diff --git a/clang/lib/CodeGen/CGBuilder.h b/clang/lib/CodeGen/CGBuilder.h
--- a/clang/lib/CodeGen/CGBuilder.h
+++ b/clang/lib/CodeGen/CGBuilder.h
@@ -132,6 +132,28 @@
     return CreateAlignedStore(getInt1(Value), Addr, CharUnits::One());
   }
 
+  // Temporarily use old signature; clang will be updated to an Address overload
+  // in a subsequent patch.
+  llvm::AtomicCmpXchgInst *
+  CreateAtomicCmpXchg(llvm::Value *Ptr, llvm::Value *Cmp, llvm::Value *New,
+                      llvm::AtomicOrdering SuccessOrdering,
+                      llvm::AtomicOrdering FailureOrdering,
+                      llvm::SyncScope::ID SSID = llvm::SyncScope::System) {
+    return CGBuilderBaseTy::CreateAtomicCmpXchg(
+        Ptr, Cmp, New, llvm::MaybeAlign(), SuccessOrdering, FailureOrdering,
+        SSID);
+  }
+
+  // Temporarily use old signature; clang will be updated to an Address overload
+  // in a subsequent patch.
+  llvm::AtomicRMWInst *
+  CreateAtomicRMW(llvm::AtomicRMWInst::BinOp Op, llvm::Value *Ptr,
+                  llvm::Value *Val, llvm::AtomicOrdering Ordering,
+                  llvm::SyncScope::ID SSID = llvm::SyncScope::System) {
+    return CGBuilderBaseTy::CreateAtomicRMW(Op, Ptr, Val, llvm::MaybeAlign(),
+                                            Ordering, SSID);
+  }
+
   using CGBuilderBaseTy::CreateBitCast;
   Address CreateBitCast(Address Addr, llvm::Type *Ty,
                         const llvm::Twine &Name = "") {
diff --git a/llvm/include/llvm/CodeGen/AtomicExpandUtils.h b/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
--- a/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
+++ b/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
@@ -22,8 +22,8 @@
 /// (the builder, %addr, %loaded, %new_val, ordering,
 ///  /* OUT */ %success, /* OUT */ %new_loaded)
 using CreateCmpXchgInstFun =
-    function_ref<void(IRBuilder<> &, Value *, Value *, Value *, AtomicOrdering,
-                      Value *&, Value *&)>;
+    function_ref<void(IRBuilder<> &, Value *, Value *, Value *, Align,
+                      AtomicOrdering, Value *&, Value *&)>;
 
 /// Expand an atomic RMW instruction into a loop utilizing
 /// cmpxchg. You'll want to make sure your target machine likes cmpxchg
diff --git a/llvm/include/llvm/IR/IRBuilder.h b/llvm/include/llvm/IR/IRBuilder.h
--- a/llvm/include/llvm/IR/IRBuilder.h
+++ b/llvm/include/llvm/IR/IRBuilder.h
@@ -1704,21 +1704,30 @@
     return Insert(new FenceInst(Context, Ordering, SSID), Name);
   }
 
-  AtomicCmpXchgInst *CreateAtomicCmpXchg(
-      Value *Ptr, Value *Cmp, Value *New, AtomicOrdering SuccessOrdering,
-      AtomicOrdering FailureOrdering, SyncScope::ID SSID = SyncScope::System) {
-    const DataLayout &DL = BB->getModule()->getDataLayout();
-    Align Alignment(DL.getTypeStoreSize(New->getType()));
-    return Insert(new AtomicCmpXchgInst(
-        Ptr, Cmp, New, Alignment, SuccessOrdering, FailureOrdering, SSID));
+  AtomicCmpXchgInst *
+  CreateAtomicCmpXchg(Value *Ptr, Value *Cmp, Value *New, MaybeAlign Align,
+                      AtomicOrdering SuccessOrdering,
+                      AtomicOrdering FailureOrdering,
+                      SyncScope::ID SSID = SyncScope::System) {
+    if (!Align) {
+      const DataLayout &DL = BB->getModule()->getDataLayout();
+      Align = llvm::Align(DL.getTypeStoreSize(New->getType()));
+    }
+
+    return Insert(new AtomicCmpXchgInst(Ptr, Cmp, New, *Align, SuccessOrdering,
+                                        FailureOrdering, SSID));
   }
 
