Update LLVM for 3.5 rebase (r209712).

Change-Id: I149556c940fb7dc92d075273c87ff584f400941f

67 files changed, 3197 insertions, 1944 deletions

diff --git a/lib/Target/ARM/A15SDOptimizer.cpp b/lib/Target/ARM/A15SDOptimizer.cpp
index 28ea879..94faf6f 100644
--- a/lib/Target/ARM/A15SDOptimizer.cpp
+++ b/lib/Target/ARM/A15SDOptimizer.cpp
@@ -24,7 +24,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "a15-sd-optimizer"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMBaseRegisterInfo.h"
@@ -39,6 +38,8 @@
 
 using namespace llvm;
 
+#define DEBUG_TYPE "a15-sd-optimizer"
+
 namespace {
   struct A15SDOptimizer : public MachineFunctionPass {
     static char ID;
@@ -90,7 +91,7 @@ namespace {
     unsigned createImplicitDef(MachineBasicBlock &MBB,
                                MachineBasicBlock::iterator InsertBefore,
                                DebugLoc DL);
-    
+
     //
     // Various property checkers
     //
@@ -259,7 +260,7 @@ unsigned A15SDOptimizer::optimizeSDPattern(MachineInstr *MI) {
       if (DPRMI && SPRMI) {
         // See if the first operand of this insert_subreg is IMPLICIT_DEF
         MachineInstr *ECDef = elideCopies(DPRMI);
-        if (ECDef != 0 && ECDef->isImplicitDef()) {
+        if (ECDef && ECDef->isImplicitDef()) {
           // Another corner case - if we're inserting something that is purely
           // a subreg copy of a DPR, just use that DPR.
 
@@ -348,10 +349,10 @@ MachineInstr *A15SDOptimizer::elideCopies(MachineInstr *MI) {
   if (!MI->isFullCopy())
     return MI;
   if (!TRI->isVirtualRegister(MI->getOperand(1).getReg()))
-    return NULL;
+    return nullptr;
   MachineInstr *Def = MRI->getVRegDef(MI->getOperand(1).getReg());
   if (!Def)
-    return NULL;
+    return nullptr;
   return elideCopies(Def);
 }
 
@@ -435,7 +436,7 @@ A15SDOptimizer::createDupLane(MachineBasicBlock &MBB,
                          Out)
                    .addReg(Reg)
                    .addImm(Lane));
- 
+
   return Out;
 }
 
@@ -601,7 +602,7 @@ bool A15SDOptimizer::runOnInstruction(MachineInstr *MI) {
   //   * INSERT_SUBREG: * If the SPR value was originally in another DPR/QPR
   //                      lane, and the other lane(s) of the DPR/QPR register
   //                      that we are inserting in are undefined, use the
-  //                      original DPR/QPR value. 
+  //                      original DPR/QPR value.
   //                    * Otherwise, fall back on the same stategy as COPY.
   //
   //   * REG_SEQUENCE:  * If all except one of the input operands are
@@ -693,7 +694,7 @@ bool A15SDOptimizer::runOnMachineFunction(MachineFunction &Fn) {
       MI != ME;) {
       Modified |= runOnInstruction(MI++);
     }
- 
+
   }
 
   for (std::set<MachineInstr *>::iterator I = DeadInstr.begin(),
diff --git a/lib/Target/ARM/ARM.h b/lib/Target/ARM/ARM.h
index 4412b45..55df29c 100644
--- a/lib/Target/ARM/ARM.h
+++ b/lib/Target/ARM/ARM.h
@@ -49,8 +49,6 @@ FunctionPass *createThumb2SizeReductionPass();
 /// \brief Creates an ARM-specific Target Transformation Info pass.
 ImmutablePass *createARMTargetTransformInfoPass(const ARMBaseTargetMachine *TM);
 
-FunctionPass *createARMAtomicExpandPass(const TargetMachine *TM);
-
 void LowerARMMachineInstrToMCInst(const MachineInstr *MI, MCInst &OutMI,
                                   ARMAsmPrinter &AP);
 
diff --git a/lib/Target/ARM/ARMAsmPrinter.cpp b/lib/Target/ARM/ARMAsmPrinter.cpp
index 0fa865f..55e9fe5 100644
--- a/lib/Target/ARM/ARMAsmPrinter.cpp
+++ b/lib/Target/ARM/ARMAsmPrinter.cpp
@@ -12,7 +12,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "asm-printer"
 #include "ARMAsmPrinter.h"
 #include "ARM.h"
 #include "ARMConstantPoolValue.h"
@@ -45,6 +44,7 @@
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/Support/ARMBuildAttributes.h"
+#include "llvm/Support/COFF.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ELF.h"
@@ -55,6 +55,8 @@
 #include <cctype>
 using namespace llvm;
 
+#define DEBUG_TYPE "asm-printer"
+
 void ARMAsmPrinter::EmitFunctionBodyEnd() {
   // Make sure to terminate any constant pools that were at the end
   // of the function.
@@ -85,7 +87,7 @@ void ARMAsmPrinter::EmitXXStructor(const Constant *CV) {
                                              ? MCSymbolRefExpr::VK_ARM_TARGET1
                                              : MCSymbolRefExpr::VK_None),
                                             OutContext);
-  
+
   OutStreamer.EmitValue(E, Size);
 }
 
@@ -96,7 +98,28 @@ bool ARMAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
   AFI = MF.getInfo<ARMFunctionInfo>();
   MCP = MF.getConstantPool();
 
-  return AsmPrinter::runOnMachineFunction(MF);
+  SetupMachineFunction(MF);
+
+  if (Subtarget->isTargetCOFF()) {
+    bool Internal = MF.getFunction()->hasInternalLinkage();
+    COFF::SymbolStorageClass Scl = Internal ? COFF::IMAGE_SYM_CLASS_STATIC
+                                            : COFF::IMAGE_SYM_CLASS_EXTERNAL;
+    int Type = COFF::IMAGE_SYM_DTYPE_FUNCTION << COFF::SCT_COMPLEX_TYPE_SHIFT;
+
+    OutStreamer.BeginCOFFSymbolDef(CurrentFnSym);
+    OutStreamer.EmitCOFFSymbolStorageClass(Scl);
+    OutStreamer.EmitCOFFSymbolType(Type);
+    OutStreamer.EndCOFFSymbolDef();
+  }
+
+  // Have common code print out the function header with linkage info etc.
+  EmitFunctionHeader();
+
+  // Emit the rest of the function body.
+  EmitFunctionBody();
+
+  // We didn't modify anything.
+  return false;
 }
 
 void ARMAsmPrinter::printOperand(const MachineInstr *MI, int OpNum,
@@ -239,7 +262,7 @@ bool ARMAsmPrinter::PrintAsmOperand(const MachineInstr *MI, unsigned OpNum,
       if (ARM::GPRPairRegClass.contains(RegBegin)) {
         const TargetRegisterInfo *TRI = MF->getTarget().getRegisterInfo();
         unsigned Reg0 = TRI->getSubReg(RegBegin, ARM::gsub_0);
-        O << ARMInstPrinter::getRegisterName(Reg0) << ", ";;
+        O << ARMInstPrinter::getRegisterName(Reg0) << ", ";
         RegBegin = TRI->getSubReg(RegBegin, ARM::gsub_1);
       }
       O << ARMInstPrinter::getRegisterName(RegBegin);
@@ -383,7 +406,7 @@ void ARMAsmPrinter::emitInlineAsmEnd(const MCSubtargetInfo &StartInfo,
   // If either end mode is unknown (EndInfo == NULL) or different than
   // the start mode, then restore the start mode.
   const bool WasThumb = isThumb(StartInfo);
-  if (EndInfo == NULL || WasThumb != isThumb(*EndInfo)) {
+  if (!EndInfo || WasThumb != isThumb(*EndInfo)) {
     OutStreamer.EmitAssemblerFlag(WasThumb ? MCAF_Code16 : MCAF_Code32);
   }
 }
@@ -456,6 +479,29 @@ void ARMAsmPrinter::EmitStartOfAsmFile(Module &M) {
     emitAttributes();
 }
 
