diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 35acc7923879..68be3b7c77fd 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -332,7 +332,7 @@ namespace {
     bool processMemCpy(MemCpyInst *M);
     bool processMemMove(MemMoveInst *M);
     bool performCallSlotOptzn(Instruction *cpy, Value *cpyDst, Value *cpySrc,
-                              uint64_t cpyLen, CallInst *C);
+                              uint64_t cpyLen, unsigned cpyAlign, CallInst *C);
     bool processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep,
                                        uint64_t MSize);
     bool processByValArgument(CallSite CS, unsigned ArgNo);
@@ -509,10 +509,18 @@ bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
       }
 
       if (C) {
+        unsigned storeAlign = SI->getAlignment();
+        if (!storeAlign)
+          storeAlign = TD->getABITypeAlignment(SI->getOperand(0)->getType());
+        unsigned loadAlign = LI->getAlignment();
+        if (!loadAlign)
+          loadAlign = TD->getABITypeAlignment(LI->getType());
+
         bool changed = performCallSlotOptzn(LI,
                         SI->getPointerOperand()->stripPointerCasts(),
                         LI->getPointerOperand()->stripPointerCasts(),
-                        TD->getTypeStoreSize(SI->getOperand(0)->getType()), C);
+                        TD->getTypeStoreSize(SI->getOperand(0)->getType()),
+                        std::min(storeAlign, loadAlign), C);
         if (changed) {
           MD->removeInstruction(SI);
           SI->eraseFromParent();
@@ -559,7 +567,8 @@ bool MemCpyOpt::processMemSet(MemSetInst *MSI, BasicBlock::iterator &BBI) {
 /// the call write its result directly into the destination of the memcpy.
 bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
                                      Value *cpyDest, Value *cpySrc,
-                                     uint64_t cpyLen, CallInst *C) {
+                                     uint64_t cpyLen, unsigned cpyAlign,
+                                     CallInst *C) {
   // The general transformation to keep in mind is
   //
   //   call @func(..., src, ...)
@@ -596,6 +605,16 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
   if (cpyLen < srcSize)
     return false;
 
+  // Check that dest points to memory that is at least as aligned as src.
+  unsigned srcAlign = srcAlloca->getAlignment();
+  if (!srcAlign)
+    srcAlign = TD->getABITypeAlignment(srcAlloca->getAllocatedType());
+  bool isDestSufficientlyAligned = srcAlign <= cpyAlign;
+  // If dest is not aligned enough and we can't increase its alignment then
+  // bail out.
+  if (!isDestSufficientlyAligned && !isa<AllocaInst>(cpyDest))
+    return false;
+
   // Check that accessing the first srcSize bytes of dest will not cause a
   // trap.  Otherwise the transform is invalid since it might cause a trap
   // to occur earlier than it otherwise would.
@@ -687,6 +706,12 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
   if (!changedArgument)
     return false;
 
+  // If the destination wasn't sufficiently aligned then increase its alignment.
+  if (!isDestSufficientlyAligned) {
+    assert(isa<AllocaInst>(cpyDest) && "Can only increase alloca alignment!");
+    cast<AllocaInst>(cpyDest)->setAlignment(srcAlign);
+  }
+
   // Drop any cached information about the call, because we may have changed
   // its dependence information by changing its parameter.
   MD->removeInstruction(C);
@@ -813,7 +838,8 @@ bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
   if (DepInfo.isClobber()) {
     if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
       if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
-                               CopySize->getZExtValue(), C)) {
+                               CopySize->getZExtValue(), M->getAlignment(),
+                               C)) {
         MD->removeInstruction(M);
         M->eraseFromParent();
         return true;
diff --git a/llvm/test/Transforms/MemCpyOpt/align.ll b/llvm/test/Transforms/MemCpyOpt/align.ll
index b1f900d9da4c..1b98f6ad383f 100644
--- a/llvm/test/Transforms/MemCpyOpt/align.ll
+++ b/llvm/test/Transforms/MemCpyOpt/align.ll
@@ -1,12 +1,15 @@
-; RUN: opt < %s -S -memcpyopt | FileCheck %s
+; RUN: opt < %s -S -basicaa -memcpyopt | FileCheck %s
 target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64"
 
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind
+
 ; The resulting memset is only 4-byte aligned, despite containing
 ; a 16-byte aligned store in the middle.
 
-; CHECK: call void @llvm.memset.p0i8.i64(i8* {{.*}}, i8 0, i64 16, i32 4, i1 false)
-
 define void @foo(i32* %p) {
+; CHECK: @foo
+; CHECK: call void @llvm.memset.p0i8.i64(i8* {{.*}}, i8 0, i64 16, i32 4, i1 false)
   %a0 = getelementptr i32* %p, i64 0
   store i32 0, i32* %a0, align 4
   %a1 = getelementptr i32* %p, i64 1
@@ -17,3 +20,18 @@ define void @foo(i32* %p) {
   store i32 0, i32* %a3, align 4
   ret void
 }
+
+; Replacing %a8 with %a4 in the memset requires boosting the alignment of %a4.
+
+define void @bar() {
+; CHECK: @bar
+; CHECK: %a4 = alloca i32, align 8
+; CHECK-NOT: memcpy
+  %a4 = alloca i32, align 4
+  %a8 = alloca i32, align 8
+  %a8.cast = bitcast i32* %a8 to i8*
+  %a4.cast = bitcast i32* %a4 to i8*
+  call void @llvm.memset.p0i8.i64(i8* %a8.cast, i8 0, i64 4, i32 8, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a4.cast, i8* %a8.cast, i64 4, i32 4, i1 false)
+  ret void
+}