-  AtomicRMWInst *CreateAtomicRMW(AtomicRMWInst::BinOp Op, Value *Ptr, Value *Val,
+  AtomicRMWInst *CreateAtomicRMW(AtomicRMWInst::BinOp Op, Value *Ptr,
+                                 Value *Val, MaybeAlign Align,
                                  AtomicOrdering Ordering,
                                  SyncScope::ID SSID = SyncScope::System) {
-    const DataLayout &DL = BB->getModule()->getDataLayout();
-    Align Alignment(DL.getTypeStoreSize(Val->getType()));
-    return Insert(new AtomicRMWInst(Op, Ptr, Val, Alignment, Ordering, SSID));
+    if (!Align) {
+      const DataLayout &DL = BB->getModule()->getDataLayout();
+      Align = llvm::Align(DL.getTypeStoreSize(Val->getType()));
+    }
+
+    return Insert(new AtomicRMWInst(Op, Ptr, Val, *Align, Ordering, SSID));
   }
 
   Value *CreateGEP(Value *Ptr, ArrayRef<Value *> IdxList,
diff --git a/llvm/lib/CodeGen/AtomicExpandPass.cpp b/llvm/lib/CodeGen/AtomicExpandPass.cpp
--- a/llvm/lib/CodeGen/AtomicExpandPass.cpp
+++ b/llvm/lib/CodeGen/AtomicExpandPass.cpp
@@ -80,10 +80,11 @@
     bool tryExpandAtomicRMW(AtomicRMWInst *AI);
     Value *
     insertRMWLLSCLoop(IRBuilder<> &Builder, Type *ResultTy, Value *Addr,
-                      AtomicOrdering MemOpOrder,
+                      Align AddrAlign, AtomicOrdering MemOpOrder,
                       function_ref<Value *(IRBuilder<> &, Value *)> PerformOp);
     void expandAtomicOpToLLSC(
-        Instruction *I, Type *ResultTy, Value *Addr, AtomicOrdering MemOpOrder,
+        Instruction *I, Type *ResultTy, Value *Addr, Align AddrAlign,
+        AtomicOrdering MemOpOrder,
         function_ref<Value *(IRBuilder<> &, Value *)> PerformOp);
     void expandPartwordAtomicRMW(
         AtomicRMWInst *I,
@@ -95,7 +96,7 @@
 
     AtomicCmpXchgInst *convertCmpXchgToIntegerType(AtomicCmpXchgInst *CI);
     static Value *insertRMWCmpXchgLoop(
-        IRBuilder<> &Builder, Type *ResultType, Value *Addr,
+        IRBuilder<> &Builder, Type *ResultType, Value *Addr, Align AddrAlign,
         AtomicOrdering MemOpOrder,
         function_ref<Value *(IRBuilder<> &, Value *)> PerformOp,
         CreateCmpXchgInstFun CreateCmpXchg);
@@ -368,7 +369,8 @@
     return false;
   case TargetLoweringBase::AtomicExpansionKind::LLSC:
     expandAtomicOpToLLSC(
-        LI, LI->getType(), LI->getPointerOperand(), LI->getOrdering(),
+        LI, LI->getType(), LI->getPointerOperand(), LI->getAlign(),
+        LI->getOrdering(),
         [](IRBuilder<> &Builder, Value *Loaded) { return Loaded; });
     return true;
   case TargetLoweringBase::AtomicExpansionKind::LLOnly:
@@ -407,7 +409,7 @@
   Constant *DummyVal = Constant::getNullValue(Ty);
 
   Value *Pair = Builder.CreateAtomicCmpXchg(
-      Addr, DummyVal, DummyVal, Order,
+      Addr, DummyVal, DummyVal, LI->getAlign(), Order,
       AtomicCmpXchgInst::getStrongestFailureOrdering(Order));
   Value *Loaded = Builder.CreateExtractValue(Pair, 0, "loaded");
 
