Index: llvm/trunk/lib/Transforms/Utils/LowerMemIntrinsics.cpp
===================================================================
--- llvm/trunk/lib/Transforms/Utils/LowerMemIntrinsics.cpp
+++ llvm/trunk/lib/Transforms/Utils/LowerMemIntrinsics.cpp
@@ -27,8 +27,11 @@
   BasicBlock *LoopBB = BasicBlock::Create(F->getContext(), "loadstoreloop",
                                           F, NewBB);
 
-  OrigBB->getTerminator()->setSuccessor(0, LoopBB);
   IRBuilder<> Builder(OrigBB->getTerminator());
+  Builder.CreateCondBr(
+      Builder.CreateICmpEQ(ConstantInt::get(TypeOfCopyLen, 0), CopyLen), NewBB,
+      LoopBB);
+  OrigBB->getTerminator()->eraseFromParent();
 
   // SrcAddr and DstAddr are expected to be pointer types,
   // so no check is made here.
@@ -167,6 +170,7 @@
 static void createMemSetLoop(Instruction *InsertBefore,
                              Value *DstAddr, Value *CopyLen, Value *SetValue,
                              unsigned Align, bool IsVolatile) {
+  Type *TypeOfCopyLen = CopyLen->getType();
   BasicBlock *OrigBB = InsertBefore->getParent();
   Function *F = OrigBB->getParent();
   BasicBlock *NewBB =
@@ -174,8 +178,11 @@
   BasicBlock *LoopBB
     = BasicBlock::Create(F->getContext(), "loadstoreloop", F, NewBB);
 
-  OrigBB->getTerminator()->setSuccessor(0, LoopBB);
   IRBuilder<> Builder(OrigBB->getTerminator());
+  Builder.CreateCondBr(
+      Builder.CreateICmpEQ(ConstantInt::get(TypeOfCopyLen, 0), CopyLen), NewBB,
+      LoopBB);
+  OrigBB->getTerminator()->eraseFromParent();
 
   // Cast pointer to the type of value getting stored
   unsigned dstAS = cast<PointerType>(DstAddr->getType())->getAddressSpace();
@@ -183,8 +190,8 @@
                                   PointerType::get(SetValue->getType(), dstAS));
 
   IRBuilder<> LoopBuilder(LoopBB);
-  PHINode *LoopIndex = LoopBuilder.CreatePHI(CopyLen->getType(), 0);
-  LoopIndex->addIncoming(ConstantInt::get(CopyLen->getType(), 0), OrigBB);
+  PHINode *LoopIndex = LoopBuilder.CreatePHI(TypeOfCopyLen, 0);
+  LoopIndex->addIncoming(ConstantInt::get(TypeOfCopyLen, 0), OrigBB);
 
   LoopBuilder.CreateStore(
       SetValue,
@@ -192,7 +199,7 @@
       IsVolatile);
 
   Value *NewIndex =
-      LoopBuilder.CreateAdd(LoopIndex, ConstantInt::get(CopyLen->getType(), 1));
+      LoopBuilder.CreateAdd(LoopIndex, ConstantInt::get(TypeOfCopyLen, 1));
   LoopIndex->addIncoming(NewIndex, LoopBB);
 
   LoopBuilder.CreateCondBr(LoopBuilder.CreateICmpULT(NewIndex, CopyLen), LoopBB,
Index: llvm/trunk/test/CodeGen/NVPTX/lower-aggr-copies.ll
===================================================================
--- llvm/trunk/test/CodeGen/NVPTX/lower-aggr-copies.ll
+++ llvm/trunk/test/CodeGen/NVPTX/lower-aggr-copies.ll
@@ -17,6 +17,8 @@
   ret i8* %dst
 
 ; IR-LABEL:   @memcpy_caller
+; IR:         [[CMPREG:%[0-9]+]] = icmp eq i64 0, %n
+; IR:         br i1 [[CMPREG]], label %split, label %loadstoreloop
 ; IR:         loadstoreloop:
 ; IR:         [[LOADPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %src, i64
 ; IR-NEXT:    [[VAL:%[0-9]+]] = load i8, i8* [[LOADPTR]]
@@ -73,6 +75,8 @@
 
 ; IR-LABEL:   @memset_caller
 ; IR:         [[VAL:%[0-9]+]] = trunc i32 %c to i8
+; IR:         [[CMPREG:%[0-9]+]] = icmp eq i64 0, %n
+; IR:         br i1 [[CMPREG]], label %split, label %loadstoreloop
 ; IR:         loadstoreloop:
 ; IR:         [[STOREPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64
 ; IR-NEXT:    store i8 [[VAL]], i8* [[STOREPTR]]