+static void
+emitNonLazySymbolPointer(MCStreamer &OutStreamer, MCSymbol *StubLabel,
+                         MachineModuleInfoImpl::StubValueTy &MCSym) {
+  // L_foo$stub:
+  OutStreamer.EmitLabel(StubLabel);
+  //   .indirect_symbol _foo
+  OutStreamer.EmitSymbolAttribute(MCSym.getPointer(), MCSA_IndirectSymbol);
+
+  if (MCSym.getInt())
+    // External to current translation unit.
+    OutStreamer.EmitIntValue(0, 4/*size*/);
+  else
+    // Internal to current translation unit.
+    //
+    // When we place the LSDA into the TEXT section, the type info
+    // pointers need to be indirect and pc-rel. We accomplish this by
+    // using NLPs; however, sometimes the types are local to the file.
+    // We need to fill in the value for the NLP in those cases.
+    OutStreamer.EmitValue(
+        MCSymbolRefExpr::Create(MCSym.getPointer(), OutStreamer.getContext()),
+        4 /*size*/);
+}
+
 
 void ARMAsmPrinter::EmitEndOfAsmFile(Module &M) {
   if (Subtarget->isTargetMachO()) {
@@ -472,27 +518,9 @@ void ARMAsmPrinter::EmitEndOfAsmFile(Module &M) {
       // Switch with ".non_lazy_symbol_pointer" directive.
       OutStreamer.SwitchSection(TLOFMacho.getNonLazySymbolPointerSection());
       EmitAlignment(2);
-      for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
-        // L_foo$stub:
-        OutStreamer.EmitLabel(Stubs[i].first);
-        //   .indirect_symbol _foo
-        MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
-        OutStreamer.EmitSymbolAttribute(MCSym.getPointer(),MCSA_IndirectSymbol);
-
-        if (MCSym.getInt())
-          // External to current translation unit.
-          OutStreamer.EmitIntValue(0, 4/*size*/);
-        else
-          // Internal to current translation unit.
-          //
-          // When we place the LSDA into the TEXT section, the type info
-          // pointers need to be indirect and pc-rel. We accomplish this by
-          // using NLPs; however, sometimes the types are local to the file.
-          // We need to fill in the value for the NLP in those cases.
-          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym.getPointer(),
-                                                        OutContext),
-                                4/*size*/);
-      }
+
+      for (auto &Stub : Stubs)
+        emitNonLazySymbolPointer(OutStreamer, Stub.first, Stub.second);
 
       Stubs.clear();
       OutStreamer.AddBlankLine();
@@ -500,17 +528,11 @@ void ARMAsmPrinter::EmitEndOfAsmFile(Module &M) {
 
     Stubs = MMIMacho.GetHiddenGVStubList();
     if (!Stubs.empty()) {
-      OutStreamer.SwitchSection(getObjFileLowering().getDataSection());
+      OutStreamer.SwitchSection(TLOFMacho.getNonLazySymbolPointerSection());
       EmitAlignment(2);
-      for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
-        // L_foo$stub:
-        OutStreamer.EmitLabel(Stubs[i].first);
-        //   .long _foo
-        OutStreamer.EmitValue(MCSymbolRefExpr::
-                              Create(Stubs[i].second.getPointer(),
-                                     OutContext),
-                              4/*size*/);
-      }
+
+      for (auto &Stub : Stubs)
+        emitNonLazySymbolPointer(OutStreamer, Stub.first, Stub.second);
 
       Stubs.clear();
       OutStreamer.AddBlankLine();
@@ -523,6 +545,28 @@ void ARMAsmPrinter::EmitEndOfAsmFile(Module &M) {
     // generates code that does this, it is always safe to set.
     OutStreamer.EmitAssemblerFlag(MCAF_SubsectionsViaSymbols);
   }
+
+  // Emit a .data.rel section containing any stubs that were created.
+  if (Subtarget->isTargetELF()) {
+    const TargetLoweringObjectFileELF &TLOFELF =
+      static_cast<const TargetLoweringObjectFileELF &>(getObjFileLowering());
+
+    MachineModuleInfoELF &MMIELF = MMI->getObjFileInfo<MachineModuleInfoELF>();
+
+    // Output stubs for external and common global variables.
+    MachineModuleInfoELF::SymbolListTy Stubs = MMIELF.GetGVStubList();
+    if (!Stubs.empty()) {
+      OutStreamer.SwitchSection(TLOFELF.getDataRelSection());
+      const DataLayout *TD = TM.getDataLayout();
+
+      for (auto &stub: Stubs) {
+        OutStreamer.EmitLabel(stub.first);
+        OutStreamer.EmitSymbolValue(stub.second.getPointer(),
+                                    TD->getPointerSize(0));
+      }
+      Stubs.clear();
+    }
+  }
 }
 
 //===----------------------------------------------------------------------===//
@@ -575,7 +619,7 @@ void ARMAsmPrinter::emitAttributes() {
                     getArchForCPU(CPUString, Subtarget));
 
   // Tag_CPU_arch_profile must have the default value of 0 when "Architecture
-  // profile is not applicable (e.g. pre v7, or cross-profile code)". 
+  // profile is not applicable (e.g. pre v7, or cross-profile code)".
   if (Subtarget->hasV7Ops()) {
     if (Subtarget->isAClass()) {
       ATS.emitAttribute(ARMBuildAttrs::CPU_arch_profile,
@@ -627,6 +671,20 @@ void ARMAsmPrinter::emitAttributes() {
       ATS.emitFPU(ARM::VFPV2);
   }
 
+  if (TM.getRelocationModel() == Reloc::PIC_) {
+    // PIC specific attributes.
+    ATS.emitAttribute(ARMBuildAttrs::ABI_PCS_RW_data,
+                      ARMBuildAttrs::AddressRWPCRel);
+    ATS.emitAttribute(ARMBuildAttrs::ABI_PCS_RO_data,
+                      ARMBuildAttrs::AddressROPCRel);
+    ATS.emitAttribute(ARMBuildAttrs::ABI_PCS_GOT_use,
+                      ARMBuildAttrs::AddressGOT);
+  } else {
+    // Allow direct addressing of imported data for all other relocation models.
+    ATS.emitAttribute(ARMBuildAttrs::ABI_PCS_GOT_use,
+                      ARMBuildAttrs::AddressDirect);
+  }
+
   // Signal various FP modes.
   if (!TM.Options.UnsafeFPMath) {
     ATS.emitAttribute(ARMBuildAttrs::ABI_FP_denormal, ARMBuildAttrs::Allowed);
@@ -723,7 +781,7 @@ MCSymbol *ARMAsmPrinter::GetARMGVSymbol(const GlobalValue *GV,
   MachineModuleInfoImpl::StubValueTy &StubSym =
     GV->hasHiddenVisibility() ? MMIMachO.getHiddenGVStubEntry(MCSym) :
     MMIMachO.getGVStubEntry(MCSym);
-  if (StubSym.getPointer() == 0)
+  if (!StubSym.getPointer())
     StubSym = MachineModuleInfoImpl::
       StubValueTy(getSymbol(GV), !GV->hasInternalLinkage());
   return MCSym;
@@ -971,7 +1029,8 @@ void ARMAsmPrinter::EmitUnwindingInstruction(const MachineInstr *MI) {
       RegList.push_back(SrcReg);
       break;
     }
-    ATS.emitRegSave(RegList, Opc == ARM::VSTMDDB_UPD);
+    if (MAI->getExceptionHandlingType() == ExceptionHandling::ARM)
+      ATS.emitRegSave(RegList, Opc == ARM::VSTMDDB_UPD);
   } else {
     // Changes of stack / frame pointer.
     if (SrcReg == ARM::SP) {
@@ -1016,18 +1075,20 @@ void ARMAsmPrinter::EmitUnwindingInstruction(const MachineInstr *MI) {
       }
       }
 