@@ -454,9 +456,9 @@
   // It is the responsibility of the target to only signal expansion via
   // shouldExpandAtomicRMW in cases where this is required and possible.
   IRBuilder<> Builder(SI);
-  AtomicRMWInst *AI =
-      Builder.CreateAtomicRMW(AtomicRMWInst::Xchg, SI->getPointerOperand(),
-                              SI->getValueOperand(), SI->getOrdering());
+  AtomicRMWInst *AI = Builder.CreateAtomicRMW(
+      AtomicRMWInst::Xchg, SI->getPointerOperand(), SI->getValueOperand(),
+      SI->getAlign(), SI->getOrdering());
   SI->eraseFromParent();
 
   // Now we have an appropriate swap instruction, lower it as usual.
@@ -464,9 +466,9 @@
 }
 
 static void createCmpXchgInstFun(IRBuilder<> &Builder, Value *Addr,
-                                 Value *Loaded, Value *NewVal,
-                                 AtomicOrdering MemOpOrder,
-                                 Value *&Success, Value *&NewLoaded) {
+                                 Value *Loaded, Value *NewVal, Align AddrAlign,
+                                 AtomicOrdering MemOpOrder, Value *&Success,
+                                 Value *&NewLoaded) {
   Type *OrigTy = NewVal->getType();
 
   // This code can go away when cmpxchg supports FP types.
@@ -479,8 +481,8 @@
     Loaded = Builder.CreateBitCast(Loaded, IntTy);
   }
 
-  Value* Pair = Builder.CreateAtomicCmpXchg(
-      Addr, Loaded, NewVal, MemOpOrder,
+  Value *Pair = Builder.CreateAtomicCmpXchg(
+      Addr, Loaded, NewVal, AddrAlign, MemOpOrder,
       AtomicCmpXchgInst::getStrongestFailureOrdering(MemOpOrder));
   Success = Builder.CreateExtractValue(Pair, 1, "success");
   NewLoaded = Builder.CreateExtractValue(Pair, 0, "newloaded");
@@ -546,7 +548,7 @@
                                AI->getValOperand());
       };
       expandAtomicOpToLLSC(AI, AI->getType(), AI->getPointerOperand(),
-                           AI->getOrdering(), PerformOp);
+                           AI->getAlign(), AI->getOrdering(), PerformOp);
     }
     return true;
   }
@@ -581,6 +583,7 @@
   Type *WordType = nullptr;
   Type *ValueType = nullptr;
   Value *AlignedAddr = nullptr;
+  Align AlignedAddrAlignment;
   // The remaining fields can be null.
   Value *ShiftAmt = nullptr;
   Value *Mask = nullptr;
@@ -603,6 +606,7 @@
   PrintObj(PMV.ValueType);
   O << "  AlignedAddr: ";
   PrintObj(PMV.AlignedAddr);
+  O << "  AlignedAddrAlignment: " << PMV.AlignedAddrAlignment.value() << '\n';
   O << "  ShiftAmt: ";
   PrintObj(PMV.ShiftAmt);
   O << "  Mask: ";
@@ -633,6 +637,7 @@
 /// Inv_Mask: The inverse of Mask.
 static PartwordMaskValues createMaskInstrs(IRBuilder<> &Builder, Instruction *I,
                                            Type *ValueType, Value *Addr,
+                                           Align AddrAlign,
                                            unsigned MinWordSize) {
   PartwordMaskValues PMV;
 
@@ -646,6 +651,7 @@
                                          : ValueType;
   if (PMV.ValueType == PMV.WordType) {
     PMV.AlignedAddr = Addr;
+    PMV.AlignedAddrAlignment = AddrAlign;
     return PMV;
   }
 
@@ -654,10 +660,12 @@
   Type *WordPtrType =
       PMV.WordType->getPointerTo(Addr->getType()->getPointerAddressSpace());
 
+  // TODO: we could skip some of this if AddrAlign >= MinWordSize.
   Value *AddrInt = Builder.CreatePtrToInt(Addr, DL.getIntPtrType(Ctx));
   PMV.AlignedAddr = Builder.CreateIntToPtr(
       Builder.CreateAnd(AddrInt, ~(uint64_t)(MinWordSize - 1)), WordPtrType,
       "AlignedAddr");
+  PMV.AlignedAddrAlignment = Align(MinWordSize);
 
