1 files changed, 73 insertions, 91 deletions

diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 77e4727..d34fab1 100644
--- a/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -172,8 +172,6 @@ Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
 Instruction *InstCombiner::visitCallInst(CallInst &CI) {
   if (isFreeCall(&CI))
     return visitFree(CI);
-  if (isMalloc(&CI))
-    return visitMalloc(CI);
 
   // If the caller function is nounwind, mark the call as nounwind, even if the
   // callee isn't.
@@ -246,78 +244,10 @@ 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) break;
-
-    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()) break;
-
-      // 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()) break;
-    }
-
-    // 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) break;
-          Size *= C->getZExtValue();
-        }
-      }
-    } else if (CallInst *MI = extractMallocCall(Op1)) {
-      // Get allocation size.
-      Type* MallocType = getMallocAllocatedType(MI);
-      if (MallocType && MallocType->isSized())
-        if (Value *NElems = getMallocArraySize(MI, TD, true))
-          if (ConstantInt *NElements = dyn_cast<ConstantInt>(NElems))
-            Size = NElements->getZExtValue() * TD->getTypeAllocSize(MallocType);
-    }
-
-    // Do not return "I don't know" here. Later optimization passes could
-    // make it possible to evaluate objectsize to a constant.
-    if (Size == -1ULL)
-      break;
-
-    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
@@ -694,6 +624,57 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
     break;
   }
 
+  case Intrinsic::arm_neon_vmulls:
+  case Intrinsic::arm_neon_vmullu: {
+    Value *Arg0 = II->getArgOperand(0);
+    Value *Arg1 = II->getArgOperand(1);
+
+    // Handle mul by zero first:
+    if (isa<ConstantAggregateZero>(Arg0) || isa<ConstantAggregateZero>(Arg1)) {
+      return ReplaceInstUsesWith(CI, ConstantAggregateZero::get(II->getType()));
+    }
+
+    // Check for constant LHS & RHS - in this case we just simplify.
+    bool Zext = (II->getIntrinsicID() == Intrinsic::arm_neon_vmullu);
+    VectorType *NewVT = cast<VectorType>(II->getType());
+    unsigned NewWidth = NewVT->getElementType()->getIntegerBitWidth();
+    if (ConstantDataVector *CV0 = dyn_cast<ConstantDataVector>(Arg0)) {
+      if (ConstantDataVector *CV1 = dyn_cast<ConstantDataVector>(Arg1)) {
+        VectorType* VT = cast<VectorType>(CV0->getType());
+        SmallVector<Constant*, 4> NewElems;
+        for (unsigned i = 0; i < VT->getNumElements(); ++i) {
+          APInt CV0E =
+            (cast<ConstantInt>(CV0->getAggregateElement(i)))->getValue();
+          CV0E = Zext ? CV0E.zext(NewWidth) : CV0E.sext(NewWidth);
+          APInt CV1E =
+            (cast<ConstantInt>(CV1->getAggregateElement(i)))->getValue();
+          CV1E = Zext ? CV1E.zext(NewWidth) : CV1E.sext(NewWidth);
+          NewElems.push_back(
+            ConstantInt::get(NewVT->getElementType(), CV0E * CV1E));
+        }
+        return ReplaceInstUsesWith(CI, ConstantVector::get(NewElems));
+      }
+
+      // Couldn't simplify - cannonicalize constant to the RHS.
+      std::swap(Arg0, Arg1);
+    }
+
+    // Handle mul by one:
+    if (ConstantDataVector *CV1 = dyn_cast<ConstantDataVector>(Arg1)) {
+      if (ConstantInt *Splat =
+            dyn_cast_or_null<ConstantInt>(CV1->getSplatValue())) {
+        if (Splat->isOne()) {
+          if (Zext)
+            return CastInst::CreateZExtOrBitCast(Arg0, II->getType());
+          // else    
+          return CastInst::CreateSExtOrBitCast(Arg0, II->getType());
+        }
+      }
+    }
+
+    break;
+  }
+
   case Intrinsic::stackrestore: {
     // If the save is right next to the restore, remove the restore.  This can
     // happen when variable allocas are DCE'd.
@@ -711,7 +692,7 @@ 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;
       }
@@ -814,7 +795,7 @@ Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const TargetData *TD) {
   if (CI->getCalledFunction() == 0) return 0;
 
