContext();
-  ImmutableCallSite CS(CI);
-
-  assert(F && "CallInst has no function associated with it.");
-
-  if (!NewFn) {
-    if (F->getName() == "llvm.x86.sse.loadu.ps" ||
-        F->getName() == "llvm.x86.sse2.loadu.dq" ||
-        F->getName() == "llvm.x86.sse2.loadu.pd") {
-      // Convert to a native, unaligned load.
-      Type *VecTy = CI->getType();
-      Type *IntTy = IntegerType::get(C, 128);
-      IRBuilder<> Builder(C);
-      Builder.SetInsertPoint(CI->getParent(), CI);
-
-      Value *BC = Builder.CreateBitCast(CI->getArgOperand(0),
-                                        PointerType::getUnqual(IntTy),
-                                        "cast");
-      LoadInst *LI = Builder.CreateLoad(BC, CI->getName());
-      LI->setAlignment(1);      // Unaligned load.
-      BC = Builder.CreateBitCast(LI, VecTy, "new.cast");
-
-      // Fix up all the uses with our new load.
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(BC);
-
-      // Remove intrinsic.
-      CI->eraseFromParent();
-    } else if (F->getName() == "llvm.x86.sse.movnt.ps" ||
-               F->getName() == "llvm.x86.sse2.movnt.dq" ||
-               F->getName() == "llvm.x86.sse2.movnt.pd" ||
-               F->getName() == "llvm.x86.sse2.movnt.i") {
-      IRBuilder<> Builder(C);
-      Builder.SetInsertPoint(CI->getParent(), CI);
-
-      Module *M = F->getParent();
-      SmallVector<Value *, 1> Elts;
-      Elts.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
-      MDNode *Node = MDNode::get(C, Elts);
-
-      Value *Arg0 = CI->getArgOperand(0);
-      Value *Arg1 = CI->getArgOperand(1);
-
-      // Convert the type of the pointer to a pointer to the stored type.
-      Value *BC = Builder.CreateBitCast(Arg0,
-                                        PointerType::getUnqual(Arg1->getType()),
-                                        "cast");
-      StoreInst *SI = Builder.CreateStore(Arg1, BC);
-      SI->setMetadata(M->getMDKindID("nontemporal"), Node);
-      SI->setAlignment(16);
-
-      // Remove intrinsic.
-      CI->eraseFromParent();
-    } else if (F->getName().startswith("llvm.atomic.cmp.swap")) {
-      IRBuilder<> Builder(C);
-      Builder.SetInsertPoint(CI->getParent(), CI);
-      Value *Val = Builder.CreateAtomicCmpXchg(CI->getArgOperand(0),
-                                               CI->getArgOperand(1),
-                                               CI->getArgOperand(2),
-                                               Monotonic);
-
-      // Replace intrinsic.
-      Val->takeName(CI);
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(Val);
-      CI->eraseFromParent();
-    } else if (F->getName().startswith("llvm.atomic")) {
-      IRBuilder<> Builder(C);
-      Builder.SetInsertPoint(CI->getParent(), CI);
-
-      AtomicRMWInst::BinOp Op;
-      if (F->getName().startswith("llvm.atomic.swap"))
-        Op = AtomicRMWInst::Xchg;
-      else if (F->getName().startswith("llvm.atomic.load.add"))
-        Op = AtomicRMWInst::Add;
-      else if (F->getName().startswith("llvm.atomic.load.sub"))
-        Op = AtomicRMWInst::Sub;
-      else if (F->getName().startswith("llvm.atomic.load.and"))
-        Op = AtomicRMWInst::And;
-      else if (F->getName().startswith("llvm.atomic.load.nand"))
-        Op = AtomicRMWInst::Nand;
-      else if (F->getName().startswith("llvm.atomic.load.or"))
-        Op = AtomicRMWInst::Or;
-      else if (F->getName().startswith("llvm.atomic.load.xor"))
-        Op = AtomicRMWInst::Xor;
-      else if (F->getName().startswith("llvm.atomic.load.max"))
-        Op = AtomicRMWInst::Max;
-      else if (F->getName().startswith("llvm.atomic.load.min"))
-        Op = AtomicRMWInst::Min;
-      else if (F->getName().startswith("llvm.atomic.load.umax"))
-        Op = AtomicRMWInst::UMax;
-      else if (F->getName().startswith("llvm.atomic.load.umin"))
-        Op = AtomicRMWInst::UMin;
-      else
-        llvm_unreachable("Unknown atomic");
-
-      Value *Val = Builder.CreateAtomicRMW(Op, CI->getArgOperand(0),
-                                           CI->getArgOperand(1),
-                                           Monotonic);
-
-      // Replace intrinsic.
-      Val->takeName(CI);
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(Val);
-      CI->eraseFromParent();
-    } else if (F->getName() == "llvm.memory.barrier") {
-      IRBuilder<> Builder(C);
-      Builder.SetInsertPoint(CI->getParent(), CI);
-
-      // Note that this conversion ignores the "device" bit; it was not really
-      // well-defined, and got abused because nobody paid enough attention to
-      // get it right. In practice, this probably doesn't matter; application
-      // code generally doesn't need anything stronger than
-      // SequentiallyConsistent (and realistically, SequentiallyConsistent
-      // is lowered to a strong enough barrier for almost anything).
-
-      if (cast<ConstantInt>(CI->getArgOperand(1))->getZExtValue())
-        Builder.CreateFence(SequentiallyConsistent);
-      else if (!cast<ConstantInt>(CI->getArgOperand(0))->getZExtValue())
-        Builder.CreateFence(Release);
-      else if (!cast<ConstantInt>(CI->getArgOperand(3))->getZExtValue())
-        Builder.CreateFence(Acquire);
-      else
-        Builder.CreateFence(AcquireRelease);
+  assert(CI->getCalledFunction() && "Intrinsic call is not direct?");
+  if (!NewFn) return;
 