-      if (DstReg == FramePtr && FramePtr != ARM::SP)
-        // Set-up of the frame pointer. Positive values correspond to "add"
-        // instruction.
-        ATS.emitSetFP(FramePtr, ARM::SP, -Offset);
-      else if (DstReg == ARM::SP) {
-        // Change of SP by an offset. Positive values correspond to "sub"
-        // instruction.
-        ATS.emitPad(Offset);
-      } else {
-        // Move of SP to a register.  Positive values correspond to an "add"
-        // instruction.
-        ATS.emitMovSP(DstReg, -Offset);
+      if (MAI->getExceptionHandlingType() == ExceptionHandling::ARM) {
+        if (DstReg == FramePtr && FramePtr != ARM::SP)
+          // Set-up of the frame pointer. Positive values correspond to "add"
+          // instruction.
+          ATS.emitSetFP(FramePtr, ARM::SP, -Offset);
+        else if (DstReg == ARM::SP) {
+          // Change of SP by an offset. Positive values correspond to "sub"
+          // instruction.
+          ATS.emitPad(Offset);
+        } else {
+          // Move of SP to a register.  Positive values correspond to an "add"
+          // instruction.
+          ATS.emitMovSP(DstReg, -Offset);
+        }
       }
     } else if (DstReg == ARM::SP) {
       MI->dump();
diff --git a/lib/Target/ARM/ARMAsmPrinter.h b/lib/Target/ARM/ARMAsmPrinter.h
index 46c2626..7c103c6 100644
--- a/lib/Target/ARM/ARMAsmPrinter.h
+++ b/lib/Target/ARM/ARMAsmPrinter.h
@@ -47,16 +47,17 @@ class LLVM_LIBRARY_VISIBILITY ARMAsmPrinter : public AsmPrinter {
   bool InConstantPool;
 public:
   explicit ARMAsmPrinter(TargetMachine &TM, MCStreamer &Streamer)
-    : AsmPrinter(TM, Streamer), AFI(NULL), MCP(NULL), InConstantPool(false) {
-      Subtarget = &TM.getSubtarget<ARMSubtarget>();
-    }
+    : AsmPrinter(TM, Streamer), AFI(nullptr), MCP(nullptr),
+      InConstantPool(false) {
+    Subtarget = &TM.getSubtarget<ARMSubtarget>();
+  }
 
   const char *getPassName() const override {
     return "ARM Assembly / Object Emitter";
   }
 
   void printOperand(const MachineInstr *MI, int OpNum, raw_ostream &O,
-                    const char *Modifier = 0);
+                    const char *Modifier = nullptr);
 