   Value *PtrLSB = Builder.CreateAnd(AddrInt, MinWordSize - 1, "PtrLSB");
   if (DL.isLittleEndian()) {
@@ -765,7 +773,7 @@
 
   PartwordMaskValues PMV =
       createMaskInstrs(Builder, AI, AI->getType(), AI->getPointerOperand(),
-                       TLI->getMinCmpXchgSizeInBits() / 8);
+                       AI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
 
   Value *ValOperand_Shifted =
       Builder.CreateShl(Builder.CreateZExt(AI->getValOperand(), PMV.WordType),
@@ -778,13 +786,14 @@
 
   Value *OldResult;
   if (ExpansionKind == TargetLoweringBase::AtomicExpansionKind::CmpXChg) {
-    OldResult =
-        insertRMWCmpXchgLoop(Builder, PMV.WordType, PMV.AlignedAddr, MemOpOrder,
-                             PerformPartwordOp, createCmpXchgInstFun);
+    OldResult = insertRMWCmpXchgLoop(Builder, PMV.WordType, PMV.AlignedAddr,
+                                     PMV.AlignedAddrAlignment, MemOpOrder,
+                                     PerformPartwordOp, createCmpXchgInstFun);
   } else {
     assert(ExpansionKind == TargetLoweringBase::AtomicExpansionKind::LLSC);
     OldResult = insertRMWLLSCLoop(Builder, PMV.WordType, PMV.AlignedAddr,
-                                  MemOpOrder, PerformPartwordOp);
+                                  PMV.AlignedAddrAlignment, MemOpOrder,
+                                  PerformPartwordOp);
   }
 
   Value *FinalOldResult = extractMaskedValue(Builder, OldResult, PMV);
@@ -803,7 +812,7 @@
 
   PartwordMaskValues PMV =
       createMaskInstrs(Builder, AI, AI->getType(), AI->getPointerOperand(),
-                       TLI->getMinCmpXchgSizeInBits() / 8);
+                       AI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
 
   Value *ValOperand_Shifted =
       Builder.CreateShl(Builder.CreateZExt(AI->getValOperand(), PMV.WordType),
@@ -817,8 +826,9 @@
   else
     NewOperand = ValOperand_Shifted;
 
-  AtomicRMWInst *NewAI = Builder.CreateAtomicRMW(Op, PMV.AlignedAddr,
-                                                 NewOperand, AI->getOrdering());
+  AtomicRMWInst *NewAI =
+      Builder.CreateAtomicRMW(Op, PMV.AlignedAddr, NewOperand,
+                              PMV.AlignedAddrAlignment, AI->getOrdering());
 
   Value *FinalOldResult = extractMaskedValue(Builder, NewAI, PMV);
   AI->replaceAllUsesWith(FinalOldResult);
@@ -871,8 +881,6 @@
   IRBuilder<> Builder(CI);
   LLVMContext &Ctx = Builder.getContext();
 
-  const int WordSize = TLI->getMinCmpXchgSizeInBits() / 8;
-
   BasicBlock *EndBB =
       BB->splitBasicBlock(CI->getIterator(), "partword.cmpxchg.end");
   auto FailureBB =
@@ -884,8 +892,9 @@
   std::prev(BB->end())->eraseFromParent();
   Builder.SetInsertPoint(BB);
 
-  PartwordMaskValues PMV = createMaskInstrs(
-      Builder, CI, CI->getCompareOperand()->getType(), Addr, WordSize);
+  PartwordMaskValues PMV =
+      createMaskInstrs(Builder, CI, CI->getCompareOperand()->getType(), Addr,
+                       CI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
 