-      // Remove intrinsic.
-      CI->eraseFromParent();
-    } else {
-      llvm_unreachable("Unknown function for CallInst upgrade.");
-    }
-    return;
-  }
+  LLVMContext &C = CI->getContext();
+  IRBuilder<> Builder(C);
+  Builder.SetInsertPoint(CI->getParent(), CI);
 
   switch (NewFn->getIntrinsicID()) {
-  case Intrinsic::prefetch: {
-    IRBuilder<> Builder(C);
-    Builder.SetInsertPoint(CI->getParent(), CI);
-    llvm::Type *I32Ty = llvm::Type::getInt32Ty(CI->getContext());
-
-    // Add the extra "data cache" argument
-    Value *Operands[4] = { CI->getArgOperand(0), CI->getArgOperand(1),
-                           CI->getArgOperand(2),
-                           llvm::ConstantInt::get(I32Ty, 1) };
-    CallInst *NewCI = CallInst::Create(NewFn, Operands,
-                                       CI->getName(), CI);
-    NewCI->setTailCall(CI->isTailCall());
-    NewCI->setCallingConv(CI->getCallingConv());
-    //  Handle any uses of the old CallInst.
-    if (!CI->use_empty())
-      //  Replace all uses of the old call with the new cast which has the
-      //  correct type.
-      CI->replaceAllUsesWith(NewCI);
-
-    //  Clean up the old call now that it has been completely upgraded.
-    CI->eraseFromParent();
-    break;
-  }
-  case Intrinsic::init_trampoline: {
-
-    //  Transform
-    //    %tramp = call i8* llvm.init.trampoline (i8* x, i8* y, i8* z)
-    //  to
-    //    call void llvm.init.trampoline (i8* %x, i8* %y, i8* %z)
-    //    %tramp = call i8* llvm.adjust.trampoline (i8* %x)
-
-    Function *AdjustTrampolineFn =
-      cast<Function>(Intrinsic::getDeclaration(F->getParent(),
-                                               Intrinsic::adjust_trampoline));
-
-    IRBuilder<> Builder(C);
-    Builder.SetInsertPoint(CI);
-
-    Builder.CreateCall3(NewFn, CI->getArgOperand(0), CI->getArgOperand(1),
-                        CI->getArgOperand(2));
-
-    CallInst *AdjustCall = Builder.CreateCall(AdjustTrampolineFn,
-                                              CI->getArgOperand(0),
-                                              CI->getName());
-    if (!CI->use_empty())
-      CI->replaceAllUsesWith(AdjustCall);
+  default:
+    llvm_unreachable("Unknown function for CallInst upgrade.");
+
+  case Intrinsic::ctlz:
+  case Intrinsic::cttz:
+    assert(CI->getNumArgOperands() == 1 &&
+           "Mismatch between function args and call args");
+    StringRef Name = CI->getName();
+    CI->setName(Name + ".old");
+    CI->replaceAllUsesWith(Builder.CreateCall2(NewFn, CI->getArgOperand(0),
+                                               Builder.getFalse(), Name));
     CI->eraseFromParent();
-    break;
-  }
+    return;
   }
 }
 