   bool PrintAsmOperand(const MachineInstr *MI, unsigned OpNum,
                        unsigned AsmVariant, const char *ExtraCode,
diff --git a/lib/Target/ARM/ARMAtomicExpandPass.cpp b/lib/Target/ARM/ARMAtomicExpandPass.cpp
deleted file mode 100644
index 18e0783..0000000
--- a/lib/Target/ARM/ARMAtomicExpandPass.cpp
+++ /dev/null
@@ -1,406 +0,0 @@
-//===-- ARMAtomicExpandPass.cpp - Expand atomic instructions --------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file contains a pass (at IR level) to replace atomic instructions with
-// appropriate (intrinsic-based) ldrex/strex loops.
-//
-//===----------------------------------------------------------------------===//
-
-#define DEBUG_TYPE "arm-atomic-expand"
-#include "ARM.h"
-#include "llvm/CodeGen/Passes.h"
-#include "llvm/IR/Function.h"
-#include "llvm/IR/IRBuilder.h"
-#include "llvm/IR/Instructions.h"
-#include "llvm/IR/Intrinsics.h"
-#include "llvm/IR/Module.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Target/TargetLowering.h"
-#include "llvm/Target/TargetMachine.h"
-using namespace llvm;
-
-namespace {
-  class ARMAtomicExpandPass : public FunctionPass {
-    const TargetLowering *TLI;
-  public:
-    static char ID; // Pass identification, replacement for typeid
-    explicit ARMAtomicExpandPass(const TargetMachine *TM = 0)
-      : FunctionPass(ID), TLI(TM->getTargetLowering()) {}
-
-    bool runOnFunction(Function &F) override;
-    bool expandAtomicInsts(Function &F);
-
-    bool expandAtomicLoad(LoadInst *LI);
-    bool expandAtomicStore(StoreInst *LI);
-    bool expandAtomicRMW(AtomicRMWInst *AI);
-    bool expandAtomicCmpXchg(AtomicCmpXchgInst *CI);
-
-    AtomicOrdering insertLeadingFence(IRBuilder<> &Builder, AtomicOrdering Ord);
-    void insertTrailingFence(IRBuilder<> &Builder, AtomicOrdering Ord);
-
-    /// Perform a load-linked operation on Addr, returning a "Value *" with the
-    /// corresponding pointee type. This may entail some non-trivial operations
-    /// to truncate or reconstruct illegal types since intrinsics must be legal
-    Value *loadLinked(IRBuilder<> &Builder, Value *Addr, AtomicOrdering Ord);
-
-    /// Perform a store-conditional operation to Addr. Return the status of the
-    /// store: 0 if the it succeeded, non-zero otherwise.
-    Value *storeConditional(IRBuilder<> &Builder, Value *Val, Value *Addr,
-                            AtomicOrdering Ord);
-
-    /// Return true if the given (atomic) instruction should be expanded by this
-    /// pass.
-    bool shouldExpandAtomic(Instruction *Inst);
-  };
-}
-
-char ARMAtomicExpandPass::ID = 0;
-
-FunctionPass *llvm::createARMAtomicExpandPass(const TargetMachine *TM) {
-  return new ARMAtomicExpandPass(TM);
-}
-
-bool ARMAtomicExpandPass::runOnFunction(Function &F) {
-  SmallVector<Instruction *, 1> AtomicInsts;
-
-  // Changing control-flow while iterating through it is a bad idea, so gather a
-  // list of all atomic instructions before we start.
-  for (BasicBlock &BB : F)
-    for (Instruction &Inst : BB) {
-      if (isa<AtomicRMWInst>(&Inst) || isa<AtomicCmpXchgInst>(&Inst) ||
-          (isa<LoadInst>(&Inst) && cast<LoadInst>(&Inst)->isAtomic()) ||
-          (isa<StoreInst>(&Inst) && cast<StoreInst>(&Inst)->isAtomic()))
-        AtomicInsts.push_back(&Inst);
-    }
-
-  bool MadeChange = false;
-  for (Instruction *Inst : AtomicInsts) {
-    if (!shouldExpandAtomic(Inst))
-      continue;
-
-    if (AtomicRMWInst *AI = dyn_cast<AtomicRMWInst>(Inst))
-      MadeChange |= expandAtomicRMW(AI);
-    else if (AtomicCmpXchgInst *CI = dyn_cast<AtomicCmpXchgInst>(Inst))
-      MadeChange |= expandAtomicCmpXchg(CI);
-    else if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
-      MadeChange |= expandAtomicLoad(LI);
-    else if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
-      MadeChange |= expandAtomicStore(SI);
-    else
-      llvm_unreachable("Unknown atomic instruction");
-  }
-
-  return MadeChange;
-}
-
-bool ARMAtomicExpandPass::expandAtomicLoad(LoadInst *LI) {
-  // Load instructions don't actually need a leading fence, even in the
-  // SequentiallyConsistent case.
-  AtomicOrdering MemOpOrder =
-    TLI->getInsertFencesForAtomic() ? Monotonic : LI->getOrdering();
-
-  // The only 64-bit load guaranteed to be single-copy atomic by the ARM ARM is
-  // an ldrexd (A3.5.3).
-  IRBuilder<> Builder(LI);
-  Value *Val = loadLinked(Builder, LI->getPointerOperand(), MemOpOrder);
-
-  insertTrailingFence(Builder, LI->getOrdering());
-
-  LI->replaceAllUsesWith(Val);
-  LI->eraseFromParent();
-
-  return true;
-}
-
-bool ARMAtomicExpandPass::expandAtomicStore(StoreInst *SI) {
-  // The only atomic 64-bit store on ARM is an strexd that succeeds, which means
-  // we need a loop and the entire instruction is essentially an "atomicrmw
-  // xchg" that ignores the value loaded.
-  IRBuilder<> Builder(SI);
-  AtomicRMWInst *AI =
-      Builder.CreateAtomicRMW(AtomicRMWInst::Xchg, SI->getPointerOperand(),
-                              SI->getValueOperand(), SI->getOrdering());
-  SI->eraseFromParent();
-
-  // Now we have an appropriate swap instruction, lower it as usual.
-  return expandAtomicRMW(AI);
-}
-
-bool ARMAtomicExpandPass::expandAtomicRMW(AtomicRMWInst *AI) {
-  AtomicOrdering Order = AI->getOrdering();
-  Value *Addr = AI->getPointerOperand();
-  BasicBlock *BB = AI->getParent();
-  Function *F = BB->getParent();
-  LLVMContext &Ctx = F->getContext();
-
-  // Given: atomicrmw some_op iN* %addr, iN %incr ordering
-  //
-  // The standard expansion we produce is:
-  //     [...]
-  //     fence?
-  // atomicrmw.start:
-  //     %loaded = @load.linked(%addr)
-  //     %new = some_op iN %loaded, %incr
-  //     %stored = @store_conditional(%new, %addr)
-  //     %try_again = icmp i32 ne %stored, 0
-  //     br i1 %try_again, label %loop, label %atomicrmw.end
-  // atomicrmw.end:
-  //     fence?
-  //     [...]
-  BasicBlock *ExitBB = BB->splitBasicBlock(AI, "atomicrmw.end");
-  BasicBlock *LoopBB =  BasicBlock::Create(Ctx, "atomicrmw.start", F, ExitBB);
-
-  // This grabs the DebugLoc from AI.
-  IRBuilder<> Builder(AI);
-
-  // The split call above "helpfully" added a branch at the end of BB (to the
-  // wrong place), but we might want a fence too. It's easiest to just remove
-  // the branch entirely.
-  std::prev(BB->end())->eraseFromParent();
-  Builder.SetInsertPoint(BB);
-  AtomicOrdering MemOpOrder = insertLeadingFence(Builder, Order);
-  Builder.CreateBr(LoopBB);
-
-  // Start the main loop block now that we've taken care of the preliminaries.
-  Builder.SetInsertPoint(LoopBB);
-  Value *Loaded = loadLinked(Builder, Addr, MemOpOrder);
-
-  Value *NewVal;
-  switch (AI->getOperation()) {
-  case AtomicRMWInst::Xchg:
-    NewVal = AI->getValOperand();
-    break;
-  case AtomicRMWInst::Add:
-    NewVal = Builder.CreateAdd(Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::Sub:
-    NewVal = Builder.CreateSub(Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::And:
-    NewVal = Builder.CreateAnd(Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::Nand:
-    NewVal = Builder.CreateAnd(Loaded, Builder.CreateNot(AI->getValOperand()),
-                               "new");
-    break;
-  case AtomicRMWInst::Or:
-    NewVal = Builder.CreateOr(Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::Xor:
-    NewVal = Builder.CreateXor(Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::Max:
-    NewVal = Builder.CreateICmpSGT(Loaded, AI->getValOperand());
-    NewVal = Builder.CreateSelect(NewVal, Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::Min:
-    NewVal = Builder.CreateICmpSLE(Loaded, AI->getValOperand());
-    NewVal = Builder.CreateSelect(NewVal, Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::UMax:
-    NewVal = Builder.CreateICmpUGT(Loaded, AI->getValOperand());
-    NewVal = Builder.CreateSelect(NewVal, Loaded, AI->getValOperand(), "new");
-    break;
-  case AtomicRMWInst::UMin:
-    NewVal = Builder.CreateICmpULE(Loaded, AI->getValOperand());
-    NewVal = Builder.CreateSelect(NewVal, Loaded, AI->getValOperand(), "new");
-    break;
-  default:
-    llvm_unreachable("Unknown atomic op");