   // Shift the incoming values over, into the right location in the word.
   Value *NewVal_Shifted =
@@ -909,8 +918,8 @@
   Value *FullWord_NewVal = Builder.CreateOr(Loaded_MaskOut, NewVal_Shifted);
   Value *FullWord_Cmp = Builder.CreateOr(Loaded_MaskOut, Cmp_Shifted);
   AtomicCmpXchgInst *NewCI = Builder.CreateAtomicCmpXchg(
-      PMV.AlignedAddr, FullWord_Cmp, FullWord_NewVal, CI->getSuccessOrdering(),
-      CI->getFailureOrdering(), CI->getSyncScopeID());
+      PMV.AlignedAddr, FullWord_Cmp, FullWord_NewVal, PMV.AlignedAddrAlignment,
+      CI->getSuccessOrdering(), CI->getFailureOrdering(), CI->getSyncScopeID());
   NewCI->setVolatile(CI->isVolatile());
   // When we're building a strong cmpxchg, we need a loop, so you
   // might think we could use a weak cmpxchg inside. But, using strong
@@ -953,11 +962,12 @@
 }
 
 void AtomicExpand::expandAtomicOpToLLSC(
-    Instruction *I, Type *ResultType, Value *Addr, AtomicOrdering MemOpOrder,
+    Instruction *I, Type *ResultType, Value *Addr, Align AddrAlign,
+    AtomicOrdering MemOpOrder,
     function_ref<Value *(IRBuilder<> &, Value *)> PerformOp) {
   IRBuilder<> Builder(I);
-  Value *Loaded =
-      insertRMWLLSCLoop(Builder, ResultType, Addr, MemOpOrder, PerformOp);
+  Value *Loaded = insertRMWLLSCLoop(Builder, ResultType, Addr, AddrAlign,
+                                    MemOpOrder, PerformOp);
 
   I->replaceAllUsesWith(Loaded);
   I->eraseFromParent();
@@ -968,7 +978,7 @@
 
   PartwordMaskValues PMV =
       createMaskInstrs(Builder, AI, AI->getType(), AI->getPointerOperand(),
-                       TLI->getMinCmpXchgSizeInBits() / 8);
+                       AI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
 
   // The value operand must be sign-extended for signed min/max so that the
   // target's signed comparison instructions can be used. Otherwise, just
@@ -994,7 +1004,7 @@
 
   PartwordMaskValues PMV = createMaskInstrs(
       Builder, CI, CI->getCompareOperand()->getType(), CI->getPointerOperand(),
-      TLI->getMinCmpXchgSizeInBits() / 8);
+      CI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
 
   Value *CmpVal_Shifted = Builder.CreateShl(
       Builder.CreateZExt(CI->getCompareOperand(), PMV.WordType), PMV.ShiftAmt,
@@ -1017,13 +1027,16 @@
 }
 
 Value *AtomicExpand::insertRMWLLSCLoop(
-    IRBuilder<> &Builder, Type *ResultTy, Value *Addr,
+    IRBuilder<> &Builder, Type *ResultTy, Value *Addr, Align AddrAlign,
     AtomicOrdering MemOpOrder,
     function_ref<Value *(IRBuilder<> &, Value *)> PerformOp) {
   LLVMContext &Ctx = Builder.getContext();
   BasicBlock *BB = Builder.GetInsertBlock();
   Function *F = BB->getParent();
 
+  assert(AddrAlign >= ResultTy->getPrimitiveSizeInBits() / 8 &&
+         "Expected at least natural alignment at this point.");
+
   // Given: atomicrmw some_op iN* %addr, iN %incr ordering
   //
   // The standard expansion we produce is:
@@ -1082,11 +1095,9 @@
   Value *NewCmp = Builder.CreatePtrToInt(CI->getCompareOperand(), NewTy);
   Value *NewNewVal = Builder.CreatePtrToInt(CI->getNewValOperand(), NewTy);
 
-
-  auto *NewCI = Builder.CreateAtomicCmpXchg(NewAddr, NewCmp, NewNewVal,
-                                            CI->getSuccessOrdering(),
-                                            CI->getFailureOrdering(),
-                                            CI->getSyncScopeID());
+  auto *NewCI = Builder.CreateAtomicCmpXchg(
+      NewAddr, NewCmp, NewNewVal, CI->getAlign(), CI->getSuccessOrdering(),
+      CI->getFailureOrdering(), CI->getSyncScopeID());
   NewCI->setVolatile(CI->isVolatile());
   NewCI->setWeak(CI->isWeak());
   LLVM_DEBUG(dbgs() << "Replaced " << *CI << " with " << *NewCI << "\n");
@@ -1211,7 +1222,7 @@
 