@@ -378,291 +131,3 @@ void llvm::UpgradeCallsToIntrinsic(Function* F) {
   }
 }
 
-/// This function strips all debug info intrinsics, except for llvm.dbg.declare.
-/// If an llvm.dbg.declare intrinsic is invalid, then this function simply
-/// strips that use.
-void llvm::CheckDebugInfoIntrinsics(Module *M) {
-  if (Function *FuncStart = M->getFunction("llvm.dbg.func.start")) {
-    while (!FuncStart->use_empty())
-      cast<CallInst>(FuncStart->use_back())->eraseFromParent();
-    FuncStart->eraseFromParent();
-  }
-  
-  if (Function *StopPoint = M->getFunction("llvm.dbg.stoppoint")) {
-    while (!StopPoint->use_empty())
-      cast<CallInst>(StopPoint->use_back())->eraseFromParent();
-    StopPoint->eraseFromParent();
-  }
-
-  if (Function *RegionStart = M->getFunction("llvm.dbg.region.start")) {
-    while (!RegionStart->use_empty())
-      cast<CallInst>(RegionStart->use_back())->eraseFromParent();
-    RegionStart->eraseFromParent();
-  }
-
-  if (Function *RegionEnd = M->getFunction("llvm.dbg.region.end")) {
-    while (!RegionEnd->use_empty())
-      cast<CallInst>(RegionEnd->use_back())->eraseFromParent();
-    RegionEnd->eraseFromParent();
-  }
-  
-  if (Function *Declare = M->getFunction("llvm.dbg.declare")) {
-    if (!Declare->use_empty()) {
-      DbgDeclareInst *DDI = cast<DbgDeclareInst>(Declare->use_back());
-      if (!isa<MDNode>(DDI->getArgOperand(0)) ||
-          !isa<MDNode>(DDI->getArgOperand(1))) {
-        while (!Declare->use_empty()) {
-          CallInst *CI = cast<CallInst>(Declare->use_back());
-          CI->eraseFromParent();
-        }
-        Declare->eraseFromParent();
-      }
-    }
-  }
-}
-
-/// FindExnAndSelIntrinsics - Find the eh_exception and eh_selector intrinsic
-/// calls reachable from the unwind basic block.
-static void FindExnAndSelIntrinsics(BasicBlock *BB, CallInst *&Exn,
-                                    CallInst *&Sel,
-                                    SmallPtrSet<BasicBlock*, 8> &Visited) {
-  if (!Visited.insert(BB)) return;
-
-  for (BasicBlock::iterator
-         I = BB->begin(), E = BB->end(); I != E; ++I) {
-    if (CallInst *CI = dyn_cast<CallInst>(I)) {
-      switch (CI->getCalledFunction()->getIntrinsicID()) {
-      default: break;
-      case Intrinsic::eh_exception:
-        assert(!Exn && "Found more than one eh.exception call!");
-        Exn = CI;
-        break;
-      case Intrinsic::eh_selector:
-        assert(!Sel && "Found more than one eh.selector call!");
-        Sel = CI;
-        break;
-      }
-
-      if (Exn && Sel) return;
-    }
-  }
-
-  if (Exn && Sel) return;
-
-  for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
-    FindExnAndSelIntrinsics(*I, Exn, Sel, Visited);
-    if (Exn && Sel) return;
-  }
-}
-
-/// TransferClausesToLandingPadInst - Transfer the exception handling clauses
-/// from the eh_selector call to the new landingpad instruction.
-static void TransferClausesToLandingPadInst(LandingPadInst *LPI,
-                                            CallInst *EHSel) {
-  LLVMContext &Context = LPI->getContext();
-  unsigned N = EHSel->getNumArgOperands();
-
-  for (unsigned i = N - 1; i > 1; --i) {
-    if (const ConstantInt *CI = dyn_cast<ConstantInt>(EHSel->getArgOperand(i))){
-      unsigned FilterLength = CI->getZExtValue();
-      unsigned FirstCatch = i + FilterLength + !FilterLength;
-      assert(FirstCatch <= N && "Invalid filter length");
-
-      if (FirstCatch < N)
-        for (unsigned j = FirstCatch; j < N; ++j) {
-          Value *Val = EHSel->getArgOperand(j);
-          if (!Val->hasName() || Val->getName() != "llvm.eh.catch.all.value") {
-            LPI->addClause(EHSel->getArgOperand(j));
-          } else {
-            GlobalVariable *GV = cast<GlobalVariable>(Val);
-            LPI->addClause(GV->getInitializer());
-          }
-        }
-
-      if (!FilterLength) {
-        // Cleanup.
-        LPI->setCleanup(true);
-      } else {
-        // Filter.
-        SmallVector<Constant *, 4> TyInfo;
-        TyInfo.reserve(FilterLength - 1);
-        for (unsigned j = i + 1; j < FirstCatch; ++j)
-          TyInfo.push_back(cast<Constant>(EHSel->getArgOperand(j)));
-        ArrayType *AType =
-          ArrayType::get(!TyInfo.empty() ? TyInfo[0]->getType() :
-                         PointerType::getUnqual(Type::getInt8Ty(Context)),
-                         TyInfo.size());
-        LPI->addClause(ConstantArray::get(AType, TyInfo));
-      }
-
-      N = i;
-    }
-  }
-
-  if (N > 2)
-    for (unsigned j = 2; j < N; ++j) {
-      Value *Val = EHSel->getArgOperand(j);
-      if (!Val->hasName() || Val->getName() != "llvm.eh.catch.all.value") {
-        LPI->addClause(EHSel->getArgOperand(j));
-      } else {
-        GlobalVariable *GV = cast<GlobalVariable>(Val);
-        LPI->addClause(GV->getInitializer());
-      }
-    }
-}
-
-/// This function upgrades the old pre-3.0 exception handling system to the new
-/// one. N.B. This will be removed in 3.1.
-void llvm::UpgradeExceptionHandling(Module *M) {
-  Function *EHException = M->getFunction("llvm.eh.exception");
-  Function *EHSelector = M->getFunction("llvm.eh.selector");
-  if (!EHException || !EHSelector)
-    return;
-
-  LLVMContext &Context = M->getContext();
-  Type *ExnTy = PointerType::getUnqual(Type::getInt8Ty(Context));
-  Type *SelTy = Type::getInt32Ty(Context);
-  Type *LPadSlotTy = StructType::get(ExnTy, SelTy, NULL);
-
-  // This map links the invoke instruction with the eh.exception and eh.selector
-  // calls associated with it.
-  DenseMap<InvokeInst*, std::pair<Value*, Value*> > InvokeToIntrinsicsMap;
-  for (Module::iterator
-         I = M->begin(), E = M->end(); I != E; ++I) {
-    Function &F = *I;
-
-    for (Function::iterator
-           II = F.begin(), IE = F.end(); II != IE; ++II) {
-      BasicBlock *BB = &*II;