-  }
-
-  Value *StoreSuccess = storeConditional(Builder, NewVal, Addr, MemOpOrder);
-  Value *TryAgain = Builder.CreateICmpNE(
-      StoreSuccess, ConstantInt::get(IntegerType::get(Ctx, 32), 0), "tryagain");
-  Builder.CreateCondBr(TryAgain, LoopBB, ExitBB);
-
-  Builder.SetInsertPoint(ExitBB, ExitBB->begin());
-  insertTrailingFence(Builder, Order);
-
-  AI->replaceAllUsesWith(Loaded);
-  AI->eraseFromParent();
-
-  return true;
-}
-
-bool ARMAtomicExpandPass::expandAtomicCmpXchg(AtomicCmpXchgInst *CI) {
-  AtomicOrdering SuccessOrder = CI->getSuccessOrdering();
-  AtomicOrdering FailureOrder = CI->getFailureOrdering();
-  Value *Addr = CI->getPointerOperand();
-  BasicBlock *BB = CI->getParent();
-  Function *F = BB->getParent();
-  LLVMContext &Ctx = F->getContext();
-
-  // Given: cmpxchg some_op iN* %addr, iN %desired, iN %new success_ord fail_ord
-  //
-  // The full expansion we produce is:
-  //     [...]
-  //     fence?
-  // cmpxchg.start:
-  //     %loaded = @load.linked(%addr)
-  //     %should_store = icmp eq %loaded, %desired
-  //     br i1 %should_store, label %cmpxchg.trystore,
-  //                          label %cmpxchg.end/%cmpxchg.barrier
-  // cmpxchg.trystore:
-  //     %stored = @store_conditional(%new, %addr)
-  //     %try_again = icmp i32 ne %stored, 0
-  //     br i1 %try_again, label %loop, label %cmpxchg.end
-  // cmpxchg.barrier:
-  //     fence?
-  //     br label %cmpxchg.end
-  // cmpxchg.end:
-  //     [...]
-  BasicBlock *ExitBB = BB->splitBasicBlock(CI, "cmpxchg.end");
-  auto BarrierBB = BasicBlock::Create(Ctx, "cmpxchg.trystore", F, ExitBB);
-  auto TryStoreBB = BasicBlock::Create(Ctx, "cmpxchg.barrier", F, BarrierBB);
-  auto LoopBB = BasicBlock::Create(Ctx, "cmpxchg.start", F, TryStoreBB);
-
-  // This grabs the DebugLoc from CI
-  IRBuilder<> Builder(CI);
-
-  // The split call above "helpfully" added a branch at the end of BB (to the
-  // wrong place), but we might want a fence too. It's easiest to just remove
-  // the branch entirely.
-  std::prev(BB->end())->eraseFromParent();
-  Builder.SetInsertPoint(BB);
-  AtomicOrdering MemOpOrder = insertLeadingFence(Builder, SuccessOrder);
-  Builder.CreateBr(LoopBB);
-
-  // Start the main loop block now that we've taken care of the preliminaries.
-  Builder.SetInsertPoint(LoopBB);
-  Value *Loaded = loadLinked(Builder, Addr, MemOpOrder);
-  Value *ShouldStore =
-      Builder.CreateICmpEQ(Loaded, CI->getCompareOperand(), "should_store");
-
-  // If the the cmpxchg doesn't actually need any ordering when it fails, we can
-  // jump straight past that fence instruction (if it exists).
-  BasicBlock *FailureBB = FailureOrder == Monotonic ? ExitBB : BarrierBB;
-  Builder.CreateCondBr(ShouldStore, TryStoreBB, FailureBB);
-
-  Builder.SetInsertPoint(TryStoreBB);
-  Value *StoreSuccess =
-      storeConditional(Builder, CI->getNewValOperand(), Addr, MemOpOrder);
-  Value *TryAgain = Builder.CreateICmpNE(
-      StoreSuccess, ConstantInt::get(Type::getInt32Ty(Ctx), 0), "success");
-  Builder.CreateCondBr(TryAgain, LoopBB, BarrierBB);
-
-  // Finally, make sure later instructions don't get reordered with a fence if
-  // necessary.
-  Builder.SetInsertPoint(BarrierBB);
-  insertTrailingFence(Builder, SuccessOrder);
-  Builder.CreateBr(ExitBB);
-
-  CI->replaceAllUsesWith(Loaded);
-  CI->eraseFromParent();
-
-  return true;
-}
-
-Value *ARMAtomicExpandPass::loadLinked(IRBuilder<> &Builder, Value *Addr,
-                                          AtomicOrdering Ord) {
-  Module *M = Builder.GetInsertBlock()->getParent()->getParent();
-  Type *ValTy = cast<PointerType>(Addr->getType())->getElementType();
-  bool IsAcquire =
-      Ord == Acquire || Ord == AcquireRelease || Ord == SequentiallyConsistent;
-
-  // Since i64 isn't legal and intrinsics don't get type-lowered, the ldrexd
-  // intrinsic must return {i32, i32} and we have to recombine them into a
-  // single i64 here.
-  if (ValTy->getPrimitiveSizeInBits() == 64) {
-    Intrinsic::ID Int =
-        IsAcquire ? Intrinsic::arm_ldaexd : Intrinsic::arm_ldrexd;
-    Function *Ldrex = llvm::Intrinsic::getDeclaration(M, Int);
-
-    Addr = Builder.CreateBitCast(Addr, Type::getInt8PtrTy(M->getContext()));
-    Value *LoHi = Builder.CreateCall(Ldrex, Addr, "lohi");
-
-    Value *Lo = Builder.CreateExtractValue(LoHi, 0, "lo");
-    Value *Hi = Builder.CreateExtractValue(LoHi, 1, "hi");
-    Lo = Builder.CreateZExt(Lo, ValTy, "lo64");
-    Hi = Builder.CreateZExt(Hi, ValTy, "hi64");
-    return Builder.CreateOr(
-        Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 32)), "val64");
-  }
-
-  Type *Tys[] = { Addr->getType() };
-  Intrinsic::ID Int = IsAcquire ? Intrinsic::arm_ldaex : Intrinsic::arm_ldrex;
-  Function *Ldrex = llvm::Intrinsic::getDeclaration(M, Int, Tys);
-
-  return Builder.CreateTruncOrBitCast(
-      Builder.CreateCall(Ldrex, Addr),
-      cast<PointerType>(Addr->getType())->getElementType());
-}
-
-Value *ARMAtomicExpandPass::storeConditional(IRBuilder<> &Builder, Value *Val,
-                                           Value *Addr, AtomicOrdering Ord) {
-  Module *M = Builder.GetInsertBlock()->getParent()->getParent();
-  bool IsRelease =
-      Ord == Release || Ord == AcquireRelease || Ord == SequentiallyConsistent;
-
-  // Since the intrinsics must have legal type, the i64 intrinsics take two
-  // parameters: "i32, i32". We must marshal Val into the appropriate form
-  // before the call.
-  if (Val->getType()->getPrimitiveSizeInBits() == 64) {
-    Intrinsic::ID Int =
-        IsRelease ? Intrinsic::arm_stlexd : Intrinsic::arm_strexd;
-    Function *Strex = Intrinsic::getDeclaration(M, Int);
-    Type *Int32Ty = Type::getInt32Ty(M->getContext());
-
-    Value *Lo = Builder.CreateTrunc(Val, Int32Ty, "lo");
-    Value *Hi = Builder.CreateTrunc(Builder.CreateLShr(Val, 32), Int32Ty, "hi");
-    Addr = Builder.CreateBitCast(Addr, Type::getInt8PtrTy(M->getContext()));
-    return Builder.CreateCall3(Strex, Lo, Hi, Addr);
-  }
-
-  Intrinsic::ID Int = IsRelease ? Intrinsic::arm_stlex : Intrinsic::arm_strex;
-  Type *Tys[] = { Addr->getType() };
-  Function *Strex = Intrinsic::getDeclaration(M, Int, Tys);
-
-  return Builder.CreateCall2(
-      Strex, Builder.CreateZExtOrBitCast(
-                 Val, Strex->getFunctionType()->getParamType(0)),
-      Addr);
-}
-
-AtomicOrdering ARMAtomicExpandPass::insertLeadingFence(IRBuilder<> &Builder,
-                                                       AtomicOrdering Ord) {
-  if (!TLI->getInsertFencesForAtomic())
-    return Ord;
-
-  if (Ord == Release || Ord == AcquireRelease || Ord == SequentiallyConsistent)
-    Builder.CreateFence(Release);
-
-  // The exclusive operations don't need any barrier if we're adding separate
-  // fences.
-  return Monotonic;
-}
-
-void ARMAtomicExpandPass::insertTrailingFence(IRBuilder<> &Builder,
-                                              AtomicOrdering Ord) {
-  if (!TLI->getInsertFencesForAtomic())
-    return;
-
-  if (Ord == Acquire || Ord == AcquireRelease)
-    Builder.CreateFence(Acquire);
-  else if (Ord == SequentiallyConsistent)
-    Builder.CreateFence(SequentiallyConsistent);
-}
-
-bool ARMAtomicExpandPass::shouldExpandAtomic(Instruction *Inst) {
-  // Loads and stores less than 64-bits are already atomic; ones above that
-  // are doomed anyway, so defer to the default libcall and blame the OS when
-  // things go wrong:
-  if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
-    return SI->getValueOperand()->getType()->getPrimitiveSizeInBits() == 64;
-  else if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
-    return LI->getType()->getPrimitiveSizeInBits() == 64;
-
-  // For the real atomic operations, we have ldrex/strex up to 64 bits.
-  return Inst->getType()->getPrimitiveSizeInBits() <= 64;
-}
diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 47f5bf9..bc266e8 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -37,11 +37,13 @@
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 
+using namespace llvm;
+
+#define DEBUG_TYPE "arm-instrinfo"
+
 #define GET_INSTRINFO_CTOR_DTOR
 #include "ARMGenInstrInfo.inc"
 