   PartwordMaskValues PMV =
       createMaskInstrs(Builder, CI, CI->getCompareOperand()->getType(), Addr,
-                       TLI->getMinCmpXchgSizeInBits() / 8);
+                       CI->getAlign(), TLI->getMinCmpXchgSizeInBits() / 8);
   Builder.CreateBr(StartBB);
 
   // Start the main loop block now that we've taken care of the preliminaries.
@@ -1379,7 +1390,7 @@
 }
 
 Value *AtomicExpand::insertRMWCmpXchgLoop(
-    IRBuilder<> &Builder, Type *ResultTy, Value *Addr,
+    IRBuilder<> &Builder, Type *ResultTy, Value *Addr, Align AddrAlign,
     AtomicOrdering MemOpOrder,
     function_ref<Value *(IRBuilder<> &, Value *)> PerformOp,
     CreateCmpXchgInstFun CreateCmpXchg) {
@@ -1411,9 +1422,7 @@
   // the branch entirely.
   std::prev(BB->end())->eraseFromParent();
   Builder.SetInsertPoint(BB);
-  LoadInst *InitLoaded = Builder.CreateLoad(ResultTy, Addr);
-  // Atomics require at least natural alignment.
-  InitLoaded->setAlignment(Align(ResultTy->getPrimitiveSizeInBits() / 8));
+  LoadInst *InitLoaded = Builder.CreateAlignedLoad(ResultTy, Addr, AddrAlign);
   Builder.CreateBr(LoopBB);
 
   // Start the main loop block now that we've taken care of the preliminaries.
@@ -1426,7 +1435,7 @@
   Value *NewLoaded = nullptr;
   Value *Success = nullptr;
 
-  CreateCmpXchg(Builder, Addr, Loaded, NewVal,
+  CreateCmpXchg(Builder, Addr, Loaded, NewVal, AddrAlign,
                 MemOpOrder == AtomicOrdering::Unordered
                     ? AtomicOrdering::Monotonic
                     : MemOpOrder,
@@ -1466,7 +1475,8 @@
                                     CreateCmpXchgInstFun CreateCmpXchg) {
   IRBuilder<> Builder(AI);
   Value *Loaded = AtomicExpand::insertRMWCmpXchgLoop(
-      Builder, AI->getType(), AI->getPointerOperand(), AI->getOrdering(),
+      Builder, AI->getType(), AI->getPointerOperand(), AI->getAlign(),
+      AI->getOrdering(),
       [&](IRBuilder<> &Builder, Value *Loaded) {
         return performAtomicOp(AI->getOperation(), Builder, Loaded,
                                AI->getValOperand());
@@ -1614,20 +1624,20 @@
   // libcalls (add/sub/etc) and we needed a generic. So, expand to a
   // CAS libcall, via a CAS loop, instead.
   if (!Success) {
-    expandAtomicRMWToCmpXchg(I, [this](IRBuilder<> &Builder, Value *Addr,
-                                       Value *Loaded, Value *NewVal,
-                                       AtomicOrdering MemOpOrder,
-                                       Value *&Success, Value *&NewLoaded) {
-      // Create the CAS instruction normally...
-      AtomicCmpXchgInst *Pair = Builder.CreateAtomicCmpXchg(
-          Addr, Loaded, NewVal, MemOpOrder,
-          AtomicCmpXchgInst::getStrongestFailureOrdering(MemOpOrder));
-      Success = Builder.CreateExtractValue(Pair, 1, "success");
-      NewLoaded = Builder.CreateExtractValue(Pair, 0, "newloaded");
-
-      // ...and then expand the CAS into a libcall.
-      expandAtomicCASToLibcall(Pair);
-    });
+    expandAtomicRMWToCmpXchg(
+        I, [this](IRBuilder<> &Builder, Value *Addr, Value *Loaded,
+                  Value *NewVal, Align Alignment, AtomicOrdering MemOpOrder,
+                  Value *&Success, Value *&NewLoaded) {
+          // Create the CAS instruction normally...
+          AtomicCmpXchgInst *Pair = Builder.CreateAtomicCmpXchg(
+              Addr, Loaded, NewVal, Alignment, MemOpOrder,
+              AtomicCmpXchgInst::getStrongestFailureOrdering(MemOpOrder));
+          Success = Builder.CreateExtractValue(Pair, 1, "success");
+          NewLoaded = Builder.CreateExtractValue(Pair, 0, "newloaded");
+
+          // ...and then expand the CAS into a libcall.
+          expandAtomicCASToLibcall(Pair);
+        });
   }
 }
 