-      InvokeInst *Inst = dyn_cast<InvokeInst>(BB->getTerminator());
-      if (!Inst) continue;
-      BasicBlock *UnwindDest = Inst->getUnwindDest();
-      if (UnwindDest->isLandingPad()) continue; // Already converted.
-
-      SmallPtrSet<BasicBlock*, 8> Visited;
-      CallInst *Exn = 0;
-      CallInst *Sel = 0;
-      FindExnAndSelIntrinsics(UnwindDest, Exn, Sel, Visited);
-      assert(Exn && Sel && "Cannot find eh.exception and eh.selector calls!");
-      InvokeToIntrinsicsMap[Inst] = std::make_pair(Exn, Sel);
-    }
-  }
-
-  // This map stores the slots where the exception object and selector value are
-  // stored within a function.
-  DenseMap<Function*, std::pair<Value*, Value*> > FnToLPadSlotMap;
-  SmallPtrSet<Instruction*, 32> DeadInsts;
-  for (DenseMap<InvokeInst*, std::pair<Value*, Value*> >::iterator
-         I = InvokeToIntrinsicsMap.begin(), E = InvokeToIntrinsicsMap.end();
-       I != E; ++I) {
-    InvokeInst *Invoke = I->first;
-    BasicBlock *UnwindDest = Invoke->getUnwindDest();
-    Function *F = UnwindDest->getParent();
-    std::pair<Value*, Value*> EHIntrinsics = I->second;
-    CallInst *Exn = cast<CallInst>(EHIntrinsics.first);
-    CallInst *Sel = cast<CallInst>(EHIntrinsics.second);
-
-    // Store the exception object and selector value in the entry block.
-    Value *ExnSlot = 0;
-    Value *SelSlot = 0;
-    if (!FnToLPadSlotMap[F].first) {
-      BasicBlock *Entry = &F->front();
-      ExnSlot = new AllocaInst(ExnTy, "exn", Entry->getTerminator());
-      SelSlot = new AllocaInst(SelTy, "sel", Entry->getTerminator());
-      FnToLPadSlotMap[F] = std::make_pair(ExnSlot, SelSlot);
-    } else {
-      ExnSlot = FnToLPadSlotMap[F].first;
-      SelSlot = FnToLPadSlotMap[F].second;
-    }
-
-    if (!UnwindDest->getSinglePredecessor()) {
-      // The unwind destination doesn't have a single predecessor. Create an
-      // unwind destination which has only one predecessor.
-      BasicBlock *NewBB = BasicBlock::Create(Context, "new.lpad",
-                                             UnwindDest->getParent());
-      BranchInst::Create(UnwindDest, NewBB);
-      Invoke->setUnwindDest(NewBB);
-
-      // Fix up any PHIs in the original unwind destination block.
-      for (BasicBlock::iterator
-             II = UnwindDest->begin(); isa<PHINode>(II); ++II) {
-        PHINode *PN = cast<PHINode>(II);
-        int Idx = PN->getBasicBlockIndex(Invoke->getParent());
-        if (Idx == -1) continue;
-        PN->setIncomingBlock(Idx, NewBB);
-      }
-
-      UnwindDest = NewBB;
-    }
-
-    IRBuilder<> Builder(Context);
-    Builder.SetInsertPoint(UnwindDest, UnwindDest->getFirstInsertionPt());
-
-    Value *PersFn = Sel->getArgOperand(1);
-    LandingPadInst *LPI = Builder.CreateLandingPad(LPadSlotTy, PersFn, 0);
-    Value *LPExn = Builder.CreateExtractValue(LPI, 0);
-    Value *LPSel = Builder.CreateExtractValue(LPI, 1);
-    Builder.CreateStore(LPExn, ExnSlot);
-    Builder.CreateStore(LPSel, SelSlot);
-
-    TransferClausesToLandingPadInst(LPI, Sel);
-
-    DeadInsts.insert(Exn);
-    DeadInsts.insert(Sel);
-  }
-
-  // Replace the old intrinsic calls with the values from the landingpad
-  // instruction(s). These values were stored in allocas for us to use here.
-  for (DenseMap<InvokeInst*, std::pair<Value*, Value*> >::iterator
-         I = InvokeToIntrinsicsMap.begin(), E = InvokeToIntrinsicsMap.end();
-       I != E; ++I) {
-    std::pair<Value*, Value*> EHIntrinsics = I->second;
-    CallInst *Exn = cast<CallInst>(EHIntrinsics.first);
-    CallInst *Sel = cast<CallInst>(EHIntrinsics.second);
-    BasicBlock *Parent = Exn->getParent();
-
-    std::pair<Value*,Value*> ExnSelSlots = FnToLPadSlotMap[Parent->getParent()];
-
-    IRBuilder<> Builder(Context);
-    Builder.SetInsertPoint(Parent, Exn);
-    LoadInst *LPExn = Builder.CreateLoad(ExnSelSlots.first, "exn.load");
-    LoadInst *LPSel = Builder.CreateLoad(ExnSelSlots.second, "sel.load");
-
-    Exn->replaceAllUsesWith(LPExn);
-    Sel->replaceAllUsesWith(LPSel);
-  }
-
-  // Remove the dead instructions.
-  for (SmallPtrSet<Instruction*, 32>::iterator
-         I = DeadInsts.begin(), E = DeadInsts.end(); I != E; ++I) {
-    Instruction *Inst = *I;
-    Inst->eraseFromParent();
-  }
-
-  // Replace calls to "llvm.eh.resume" with the 'resume' instruction. Load the
-  // exception and selector values from the stored place.
-  Function *EHResume = M->getFunction("llvm.eh.resume");
-  if (!EHResume) return;
-
-  while (!EHResume->use_empty()) {
-    CallInst *Resume = cast<CallInst>(EHResume->use_back());
-    BasicBlock *BB = Resume->getParent();
-
-    IRBuilder<> Builder(Context);
-    Builder.SetInsertPoint(BB, Resume);
-
-    Value *LPadVal =
-      Builder.CreateInsertValue(UndefValue::get(LPadSlotTy),
-                                Resume->getArgOperand(0), 0, "lpad.val");
-    LPadVal = Builder.CreateInsertValue(LPadVal, Resume->getArgOperand(1),
-                                        1, "lpad.val");
-    Builder.CreateResume(LPadVal);
-
-    // Remove all instructions after the 'resume.'
-    BasicBlock::iterator I = Resume;
-    while (I != BB->end()) {
-      Instruction *Inst = &*I++;
-      Inst->eraseFromParent();
-    }
-  }
-}
diff --git a/lib/VMCore/CMakeLists.txt b/lib/VMCore/CMakeLists.txt
index 0404297..99eeba1 100644
--- a/lib/VMCore/CMakeLists.txt
+++ b/lib/VMCore/CMakeLists.txt
@@ -37,5 +37,3 @@ add_llvm_library(LLVMCore
   ValueTypes.cpp
   Verifier.cpp
   )
-
-add_llvm_library_dependencies(LLVMCore LLVMSupport)
diff --git a/lib/VMCore/ConstantFold.cpp b/lib/VMCore/ConstantFold.cpp
index 30bae71..d1a9e7a 100644
--- a/lib/VMCore/ConstantFold.cpp
+++ b/lib/VMCore/ConstantFold.cpp
@@ -2209,7 +2209,7 @@ static Constant *ConstantFoldGetElementPtrImpl(Constant *C,
            I != E; ++I)
         LastTy = *I;
 