-using namespace llvm;
-
 static cl::opt<bool>
 EnableARM3Addr("enable-arm-3-addr-conv", cl::Hidden,
                cl::desc("Enable ARM 2-addr to 3-addr conv"));
@@ -125,14 +127,14 @@ ARMBaseInstrInfo::convertToThreeAddress(MachineFunction::iterator &MFI,
   // FIXME: Thumb2 support.
 
   if (!EnableARM3Addr)
-    return NULL;
+    return nullptr;
 
   MachineInstr *MI = MBBI;
   MachineFunction &MF = *MI->getParent()->getParent();
   uint64_t TSFlags = MI->getDesc().TSFlags;
   bool isPre = false;
   switch ((TSFlags & ARMII::IndexModeMask) >> ARMII::IndexModeShift) {
-  default: return NULL;
+  default: return nullptr;
   case ARMII::IndexModePre:
     isPre = true;
     break;
@@ -144,10 +146,10 @@ ARMBaseInstrInfo::convertToThreeAddress(MachineFunction::iterator &MFI,
   // operation.
   unsigned MemOpc = getUnindexedOpcode(MI->getOpcode());
   if (MemOpc == 0)
-    return NULL;
+    return nullptr;
 
-  MachineInstr *UpdateMI = NULL;
-  MachineInstr *MemMI = NULL;
+  MachineInstr *UpdateMI = nullptr;
+  MachineInstr *MemMI = nullptr;
   unsigned AddrMode = (TSFlags & ARMII::AddrModeMask);
   const MCInstrDesc &MCID = MI->getDesc();
   unsigned NumOps = MCID.getNumOperands();
@@ -169,7 +171,7 @@ ARMBaseInstrInfo::convertToThreeAddress(MachineFunction::iterator &MFI,
       if (ARM_AM::getSOImmVal(Amt) == -1)
         // Can't encode it in a so_imm operand. This transformation will
         // add more than 1 instruction. Abandon!
-        return NULL;
+        return nullptr;
       UpdateMI = BuildMI(MF, MI->getDebugLoc(),
                          get(isSub ? ARM::SUBri : ARM::ADDri), WBReg)
         .addReg(BaseReg).addImm(Amt)
@@ -273,8 +275,8 @@ ARMBaseInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
                                 MachineBasicBlock *&FBB,
                                 SmallVectorImpl<MachineOperand> &Cond,
                                 bool AllowModify) const {
-  TBB = 0;
-  FBB = 0;
+  TBB = nullptr;
+  FBB = nullptr;
 
   MachineBasicBlock::iterator I = MBB.end();
   if (I == MBB.begin())
@@ -331,7 +333,7 @@ ARMBaseInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
            I->isReturn())) {
       // Forget any previous condition branch information - it no longer applies.
       Cond.clear();
-      FBB = 0;
+      FBB = nullptr;
 
       // If we can modify the function, delete everything below this
       // unconditional branch.
@@ -405,7 +407,7 @@ ARMBaseInstrInfo::InsertBranch(MachineBasicBlock &MBB, MachineBasicBlock *TBB,
   assert((Cond.size() == 2 || Cond.size() == 0) &&
          "ARM branch conditions have two components!");
 
-  if (FBB == 0) {
+  if (!FBB) {
     if (Cond.empty()) { // Unconditional branch?
       if (isThumb)
         BuildMI(&MBB, DL, get(BOpc)).addMBB(TBB).addImm(ARMCC::AL).addReg(0);
@@ -535,7 +537,8 @@ bool ARMBaseInstrInfo::isPredicable(MachineInstr *MI) const {
   return true;
 }
 
-template<> bool IsCPSRDead<MachineInstr>(MachineInstr* MI) {
+namespace llvm {
+template <> bool IsCPSRDead<MachineInstr>(MachineInstr *MI) {
   for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = MI->getOperand(i);
     if (!MO.isReg() || MO.isUndef() || MO.isUse())
@@ -548,6 +551,7 @@ template<> bool IsCPSRDead<MachineInstr>(MachineInstr* MI) {
   // all definitions of CPSR are dead
   return true;
 }
+}
 
 /// FIXME: Works around a gcc miscompilation with -fstrict-aliasing.
 LLVM_ATTRIBUTE_NOINLINE
@@ -620,7 +624,7 @@ unsigned ARMBaseInstrInfo::GetInstSizeInBytes(const MachineInstr *MI) const {
       MI->getOperand(NumOps - (MI->isPredicable() ? 3 : 2));
     unsigned JTI = JTOP.getIndex();
     const MachineJumpTableInfo *MJTI = MF->getJumpTableInfo();
-    assert(MJTI != 0);
+    assert(MJTI != nullptr);
     const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();
     assert(JTI < JT.size());
     // Thumb instructions are 2 byte aligned, but JT entries are 4 byte
@@ -1248,7 +1252,7 @@ static unsigned duplicateCPV(MachineFunction &MF, unsigned &CPI) {
     static_cast<ARMConstantPoolValue*>(MCPE.Val.MachineCPVal);
 
   unsigned PCLabelId = AFI->createPICLabelUId();
-  ARMConstantPoolValue *NewCPV = 0;
+  ARMConstantPoolValue *NewCPV = nullptr;
 
   // FIXME: The below assumes PIC relocation model and that the function
   // is Thumb mode (t1 or t2). PCAdjustment would be 8 for ARM mode PIC, and
@@ -1659,10 +1663,10 @@ ARMBaseInstrInfo::commuteInstruction(MachineInstr *MI, bool NewMI) const {
     ARMCC::CondCodes CC = getInstrPredicate(MI, PredReg);
     // MOVCC AL can't be inverted. Shouldn't happen.
     if (CC == ARMCC::AL || PredReg != ARM::CPSR)
-      return NULL;
+      return nullptr;
     MI = TargetInstrInfo::commuteInstruction(MI, NewMI);
     if (!MI)
-      return NULL;
+      return nullptr;
     // After swapping the MOVCC operands, also invert the condition.
     MI->getOperand(MI->findFirstPredOperandIdx())
       .setImm(ARMCC::getOppositeCondition(CC));
@@ -1678,35 +1682,36 @@ static MachineInstr *canFoldIntoMOVCC(unsigned Reg,
                                       const MachineRegisterInfo &MRI,
                                       const TargetInstrInfo *TII) {
   if (!TargetRegisterInfo::isVirtualRegister(Reg))
-    return 0;
+    return nullptr;
   if (!MRI.hasOneNonDBGUse(Reg))
-    return 0;
+    return nullptr;
   MachineInstr *MI = MRI.getVRegDef(Reg);
   if (!MI)
-    return 0;
+    return nullptr;
   // MI is folded into the MOVCC by predicating it.
   if (!MI->isPredicable())
-    return 0;
+    return nullptr;
   // Check if MI has any non-dead defs or physreg uses. This also detects
   // predicated instructions which will be reading CPSR.
   for (unsigned i = 1, e = MI->getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = MI->getOperand(i);
     // Reject frame index operands, PEI can't handle the predicated pseudos.
     if (MO.isFI() || MO.isCPI() || MO.isJTI())
-      return 0;
+      return nullptr;
     if (!MO.isReg())
       continue;
     // MI can't have any tied operands, that would conflict with predication.
     if (MO.isTied())
-      return 0;
+      return nullptr;
     if (TargetRegisterInfo::isPhysicalRegister(MO.getReg()))
-      return 0;
+      return nullptr;
     if (MO.isDef() && !MO.isDead())
-      return 0;
+      return nullptr;
   }
   bool DontMoveAcrossStores = true;
-  if (!MI->isSafeToMove(TII, /* AliasAnalysis = */ 0, DontMoveAcrossStores))
-    return 0;
+  if (!MI->isSafeToMove(TII, /* AliasAnalysis = */ nullptr,
+                        DontMoveAcrossStores))
+    return nullptr;
   return MI;
 }
 
@@ -1741,14 +1746,14 @@ MachineInstr *ARMBaseInstrInfo::optimizeSelect(MachineInstr *MI,
   if (!DefMI)
     DefMI = canFoldIntoMOVCC(MI->getOperand(1).getReg(), MRI, this);
   if (!DefMI)
-    return 0;
+    return nullptr;
 
   // Find new register class to use.
   MachineOperand FalseReg = MI->getOperand(Invert ? 2 : 1);
   unsigned       DestReg  = MI->getOperand(0).getReg();
   const TargetRegisterClass *PreviousClass = MRI.getRegClass(FalseReg.getReg());
   if (!MRI.constrainRegClass(DestReg, PreviousClass))
-    return 0;
+    return nullptr;
 
