Fix various issues (or do cleanups) found by enabling certain MSVC warnings.

[oota-llvm.git] / lib / Transforms / Scalar / MemCpyOptimizer.cpp

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index be5aa2ea5832feb0c5047fe9e0dd9732de826796..a87cce3f9d3eda7dee5573979a3c19754c13c98a 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -54,7 +54,7 @@ static int64_t GetOffsetFromIndex(const GetElementPtrInst *GEP, unsigned Idx,
     if (OpC->isZero()) continue;  // No offset.
 
     // Handle struct indices, which add their field offset to the pointer.
-    if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
+    if (StructType *STy = dyn_cast<StructType>(*GTI)) {
       Offset += TD.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
       continue;
     }
@@ -147,8 +147,8 @@ struct MemsetRange {
 } // end anon namespace
 
 bool MemsetRange::isProfitableToUseMemset(const TargetData &TD) const {
-  // If we found more than 8 stores to merge or 64 bytes, use memset.
-  if (TheStores.size() >= 8 || End-Start >= 64) return true;
+  // If we found more than 4 stores to merge or 16 bytes, use memset.
+  if (TheStores.size() >= 4 || End-Start >= 16) return true;
 
   // If there is nothing to merge, don't do anything.
   if (TheStores.size() < 2) return false;
@@ -384,7 +384,7 @@ Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
     
     if (StoreInst *NextStore = dyn_cast<StoreInst>(BI)) {
       // If this is a store, see if we can merge it in.
-      if (NextStore->isVolatile()) break;
+      if (!NextStore->isSimple()) break;
     
       // Check to see if this stored value is of the same byte-splattable value.
       if (ByteVal != isBytewiseValue(NextStore->getOperand(0)))
@@ -448,7 +448,7 @@ Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
     // Determine alignment
     unsigned Alignment = Range.Alignment;
     if (Alignment == 0) {
-      const Type *EltType = 
+      Type *EltType = 
         cast<PointerType>(StartPtr->getType())->getElementType();
       Alignment = TD->getABITypeAlignment(EltType);
     }
@@ -479,7 +479,7 @@ Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
 
 
 bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
-  if (SI->isVolatile()) return false;
+  if (!SI->isSimple()) return false;
   
   if (TD == 0) return false;
 
@@ -487,7 +487,8 @@ bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
   // happen to be using a load-store pair to implement it, rather than
   // a memcpy.
   if (LoadInst *LI = dyn_cast<LoadInst>(SI->getOperand(0))) {
-    if (!LI->isVolatile() && LI->hasOneUse()) {
+    if (LI->isSimple() && LI->hasOneUse() &&
+        LI->getParent() == SI->getParent()) {
       MemDepResult ldep = MD->getDependency(LI);
       CallInst *C = 0;
       if (ldep.isClobber() && !isa<MemCpyInst>(ldep.getInst()))
@@ -496,17 +497,14 @@ bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
       if (C) {
         // Check that nothing touches the dest of the "copy" between
         // the call and the store.
-        MemDepResult sdep = MD->getDependency(SI);
-        if (!sdep.isNonLocal()) {
-          bool FoundCall = false;
-          for (BasicBlock::iterator I = SI, E = sdep.getInst(); I != E; --I) {
-            if (&*I == C) {
-              FoundCall = true;
-              break;
-            }
-          }
-          if (!FoundCall)
+        AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
+        AliasAnalysis::Location StoreLoc = AA.getLocation(SI);
+        for (BasicBlock::iterator I = --BasicBlock::iterator(SI),
+                                  E = C; I != E; --I) {
+          if (AA.getModRefInfo(&*I, StoreLoc) != AliasAnalysis::NoModRef) {
             C = 0;
+            break;
+          }
         }
       }
 
@@ -618,7 +616,7 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
     if (!A->hasStructRetAttr())
       return false;
 
-    const Type *StructTy = cast<PointerType>(A->getType())->getElementType();
+    Type *StructTy = cast<PointerType>(A->getType())->getElementType();
     uint64_t destSize = TD->getTypeAllocSize(StructTy);
 
     if (destSize < srcSize)
@@ -808,21 +806,25 @@ bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
   //   a) memcpy-memcpy xform which exposes redundance for DSE.
   //   b) call-memcpy xform for return slot optimization.
   MemDepResult DepInfo = MD->getDependency(M);
-  if (!DepInfo.isClobber())
-    return false;
-  
-  if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(DepInfo.getInst()))
-    return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
-    
-  if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
-    if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
-                             CopySize->getZExtValue(), C)) {
-      MD->removeInstruction(M);
-      M->eraseFromParent();
-      return true;
+  if (DepInfo.isClobber()) {
+    if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
+      if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
+                               CopySize->getZExtValue(), C)) {
+        MD->removeInstruction(M);
+        M->eraseFromParent();
+        return true;
+      }
     }
   }
-  
+
+  AliasAnalysis::Location SrcLoc = AliasAnalysis::getLocationForSource(M);
+  MemDepResult SrcDepInfo = MD->getPointerDependencyFrom(SrcLoc, true,
+                                                         M, M->getParent());
+  if (SrcDepInfo.isClobber()) {
+    if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(SrcDepInfo.getInst()))
+      return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
+  }
+
   return false;
 }
 
@@ -842,11 +844,11 @@ bool MemCpyOpt::processMemMove(MemMoveInst *M) {
   
   // If not, then we know we can transform this.
   Module *Mod = M->getParent()->getParent()->getParent();
-  const Type *ArgTys[3] = { M->getRawDest()->getType(),
-                            M->getRawSource()->getType(),
-                            M->getLength()->getType() };
+  Type *ArgTys[3] = { M->getRawDest()->getType(),
+                      M->getRawSource()->getType(),
+                      M->getLength()->getType() };
   M->setCalledFunction(Intrinsic::getDeclaration(Mod, Intrinsic::memcpy,
-                                                 ArgTys, 3));
+                                                 ArgTys));
 
   // MemDep may have over conservative information about this instruction, just
   // conservatively flush it from the cache.
@@ -862,7 +864,7 @@ bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
 
   // Find out what feeds this byval argument.
   Value *ByValArg = CS.getArgument(ArgNo);
-  const Type *ByValTy =cast<PointerType>(ByValArg->getType())->getElementType();
+  Type *ByValTy = cast<PointerType>(ByValArg->getType())->getElementType();
   uint64_t ByValSize = TD->getTypeAllocSize(ByValTy);
   MemDepResult DepInfo =
     MD->getPointerDependencyFrom(AliasAnalysis::Location(ByValArg, ByValSize),
@@ -947,7 +949,7 @@ bool MemCpyOpt::iterateOnFunction(Function &F) {
         RepeatInstruction = processMemMove(M);
       else if (CallSite CS = (Value*)I) {
         for (unsigned i = 0, e = CS.arg_size(); i != e; ++i)
-          if (CS.paramHasAttr(i+1, Attribute::ByVal))
+          if (CS.isByValArgument(i))
             MadeChange |= processByValArgument(CS, i);
       }