Add in the first iteration of support for llvm/clang/lldb to allow variable per address space pointer sizes to be optimized correctly.

llvm-svn: 165726

2 files changed, 13 insertions, 6 deletions

diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 517657cf526..97fff9edd68 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -174,10 +174,11 @@ bool MemsetRange::isProfitableToUseMemset(const DataLayout &TD) const {
   // this width can be stored.  If so, check to see whether we will end up
   // actually reducing the number of stores used.
   unsigned Bytes = unsigned(End-Start);
-  unsigned NumPointerStores = Bytes/TD.getPointerSize();
+  unsigned AS = cast<StoreInst>(TheStores[0])->getPointerAddressSpace();
+  unsigned NumPointerStores = Bytes/TD.getPointerSize(AS);
 
   // Assume the remaining bytes if any are done a byte at a time.
-  unsigned NumByteStores = Bytes - NumPointerStores*TD.getPointerSize();
+  unsigned NumByteStores = Bytes - NumPointerStores*TD.getPointerSize(AS);
 
   // If we will reduce the # stores (according to this heuristic), do the
   // transformation.  This encourages merging 4 x i8 -> i32 and 2 x i16 -> i32
diff --git a/llvm/lib/Transforms/Scalar/SROA.cpp b/llvm/lib/Transforms/Scalar/SROA.cpp
index ca762514929..f43bb96facc 100644
--- a/llvm/lib/Transforms/Scalar/SROA.cpp
+++ b/llvm/lib/Transforms/Scalar/SROA.cpp
@@ -447,6 +447,7 @@ protected:
 
   bool computeConstantGEPOffset(GetElementPtrInst &GEPI, int64_t &GEPOffset) {
     GEPOffset = Offset;
+    unsigned int AS = GEPI.getPointerAddressSpace();
     for (gep_type_iterator GTI = gep_type_begin(GEPI), GTE = gep_type_end(GEPI);
          GTI != GTE; ++GTI) {
       ConstantInt *OpC = dyn_cast<ConstantInt>(GTI.getOperand());
@@ -476,7 +477,7 @@ protected:
         continue;
       }
 
-      APInt Index = OpC->getValue().sextOrTrunc(TD.getPointerSizeInBits());
+      APInt Index = OpC->getValue().sextOrTrunc(TD.getPointerSizeInBits(AS));
       Index *= APInt(Index.getBitWidth(),
                      TD.getTypeAllocSize(GTI.getIndexedType()));
       Index += APInt(Index.getBitWidth(), (uint64_t)GEPOffset,
@@ -1784,7 +1785,9 @@ static Value *getNaturalGEPWithType(IRBuilder<> &IRB, const DataLayout &TD,
       break;
     if (SequentialType *SeqTy = dyn_cast<SequentialType>(ElementTy)) {
       ElementTy = SeqTy->getElementType();
-      Indices.push_back(IRB.getInt(APInt(TD.getPointerSizeInBits(), 0)));
+      Indices.push_back(IRB.getInt(APInt(TD.getPointerSizeInBits(
+                ElementTy->isPointerTy() ? 
+                cast<PointerType>(ElementTy)->getAddressSpace(): 0), 0)));
     } else if (StructType *STy = dyn_cast<StructType>(ElementTy)) {
       if (STy->element_begin() == STy->element_end())
         break; // Nothing left to descend into.
@@ -2239,7 +2242,8 @@ private:
 
   Value *getAdjustedAllocaPtr(IRBuilder<> &IRB, Type *PointerTy) {
     assert(BeginOffset >= NewAllocaBeginOffset);
-    APInt Offset(TD.getPointerSizeInBits(), BeginOffset - NewAllocaBeginOffset);
+    unsigned AS = cast<PointerType>(PointerTy)->getAddressSpace();
+    APInt Offset(TD.getPointerSizeInBits(AS), BeginOffset - NewAllocaBeginOffset);
     return getAdjustedPtr(IRB, TD, &NewAI, Offset, PointerTy, getName(""));
   }
 
@@ -2578,8 +2582,10 @@ private:
     const AllocaPartitioning::MemTransferOffsets &MTO
       = P.getMemTransferOffsets(II);
 
+    assert(OldPtr->getType()->isPointerTy() && "Must be a pointer type!");
+    unsigned AS = cast<PointerType>(OldPtr->getType())->getAddressSpace();
     // Compute the relative offset within the transfer.
-    unsigned IntPtrWidth = TD.getPointerSizeInBits();
+    unsigned IntPtrWidth = TD.getPointerSizeInBits(AS);
     APInt RelOffset(IntPtrWidth, BeginOffset - (IsDest ? MTO.DestBegin
                                                        : MTO.SourceBegin));