   // Create a new predicated version of DefMI.
   // Rfalse is the first use.
@@ -2254,7 +2259,7 @@ optimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg, unsigned SrcReg2,
   // Masked compares sometimes use the same register as the corresponding 'and'.
   if (CmpMask != ~0) {
     if (!isSuitableForMask(MI, SrcReg, CmpMask, false) || isPredicated(MI)) {
-      MI = 0;
+      MI = nullptr;
       for (MachineRegisterInfo::use_instr_iterator
            UI = MRI->use_instr_begin(SrcReg), UE = MRI->use_instr_end();
            UI != UE; ++UI) {
@@ -2281,17 +2286,17 @@ optimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg, unsigned SrcReg2,
   // One is MI, the other is a SUB instruction.
   // For CMPrr(r1,r2), we are looking for SUB(r1,r2) or SUB(r2,r1).
   // For CMPri(r1, CmpValue), we are looking for SUBri(r1, CmpValue).
-  MachineInstr *Sub = NULL;
+  MachineInstr *Sub = nullptr;
   if (SrcReg2 != 0)
     // MI is not a candidate for CMPrr.
-    MI = NULL;
+    MI = nullptr;
   else if (MI->getParent() != CmpInstr->getParent() || CmpValue != 0) {
     // Conservatively refuse to convert an instruction which isn't in the same
     // BB as the comparison.
     // For CMPri, we need to check Sub, thus we can't return here.
     if (CmpInstr->getOpcode() == ARM::CMPri ||
        CmpInstr->getOpcode() == ARM::t2CMPri)
-      MI = NULL;
+      MI = nullptr;
     else
       return false;
   }
@@ -3295,7 +3300,7 @@ static const MachineInstr *getBundledUseMI(const TargetRegisterInfo *TRI,
 
   if (Idx == -1) {
     Dist = 0;
-    return 0;
+    return nullptr;
   }
 
   UseIdx = Idx;
diff --git a/lib/Target/ARM/ARMBaseInstrInfo.h b/lib/Target/ARM/ARMBaseInstrInfo.h
index 3ddddcb..4b3e740 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.h
+++ b/lib/Target/ARM/ARMBaseInstrInfo.h
@@ -261,7 +261,7 @@ private:
 
   unsigned getInstrLatency(const InstrItineraryData *ItinData,
                            const MachineInstr *MI,
-                           unsigned *PredCost = 0) const override;
+                           unsigned *PredCost = nullptr) const override;
 
   int getInstrLatency(const InstrItineraryData *ItinData,
                       SDNode *Node) const override;
diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.cpp b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
index 8130a2d..a2eee9f 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.cpp
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
@@ -44,14 +44,18 @@
 using namespace llvm;
 
 ARMBaseRegisterInfo::ARMBaseRegisterInfo(const ARMSubtarget &sti)
-  : ARMGenRegisterInfo(ARM::LR, 0, 0, ARM::PC), STI(sti),
-    FramePtr((STI.isTargetMachO() || STI.isThumb()) ? ARM::R7 : ARM::R11),
-    BasePtr(ARM::R6) {
+    : ARMGenRegisterInfo(ARM::LR, 0, 0, ARM::PC), STI(sti), BasePtr(ARM::R6) {
+  if (STI.isTargetMachO())
+    FramePtr = ARM::R7;
+  else if (STI.isTargetWindows())
+    FramePtr = ARM::R11;
+  else // ARM EABI
+    FramePtr = STI.isThumb() ? ARM::R7 : ARM::R11;
 }
 
-const uint16_t*
+const MCPhysReg*
 ARMBaseRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
-  const uint16_t *RegList = (STI.isTargetIOS() && !STI.isAAPCS_ABI())
+  const MCPhysReg *RegList = (STI.isTargetIOS() && !STI.isAAPCS_ABI())
                                 ? CSR_iOS_SaveList
                                 : CSR_AAPCS_SaveList;
 
@@ -107,7 +111,7 @@ ARMBaseRegisterInfo::getThisReturnPreservedMask(CallingConv::ID CC) const {
   // should return NULL
   if (CC == CallingConv::GHC)
     // This is academic becase all GHC calls are (supposed to be) tail calls
-    return NULL;
+    return nullptr;
   return (STI.isTargetIOS() && !STI.isAAPCS_ABI())
     ? CSR_iOS_ThisReturn_RegMask : CSR_AAPCS_ThisReturn_RegMask;
 }
@@ -173,7 +177,7 @@ ARMBaseRegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind
 const TargetRegisterClass *
 ARMBaseRegisterInfo::getCrossCopyRegClass(const TargetRegisterClass *RC) const {
   if (RC == &ARM::CCRRegClass)
-    return 0;  // Can't copy CCR registers.
+    return nullptr;  // Can't copy CCR registers.
   return RC;
 }
 
diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.h b/lib/Target/ARM/ARMBaseRegisterInfo.h
index 66b3c82..91df565 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.h
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.h
@@ -100,8 +100,8 @@ protected:
 
 public:
   /// Code Generation virtual methods...
-  const uint16_t *
-  getCalleeSavedRegs(const MachineFunction *MF = 0) const override;
+  const MCPhysReg *
+  getCalleeSavedRegs(const MachineFunction *MF = nullptr) const override;
   const uint32_t *getCallPreservedMask(CallingConv::ID) const override;
   const uint32_t *getNoPreservedMask() const;
 
@@ -186,7 +186,7 @@ public:
 
   void eliminateFrameIndex(MachineBasicBlock::iterator II,
                            int SPAdj, unsigned FIOperandNum,
-                           RegScavenger *RS = NULL) const override;
+                           RegScavenger *RS = nullptr) const override;
 };
 
 } // end namespace llvm
diff --git a/lib/Target/ARM/ARMCallingConv.h b/lib/Target/ARM/ARMCallingConv.h
index 4f94ad2..dc41c1c 100644
--- a/lib/Target/ARM/ARMCallingConv.h
+++ b/lib/Target/ARM/ARMCallingConv.h
@@ -28,7 +28,7 @@ namespace llvm {
 static bool f64AssignAPCS(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
                           CCValAssign::LocInfo &LocInfo,
                           CCState &State, bool CanFail) {
-  static const uint16_t RegList[] = { ARM::R0, ARM::R1, ARM::R2, ARM::R3 };
+  static const MCPhysReg RegList[] = { ARM::R0, ARM::R1, ARM::R2, ARM::R3 };
 
   // Try to get the first register.
   if (unsigned Reg = State.AllocateReg(RegList, 4))
@@ -71,10 +71,10 @@ static bool CC_ARM_APCS_Custom_f64(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
 static bool f64AssignAAPCS(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
                            CCValAssign::LocInfo &LocInfo,
                            CCState &State, bool CanFail) {
-  static const uint16_t HiRegList[] = { ARM::R0, ARM::R2 };
-  static const uint16_t LoRegList[] = { ARM::R1, ARM::R3 };
-  static const uint16_t ShadowRegList[] = { ARM::R0, ARM::R1 };
-  static const uint16_t GPRArgRegs[] = { ARM::R0, ARM::R1, ARM::R2, ARM::R3 };
+  static const MCPhysReg HiRegList[] = { ARM::R0, ARM::R2 };
+  static const MCPhysReg LoRegList[] = { ARM::R1, ARM::R3 };
+  static const MCPhysReg ShadowRegList[] = { ARM::R0, ARM::R1 };
+  static const MCPhysReg GPRArgRegs[] = { ARM::R0, ARM::R1, ARM::R2, ARM::R3 };
 
   unsigned Reg = State.AllocateReg(HiRegList, ShadowRegList, 2);
   if (Reg == 0) {
@@ -123,8 +123,8 @@ static bool CC_ARM_AAPCS_Custom_f64(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
 
 static bool f64RetAssign(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
                          CCValAssign::LocInfo &LocInfo, CCState &State) {
-  static const uint16_t HiRegList[] = { ARM::R0, ARM::R2 };
-  static const uint16_t LoRegList[] = { ARM::R1, ARM::R3 };
+  static const MCPhysReg HiRegList[] = { ARM::R0, ARM::R2 };
+  static const MCPhysReg LoRegList[] = { ARM::R1, ARM::R3 };
 
   unsigned Reg = State.AllocateReg(HiRegList, LoRegList, 2);
   if (Reg == 0)
@@ -160,6 +160,105 @@ static bool RetCC_ARM_AAPCS_Custom_f64(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
                                    State);
 }
 