-      if ((LastTy && LastTy->isArrayTy()) || Idx0->isNullValue()) {
+      if ((LastTy && isa<SequentialType>(LastTy)) || Idx0->isNullValue()) {
         SmallVector<Value*, 16> NewIndices;
         NewIndices.reserve(Idxs.size() + CE->getNumOperands());
         for (unsigned i = 1, e = CE->getNumOperands()-1; i != e; ++i)
diff --git a/lib/VMCore/Constants.cpp b/lib/VMCore/Constants.cpp
index cd94da1..a148912 100644
--- a/lib/VMCore/Constants.cpp
+++ b/lib/VMCore/Constants.cpp
@@ -1398,14 +1398,22 @@ Constant *ConstantExpr::getFPToSI(Constant *C, Type *Ty) {
 }
 
 Constant *ConstantExpr::getPtrToInt(Constant *C, Type *DstTy) {
-  assert(C->getType()->isPointerTy() && "PtrToInt source must be pointer");
-  assert(DstTy->isIntegerTy() && "PtrToInt destination must be integral");
+  assert(C->getType()->getScalarType()->isPointerTy() &&
+         "PtrToInt source must be pointer or pointer vector");
+  assert(DstTy->getScalarType()->isIntegerTy() && 
+         "PtrToInt destination must be integer or integer vector");
+  assert(C->getType()->getNumElements() == DstTy->getNumElements() &&
+    "Invalid cast between a different number of vector elements");
   return getFoldedCast(Instruction::PtrToInt, C, DstTy);
 }
 
 Constant *ConstantExpr::getIntToPtr(Constant *C, Type *DstTy) {
-  assert(C->getType()->isIntegerTy() && "IntToPtr source must be integral");
-  assert(DstTy->isPointerTy() && "IntToPtr destination must be a pointer");
+  assert(C->getType()->getScalarType()->isIntegerTy() &&
+         "IntToPtr source must be integer or integer vector");
+  assert(DstTy->getScalarType()->isPointerTy() &&
+         "IntToPtr destination must be a pointer or pointer vector");
+  assert(C->getType()->getNumElements() == DstTy->getNumElements() &&
+    "Invalid cast between a different number of vector elements");
   return getFoldedCast(Instruction::IntToPtr, C, DstTy);
 }
 
diff --git a/lib/VMCore/Instruction.cpp b/lib/VMCore/Instruction.cpp
index 73191c1..8c8fbf9 100644
--- a/lib/VMCore/Instruction.cpp
+++ b/lib/VMCore/Instruction.cpp
@@ -391,59 +391,6 @@ bool Instruction::isCommutative(unsigned op) {
   }
 }
 