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -3594,7 +3594,7 @@
                           Name.startswith("atomic.load.add.f64.p"))) {
       Value *Ptr = CI->getArgOperand(0);
       Value *Val = CI->getArgOperand(1);
-      Rep = Builder.CreateAtomicRMW(AtomicRMWInst::FAdd, Ptr, Val,
+      Rep = Builder.CreateAtomicRMW(AtomicRMWInst::FAdd, Ptr, Val, MaybeAlign(),
                                     AtomicOrdering::SequentiallyConsistent);
     } else if (IsNVVM && (Name == "max.i" || Name == "max.ll" ||
                           Name == "max.ui" || Name == "max.ull")) {
diff --git a/llvm/lib/IR/Core.cpp b/llvm/lib/IR/Core.cpp
--- a/llvm/lib/IR/Core.cpp
+++ b/llvm/lib/IR/Core.cpp
@@ -3977,9 +3977,10 @@
                                LLVMAtomicOrdering ordering,
                                LLVMBool singleThread) {
   AtomicRMWInst::BinOp intop = mapFromLLVMRMWBinOp(op);
-  return wrap(unwrap(B)->CreateAtomicRMW(intop, unwrap(PTR), unwrap(Val),
-    mapFromLLVMOrdering(ordering), singleThread ? SyncScope::SingleThread
-                                                : SyncScope::System));
+  return wrap(unwrap(B)->CreateAtomicRMW(
+      intop, unwrap(PTR), unwrap(Val), MaybeAlign(),
+      mapFromLLVMOrdering(ordering),
+      singleThread ? SyncScope::SingleThread : SyncScope::System));
 }
 
 LLVMValueRef LLVMBuildAtomicCmpXchg(LLVMBuilderRef B, LLVMValueRef Ptr,
@@ -3988,10 +3989,11 @@
                                     LLVMAtomicOrdering FailureOrdering,
                                     LLVMBool singleThread) {
 
-  return wrap(unwrap(B)->CreateAtomicCmpXchg(unwrap(Ptr), unwrap(Cmp),
-                unwrap(New), mapFromLLVMOrdering(SuccessOrdering),
-                mapFromLLVMOrdering(FailureOrdering),
-                singleThread ? SyncScope::SingleThread : SyncScope::System));
+  return wrap(unwrap(B)->CreateAtomicCmpXchg(
+      unwrap(Ptr), unwrap(Cmp), unwrap(New), MaybeAlign(),
+      mapFromLLVMOrdering(SuccessOrdering),
+      mapFromLLVMOrdering(FailureOrdering),
+      singleThread ? SyncScope::SingleThread : SyncScope::System));
 }
 