   InstCombineFortifiedLibCalls Simplifier(this);
-  Simplifier.fold(CI, TD);
+  Simplifier.fold(CI, TD, TLI);
   return Simplifier.NewInstruction;
 }
 
@@ -898,6 +879,9 @@ static IntrinsicInst *FindInitTrampoline(Value *Callee) {
 // visitCallSite - Improvements for call and invoke instructions.
 //
 Instruction *InstCombiner::visitCallSite(CallSite CS) {
+  if (isAllocLikeFn(CS.getInstruction()))
+    return visitAllocSite(*CS.getInstruction());
+
   bool Changed = false;
 
   // If the callee is a pointer to a function, attempt to move any casts to the
@@ -933,24 +917,24 @@ Instruction *InstCombiner::visitCallSite(CallSite CS) {
     }
 
   if (isa<ConstantPointerNull>(Callee) || isa<UndefValue>(Callee)) {
-    // This instruction is not reachable, just remove it.  We insert a store to
-    // undef so that we know that this code is not reachable, despite the fact
-    // that we can't modify the CFG here.
-    new StoreInst(ConstantInt::getTrue(Callee->getContext()),
-               UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
-                  CS.getInstruction());
-
     // If CS does not return void then replaceAllUsesWith undef.
     // This allows ValueHandlers and custom metadata to adjust itself.
     if (!CS.getInstruction()->getType()->isVoidTy())
       ReplaceInstUsesWith(*CS.getInstruction(),
                           UndefValue::get(CS.getInstruction()->getType()));
 
-    if (InvokeInst *II = dyn_cast<InvokeInst>(CS.getInstruction())) {
-      // Don't break the CFG, insert a dummy cond branch.
-      BranchInst::Create(II->getNormalDest(), II->getUnwindDest(),
-                         ConstantInt::getTrue(Callee->getContext()), II);
+    if (isa<InvokeInst>(CS.getInstruction())) {
+      // Can't remove an invoke because we cannot change the CFG.
+      return 0;
     }
+
+    // This instruction is not reachable, just remove it.  We insert a store to
+    // undef so that we know that this code is not reachable, despite the fact
+    // that we can't modify the CFG here.
+    new StoreInst(ConstantInt::getTrue(Callee->getContext()),
+                  UndefValue::get(Type::getInt1PtrTy(Callee->getContext())),
+                  CS.getInstruction());
+
     return EraseInstFromFunction(*CS.getInstruction());
   }
 
@@ -1194,8 +1178,7 @@ bool InstCombiner::transformConstExprCastCall(CallSite CS) {
   if (NewRetTy->isVoidTy())
     Caller->setName("");   // Void type should not have a name.
 
-  const AttrListPtr &NewCallerPAL = AttrListPtr::get(attrVec.begin(),
-                                                     attrVec.end());
+  const AttrListPtr &NewCallerPAL = AttrListPtr::get(attrVec);
 
   Instruction *NC;
   if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {
@@ -1367,8 +1350,7 @@ InstCombiner::transformCallThroughTrampoline(CallSite CS,
         NestF->getType() == PointerType::getUnqual(NewFTy) ?
         NestF : ConstantExpr::getBitCast(NestF,
                                          PointerType::getUnqual(NewFTy));
-      const AttrListPtr &NewPAL = AttrListPtr::get(NewAttrs.begin(),
-                                                   NewAttrs.end());
+      const AttrListPtr &NewPAL = AttrListPtr::get(NewAttrs);
 
       Instruction *NewCaller;
       if (InvokeInst *II = dyn_cast<InvokeInst>(Caller)) {