-bool Instruction::isSafeToSpeculativelyExecute() const {
-  for (unsigned i = 0, e = getNumOperands(); i != e; ++i)
-    if (Constant *C = dyn_cast<Constant>(getOperand(i)))
-      if (C->canTrap())
-        return false;
-
-  switch (getOpcode()) {
-  default:
-    return true;
-  case UDiv:
-  case URem: {
-    // x / y is undefined if y == 0, but calcuations like x / 3 are safe.
-    ConstantInt *Op = dyn_cast<ConstantInt>(getOperand(1));
-    return Op && !Op->isNullValue();
-  }
-  case SDiv:
-  case SRem: {
-    // x / y is undefined if y == 0, and might be undefined if y == -1,
-    // but calcuations like x / 3 are safe.
-    ConstantInt *Op = dyn_cast<ConstantInt>(getOperand(1));
-    return Op && !Op->isNullValue() && !Op->isAllOnesValue();
-  }
-  case Load: {
-    const LoadInst *LI = cast<LoadInst>(this);
-    if (!LI->isUnordered())
-      return false;
-    return LI->getPointerOperand()->isDereferenceablePointer();
-  }
-  case Call:
-    return false; // The called function could have undefined behavior or
-                  // side-effects.
-                  // FIXME: We should special-case some intrinsics (bswap,
-                  // overflow-checking arithmetic, etc.)
-  case VAArg:
-  case Alloca:
-  case Invoke:
-  case PHI:
-  case Store:
-  case Ret:
-  case Br:
-  case IndirectBr:
-  case Switch:
-  case Unwind:
-  case Unreachable:
-  case Fence:
-  case LandingPad:
-  case AtomicRMW:
-  case AtomicCmpXchg:
-  case Resume:
-    return false; // Misc instructions which have effects
-  }
-}
-
 Instruction *Instruction::clone() const {
   Instruction *New = clone_impl();
   New->SubclassOptionalData = SubclassOptionalData;
diff --git a/lib/VMCore/Instructions.cpp b/lib/VMCore/Instructions.cpp
index c8dcdc8..4784f0c 100644
--- a/lib/VMCore/Instructions.cpp
+++ b/lib/VMCore/Instructions.cpp
@@ -1359,6 +1359,15 @@ GetElementPtrInst::GetElementPtrInst(const GetElementPtrInst &GEPI)
 ///
 template <typename IndexTy>
 static Type *getIndexedTypeInternal(Type *Ptr, ArrayRef<IndexTy> IdxList) {
+  if (Ptr->isVectorTy()) {
+    assert(IdxList.size() == 1 &&
+      "GEP with vector pointers must have a single index");
+    PointerType *PTy = dyn_cast<PointerType>(
+        cast<VectorType>(Ptr)->getElementType());
+    assert(PTy && "Gep with invalid vector pointer found");
+    return PTy->getElementType();
+  }
+
   PointerType *PTy = dyn_cast<PointerType>(Ptr);
   if (!PTy) return 0;   // Type isn't a pointer type!
   Type *Agg = PTy->getElementType();
@@ -1366,7 +1375,7 @@ static Type *getIndexedTypeInternal(Type *Ptr, ArrayRef<IndexTy> IdxList) {
   // Handle the special case of the empty set index set, which is always valid.
   if (IdxList.empty())
     return Agg;
-  
+
   // If there is at least one index, the top level type must be sized, otherwise
   // it cannot be 'stepped over'.
   if (!Agg->isSized())
@@ -1396,6 +1405,19 @@ Type *GetElementPtrInst::getIndexedType(Type *Ptr, ArrayRef<uint64_t> IdxList) {
   return getIndexedTypeInternal(Ptr, IdxList);
 }
 
+unsigned GetElementPtrInst::getAddressSpace(Value *Ptr) {
+  Type *Ty = Ptr->getType();
+
+  if (VectorType *VTy = dyn_cast<VectorType>(Ty))
+    Ty = VTy->getElementType();
+
+  if (PointerType *PTy = dyn_cast<PointerType>(Ty))
+    return PTy->getAddressSpace();
+
+  assert(false && "Invalid GEP pointer type");
+  return 0;
+}
+
 /// hasAllZeroIndices - Return true if all of the indices of this GEP are
 /// zeros.  If so, the result pointer and the first operand have the same
 /// value, just potentially different types.
@@ -2005,6 +2027,8 @@ bool BinaryOperator::isExact() const {
 //                                CastInst Class
 //===----------------------------------------------------------------------===//
 
+void CastInst::anchor() {}
+
 // Just determine if this cast only deals with integral->integral conversion.
 bool CastInst::isIntegerCast() const {
   switch (getOpcode()) {
@@ -2652,9 +2676,15 @@ CastInst::castIsValid(Instruction::CastOps op, Value *S, Type *DstTy) {
     return SrcTy->isFPOrFPVectorTy() && DstTy->isIntOrIntVectorTy() &&
       SrcLength == DstLength;
   case Instruction::PtrToInt:
-    return SrcTy->isPointerTy() && DstTy->isIntegerTy();
+    if (SrcTy->getNumElements() != DstTy->getNumElements())
+      return false;
+    return SrcTy->getScalarType()->isPointerTy() &&
+           DstTy->getScalarType()->isIntegerTy();
   case Instruction::IntToPtr:
-    return SrcTy->isIntegerTy() && DstTy->isPointerTy();
+    if (SrcTy->getNumElements() != DstTy->getNumElements())
+      return false;
+    return SrcTy->getScalarType()->isIntegerTy() &&
+           DstTy->getScalarType()->isPointerTy();
   case Instruction::BitCast:
     // BitCast implies a no-op cast of type only. No bits change.
     // However, you can't cast pointers to anything but pointers.
diff --git a/lib/VMCore/LLVMBuild.txt b/lib/VMCore/LLVMBuild.txt
index 45f528e..bca8b2c 100644
--- a/lib/VMCore/LLVMBuild.txt
+++ b/lib/VMCore/LLVMBuild.txt
@@ -20,4 +20,3 @@ type = Library
 name = Core
 parent = Libraries
 required_libraries = Support
-
diff --git a/lib/VMCore/Metadata.cpp b/lib/VMCore/Metadata.cpp
index ace4dc2..8debd7c 100644
--- a/lib/VMCore/Metadata.cpp
+++ b/lib/VMCore/Metadata.cpp
@@ -425,12 +425,12 @@ StringRef NamedMDNode::getName() const {
 // Instruction Metadata method implementations.
 //
 