 unsigned LLVMGetNumMaskElements(LLVMValueRef SVInst) {
diff --git a/llvm/lib/Transforms/Instrumentation/GCOVProfiling.cpp b/llvm/lib/Transforms/Instrumentation/GCOVProfiling.cpp
--- a/llvm/lib/Transforms/Instrumentation/GCOVProfiling.cpp
+++ b/llvm/lib/Transforms/Instrumentation/GCOVProfiling.cpp
@@ -965,7 +965,7 @@
               Counters->getValueType(), Counters, 0, I);
           if (Options.Atomic) {
             Builder.CreateAtomicRMW(AtomicRMWInst::Add, V, Builder.getInt64(1),
-                                    AtomicOrdering::Monotonic);
+                                    MaybeAlign(), AtomicOrdering::Monotonic);
           } else {
             Value *Count =
                 Builder.CreateLoad(Builder.getInt64Ty(), V, "gcov_ctr");
diff --git a/llvm/lib/Transforms/Instrumentation/InstrOrderFile.cpp b/llvm/lib/Transforms/Instrumentation/InstrOrderFile.cpp
--- a/llvm/lib/Transforms/Instrumentation/InstrOrderFile.cpp
+++ b/llvm/lib/Transforms/Instrumentation/InstrOrderFile.cpp
@@ -144,7 +144,7 @@
     // Fill up UpdateOrderFileBB: grab the index, update the buffer!
     Value *IdxVal = updateB.CreateAtomicRMW(
         AtomicRMWInst::Add, BufferIdx, ConstantInt::get(Int32Ty, 1),
-        AtomicOrdering::SequentiallyConsistent);
+        MaybeAlign(), AtomicOrdering::SequentiallyConsistent);
     // We need to wrap around the index to fit it inside the buffer.
     Value *WrappedIdx = updateB.CreateAnd(
         IdxVal, ConstantInt::get(Int32Ty, INSTR_ORDER_FILE_BUFFER_MASK));
diff --git a/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp b/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp
--- a/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp
+++ b/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp
@@ -205,6 +205,7 @@
         // automic update currently can only be promoted across the current
         // loop, not the whole loop nest.
         Builder.CreateAtomicRMW(AtomicRMWInst::Add, Addr, LiveInValue,
+                                MaybeAlign(),
                                 AtomicOrdering::SequentiallyConsistent);
       else {
         LoadInst *OldVal = Builder.CreateLoad(Ty, Addr, "pgocount.promoted");
@@ -702,7 +703,7 @@
   if (Options.Atomic || AtomicCounterUpdateAll ||
       (Index == 0 && AtomicFirstCounter)) {
     Builder.CreateAtomicRMW(AtomicRMWInst::Add, Addr, Inc->getStep(),
-                            AtomicOrdering::Monotonic);
+                            MaybeAlign(), AtomicOrdering::Monotonic);
   } else {
     Value *IncStep = Inc->getStep();
     Value *Load = Builder.CreateLoad(IncStep->getType(), Addr, "pgocount");
diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
@@ -1526,6 +1526,7 @@
 
 def LLVM_AtomicRMWType : AnyTypeOf<[LLVM_AnyFloat, LLVM_AnyInteger]>;
 
+// FIXME: Need to add alignment attribute to MLIR atomicrmw operation.
 def LLVM_AtomicRMWOp : LLVM_Op<"atomicrmw"> {
   let arguments = (ins AtomicBinOp:$bin_op,
                    LLVM_PointerTo<LLVM_AtomicRMWType>:$ptr,
@@ -1533,6 +1534,7 @@
   let results = (outs LLVM_AtomicRMWType:$res);
   let llvmBuilder = [{
     $res = builder.CreateAtomicRMW(getLLVMAtomicBinOp($bin_op), $ptr, $val,
+                                   llvm::MaybeAlign(),
                                    getLLVMAtomicOrdering($ordering));
   }];
   let parser = [{ return parseAtomicRMWOp(parser, result); }];
@@ -1553,6 +1555,7 @@
  "an LLVM struct type with any integer or pointer followed by a single-bit "
  "integer">;
 
+// FIXME: Need to add alignment attribute to MLIR cmpxchg operation.
 def LLVM_AtomicCmpXchgOp : LLVM_Op<"cmpxchg"> {
   let arguments = (ins LLVM_PointerTo<LLVM_AtomicCmpXchgType>:$ptr,
                    LLVM_AtomicCmpXchgType:$cmp, LLVM_AtomicCmpXchgType:$val,
@@ -1560,7 +1563,7 @@
                    AtomicOrdering:$failure_ordering);
   let results = (outs LLVM_AtomicCmpXchgResultType:$res);
   let llvmBuilder = [{
-    $res = builder.CreateAtomicCmpXchg($ptr, $cmp, $val,
+    $res = builder.CreateAtomicCmpXchg($ptr, $cmp, $val, llvm::MaybeAlign(),
                    getLLVMAtomicOrdering($success_ordering),
                    getLLVMAtomicOrdering($failure_ordering));
   }];