-void Instruction::setMetadata(const char *Kind, MDNode *Node) {
+void Instruction::setMetadata(StringRef Kind, MDNode *Node) {
   if (Node == 0 && !hasMetadata()) return;
   setMetadata(getContext().getMDKindID(Kind), Node);
 }
 
-MDNode *Instruction::getMetadataImpl(const char *Kind) const {
+MDNode *Instruction::getMetadataImpl(StringRef Kind) const {
   return getMetadataImpl(getContext().getMDKindID(Kind));
 }
 
diff --git a/lib/VMCore/Type.cpp b/lib/VMCore/Type.cpp
index 10184bc..469defd 100644
--- a/lib/VMCore/Type.cpp
+++ b/lib/VMCore/Type.cpp
@@ -46,6 +46,14 @@ Type *Type::getScalarType() {
   return this;
 }
 
+/// getNumElements - If this is a vector type, return the number of elements,
+/// otherwise return zero.
+unsigned Type::getNumElements() {
+  if (VectorType *VTy = dyn_cast<VectorType>(this))
+    return VTy->getNumElements();
+  return 0;
+}
+
 /// isIntegerTy - Return true if this is an IntegerType of the specified width.
 bool Type::isIntegerTy(unsigned Bitwidth) const {
   return isIntegerTy() && cast<IntegerType>(this)->getBitWidth() == Bitwidth;
@@ -664,6 +672,8 @@ VectorType *VectorType::get(Type *elementType, unsigned NumElements) {
 }
 
 bool VectorType::isValidElementType(Type *ElemTy) {
+  if (PointerType *PTy = dyn_cast<PointerType>(ElemTy))
+    ElemTy = PTy->getElementType();
   return ElemTy->isIntegerTy() || ElemTy->isFloatingPointTy();
 }
 
diff --git a/lib/VMCore/User.cpp b/lib/VMCore/User.cpp
index f01fa34..5f35ce4 100644
--- a/lib/VMCore/User.cpp
+++ b/lib/VMCore/User.cpp
@@ -17,6 +17,8 @@ namespace llvm {
 //                                 User Class
 //===----------------------------------------------------------------------===//
 
+void User::anchor() {}
+
 // replaceUsesOfWith - Replaces all references to the "From" definition with
 // references to the "To" definition.
 //
diff --git a/lib/VMCore/Value.cpp b/lib/VMCore/Value.cpp
index 291df91..a5f1918 100644
--- a/lib/VMCore/Value.cpp
+++ b/lib/VMCore/Value.cpp
@@ -108,6 +108,19 @@ bool Value::hasNUsesOrMore(unsigned N) const {
 /// isUsedInBasicBlock - Return true if this value is used in the specified
 /// basic block.
 bool Value::isUsedInBasicBlock(const BasicBlock *BB) const {
+  // Start by scanning over the instructions looking for a use before we start
+  // the expensive use iteration.
+  unsigned MaxBlockSize = 3;
+  for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
+    if (std::find(I->op_begin(), I->op_end(), this) != I->op_end())
+      return true;
+    if (MaxBlockSize-- == 0) // If the block is larger fall back to use_iterator
+      break;
+  }
+
+  if (MaxBlockSize != 0) // We scanned the entire block and found no use.
+    return false;
+
   for (const_use_iterator I = use_begin(), E = use_end(); I != E; ++I) {
     const Instruction *User = dyn_cast<Instruction>(*I);
     if (User && User->getParent() == BB)
diff --git a/lib/VMCore/Verifier.cpp b/lib/VMCore/Verifier.cpp
index 9564b7d..003de44 100644
--- a/lib/VMCore/Verifier.cpp
+++ b/lib/VMCore/Verifier.cpp
@@ -1035,8 +1035,19 @@ void Verifier::visitPtrToIntInst(PtrToIntInst &I) {
   Type *SrcTy = I.getOperand(0)->getType();
   Type *DestTy = I.getType();
 
-  Assert1(SrcTy->isPointerTy(), "PtrToInt source must be pointer", &I);
-  Assert1(DestTy->isIntegerTy(), "PtrToInt result must be integral", &I);
+  Assert1(SrcTy->getScalarType()->isPointerTy(),
+          "PtrToInt source must be pointer", &I);
+  Assert1(DestTy->getScalarType()->isIntegerTy(),
+          "PtrToInt result must be integral", &I);
+  Assert1(SrcTy->isVectorTy() == DestTy->isVectorTy(),
+          "PtrToInt type mismatch", &I);
+
+  if (SrcTy->isVectorTy()) {
+    VectorType *VSrc = dyn_cast<VectorType>(SrcTy);
+    VectorType *VDest = dyn_cast<VectorType>(DestTy);
+    Assert1(VSrc->getNumElements() == VDest->getNumElements(),
+          "PtrToInt Vector width mismatch", &I);
+  }
 
   visitInstruction(I);
 }
@@ -1046,9 +1057,18 @@ void Verifier::visitIntToPtrInst(IntToPtrInst &I) {
   Type *SrcTy = I.getOperand(0)->getType();
   Type *DestTy = I.getType();
 
-  Assert1(SrcTy->isIntegerTy(), "IntToPtr source must be an integral", &I);
-  Assert1(DestTy->isPointerTy(), "IntToPtr result must be a pointer",&I);
-
+  Assert1(SrcTy->getScalarType()->isIntegerTy(),
+          "IntToPtr source must be an integral", &I);
+  Assert1(DestTy->getScalarType()->isPointerTy(),
+          "IntToPtr result must be a pointer",&I);
+  Assert1(SrcTy->isVectorTy() == DestTy->isVectorTy(),
+          "IntToPtr type mismatch", &I);
+  if (SrcTy->isVectorTy()) {
+    VectorType *VSrc = dyn_cast<VectorType>(SrcTy);
+    VectorType *VDest = dyn_cast<VectorType>(DestTy);
+    Assert1(VSrc->getNumElements() == VDest->getNumElements(),
+          "IntToPtr Vector width mismatch", &I);
+  }
   visitInstruction(I);
 }
 
@@ -1245,7 +1265,7 @@ void Verifier::visitICmpInst(ICmpInst &IC) {
   Assert1(Op0Ty == Op1Ty,
           "Both operands to ICmp instruction are not of the same type!", &IC);
   // Check that the operands are the right type
-  Assert1(Op0Ty->isIntOrIntVectorTy() || Op0Ty->isPointerTy(),
+  Assert1(Op0Ty->isIntOrIntVectorTy() || Op0Ty->getScalarType()->isPointerTy(),
           "Invalid operand types for ICmp instruction", &IC);
   // Check that the predicate is valid.
   Assert1(IC.getPredicate() >= CmpInst::FIRST_ICMP_PREDICATE &&
@@ -1295,17 +1315,43 @@ void Verifier::visitShuffleVectorInst(ShuffleVectorInst &SV) {
 }
 
 void Verifier::visitGetElementPtrInst(GetElementPtrInst &GEP) {
-  Assert1(cast<PointerType>(GEP.getOperand(0)->getType())
-            ->getElementType()->isSized(),
+  Type *TargetTy = GEP.getPointerOperandType();
+  if (VectorType *VTy = dyn_cast<VectorType>(TargetTy))
+    TargetTy = VTy->getElementType();
+
+  Assert1(dyn_cast<PointerType>(TargetTy),
+    "GEP base pointer is not a vector or a vector of pointers", &GEP);
+  Assert1(cast<PointerType>(TargetTy)->getElementType()->isSized(),
           "GEP into unsized type!", &GEP);
-  
+
   SmallVector<Value*, 16> Idxs(GEP.idx_begin(), GEP.idx_end());
   Type *ElTy =
-    GetElementPtrInst::getIndexedType(GEP.getOperand(0)->getType(), Idxs);
+    GetElementPtrInst::getIndexedType(GEP.getPointerOperandType(), Idxs);
   Assert1(ElTy, "Invalid indices for GEP pointer type!", &GEP);
-  Assert2(GEP.getType()->isPointerTy() &&
-          cast<PointerType>(GEP.getType())->getElementType() == ElTy,
-          "GEP is not of right type for indices!", &GEP, ElTy);
+
+  if (GEP.getPointerOperandType()->isPointerTy()) {
+    // Validate GEPs with scalar indices.
+    Assert2(GEP.getType()->isPointerTy() &&
+           cast<PointerType>(GEP.getType())->getElementType() == ElTy,
+           "GEP is not of right type for indices!", &GEP, ElTy);
+  } else {
+    // Validate GEPs with a vector index.
+    Assert1(Idxs.size() == 1, "Invalid number of indices!", &GEP);
+    Value *Index = Idxs[0];
+    Type  *IndexTy = Index->getType();
+    Assert1(IndexTy->isVectorTy(),
+      "Vector GEP must have vector indices!", &GEP);
+    Assert1(GEP.getType()->isVectorTy(),
+      "Vector GEP must return a vector value", &GEP);
+    Type *ElemPtr = cast<VectorType>(GEP.getType())->getElementType();
+    Assert1(ElemPtr->isPointerTy(),
+      "Vector GEP pointer operand is not a pointer!", &GEP);
+    unsigned IndexWidth = cast<VectorType>(IndexTy)->getNumElements();
+    unsigned GepWidth = cast<VectorType>(GEP.getType())->getNumElements();
+    Assert1(IndexWidth == GepWidth, "Invalid GEP index vector width", &GEP);
+    Assert1(ElTy == cast<PointerType>(ElemPtr)->getElementType(),
+      "Vector GEP type does not match pointer type!", &GEP);
+  }
   visitInstruction(GEP);
 }
 
@@ -1642,6 +1688,12 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallInst &CI) {
   switch (ID) {
   default:
     break;
+  case Intrinsic::ctlz:  // llvm.ctlz
+  case Intrinsic::cttz:  // llvm.cttz
+    Assert1(isa<ConstantInt>(CI.getArgOperand(1)),
+            "is_zero_undef argument of bit counting intrinsics must be a "
+            "constant int", &CI);
+    break;
   case Intrinsic::dbg_declare: {  // llvm.dbg.declare
     Assert1(CI.getArgOperand(0) && isa<MDNode>(CI.getArgOperand(0)),
                 "invalid llvm.dbg.declare intrinsic call 1", &CI);