+static const uint16_t SRegList[] = { ARM::S0,  ARM::S1,  ARM::S2,  ARM::S3,
+                                     ARM::S4,  ARM::S5,  ARM::S6,  ARM::S7,
+                                     ARM::S8,  ARM::S9,  ARM::S10, ARM::S11,
+                                     ARM::S12, ARM::S13, ARM::S14,  ARM::S15 };
+static const uint16_t DRegList[] = { ARM::D0, ARM::D1, ARM::D2, ARM::D3,
+                                     ARM::D4, ARM::D5, ARM::D6, ARM::D7 };
+static const uint16_t QRegList[] = { ARM::Q0, ARM::Q1, ARM::Q2, ARM::Q3 };
+
+// Allocate part of an AAPCS HFA or HVA. We assume that each member of the HA
+// has InConsecutiveRegs set, and that the last member also has
+// InConsecutiveRegsLast set. We must process all members of the HA before
+// we can allocate it, as we need to know the total number of registers that
+// will be needed in order to (attempt to) allocate a contiguous block.
+static bool CC_ARM_AAPCS_Custom_HA(unsigned &ValNo, MVT &ValVT, MVT &LocVT,
+                                   CCValAssign::LocInfo &LocInfo,
+                                   ISD::ArgFlagsTy &ArgFlags, CCState &State) {
+  SmallVectorImpl<CCValAssign> &PendingHAMembers = State.getPendingLocs();
+  // AAPCS HFAs must have 1-4 elements, all of the same type
+  assert(PendingHAMembers.size() < 8);
+  if (PendingHAMembers.size() > 0)
+    assert(PendingHAMembers[0].getLocVT() == LocVT);
+
+  // Add the argument to the list to be allocated once we know the size of the
+  // HA
+  PendingHAMembers.push_back(
+      CCValAssign::getPending(ValNo, ValVT, LocVT, LocInfo));
+
+  if (ArgFlags.isInConsecutiveRegsLast()) {
+    assert(PendingHAMembers.size() > 0 && PendingHAMembers.size() <= 8 &&
+           "Homogeneous aggregates must have between 1 and 4 members");
+
+    // Try to allocate a contiguous block of registers, each of the correct
+    // size to hold one member.
+    const uint16_t *RegList;
+    unsigned NumRegs;
+    switch (LocVT.SimpleTy) {
+    case MVT::i32:
+    case MVT::f32:
+      RegList = SRegList;
+      NumRegs = 16;
+      break;
+    case MVT::f64:
+      RegList = DRegList;
+      NumRegs = 8;
+      break;
+    case MVT::v2f64:
+      RegList = QRegList;
+      NumRegs = 4;
+      break;
+    default:
+      llvm_unreachable("Unexpected member type for HA");
+      break;
+    }
+
+    unsigned RegResult =
+        State.AllocateRegBlock(RegList, NumRegs, PendingHAMembers.size());
+
+    if (RegResult) {
+      for (SmallVectorImpl<CCValAssign>::iterator It = PendingHAMembers.begin();
+           It != PendingHAMembers.end(); ++It) {
+        It->convertToReg(RegResult);
+        State.addLoc(*It);
+        ++RegResult;
+      }
+      PendingHAMembers.clear();
+      return true;
+    }
+
+    // Register allocation failed, fall back to the stack
+
+    // Mark all VFP regs as unavailable (AAPCS rule C.2.vfp)
+    for (unsigned regNo = 0; regNo < 16; ++regNo)
+      State.AllocateReg(SRegList[regNo]);
+
+    unsigned Size = LocVT.getSizeInBits() / 8;
+    unsigned Align = Size;
+
+    if (LocVT.SimpleTy == MVT::v2f64 || LocVT.SimpleTy == MVT::i32) {
+      // Vectors are always aligned to 8 bytes. If we've seen an i32 here
+      // it's because it's been split from a larger type, also with align 8.
+      Align = 8;
+    }
+
+    for (auto It : PendingHAMembers) {
+      It.convertToMem(State.AllocateStack(Size, Align));
+      State.addLoc(It);
+
+      // Only the first member needs to be aligned.
+      Align = 1;
+    }
+
+    // All pending members have now been allocated
+    PendingHAMembers.clear();
+  }
+
+  // This will be allocated by the last member of the HA
+  return true;
+}
+
 } // End llvm namespace
 
 #endif
diff --git a/lib/Target/ARM/ARMCallingConv.td b/lib/Target/ARM/ARMCallingConv.td
index 7cffd82..526089b 100644
--- a/lib/Target/ARM/ARMCallingConv.td
+++ b/lib/Target/ARM/ARMCallingConv.td
@@ -174,6 +174,9 @@ def CC_ARM_AAPCS_VFP : CallingConv<[
   CCIfType<[v1i64, v2i32, v4i16, v8i8, v2f32], CCBitConvertToType<f64>>,
   CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32], CCBitConvertToType<v2f64>>,
 
+  // HFAs are passed in a contiguous block of registers, or on the stack
+  CCIfConsecutiveRegs<CCCustom<"CC_ARM_AAPCS_Custom_HA">>,
+
   CCIfType<[v2f64], CCAssignToReg<[Q0, Q1, Q2, Q3]>>,
   CCIfType<[f64], CCAssignToReg<[D0, D1, D2, D3, D4, D5, D6, D7]>>,
   CCIfType<[f32], CCAssignToReg<[S0, S1, S2, S3, S4, S5, S6, S7, S8,
diff --git a/lib/Target/ARM/ARMCodeEmitter.cpp b/lib/Target/ARM/ARMCodeEmitter.cpp
index 7359a11..2fd7edd 100644
--- a/lib/Target/ARM/ARMCodeEmitter.cpp
+++ b/lib/Target/ARM/ARMCodeEmitter.cpp
@@ -12,7 +12,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "jit"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMConstantPoolValue.h"
@@ -40,6 +39,8 @@
 #endif
 using namespace llvm;
 
+#define DEBUG_TYPE "jit"
+
 STATISTIC(NumEmitted, "Number of machine instructions emitted");
 
 namespace {
@@ -65,10 +66,10 @@ namespace {
     static char ID;
   public:
     ARMCodeEmitter(TargetMachine &tm, JITCodeEmitter &mce)
-      : MachineFunctionPass(ID), JTI(0),
+      : MachineFunctionPass(ID), JTI(nullptr),
         II((const ARMBaseInstrInfo *)tm.getInstrInfo()),
         TD(tm.getDataLayout()), TM(tm),
-        MCE(mce), MCPEs(0), MJTEs(0),
+        MCE(mce), MCPEs(nullptr), MJTEs(nullptr),
         IsPIC(TM.getRelocationModel() == Reloc::PIC_), IsThumb(false) {}
 
     /// getBinaryCodeForInstr - This function, generated by the
@@ -373,7 +374,7 @@ bool ARMCodeEmitter::runOnMachineFunction(MachineFunction &MF) {
 
   Subtarget = &TM.getSubtarget<ARMSubtarget>();
   MCPEs = &MF.getConstantPool()->getConstants();
-  MJTEs = 0;
+  MJTEs = nullptr;
   if (MF.getJumpTableInfo()) MJTEs = &MF.getJumpTableInfo()->getJumpTables();
   IsPIC = TM.getRelocationModel() == Reloc::PIC_;
   IsThumb = MF.getInfo<ARMFunctionInfo>()->isThumbFunction();
diff --git a/lib/Target/ARM/ARMConstantIslandPass.cpp b/lib/Target/ARM/ARMConstantIslandPass.cpp
index ba05171..ce264ee 100644
--- a/lib/Target/ARM/ARMConstantIslandPass.cpp
+++ b/lib/Target/ARM/ARMConstantIslandPass.cpp
@@ -13,7 +13,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-cp-islands"
 #include "ARM.h"
 #include "ARMMachineFunctionInfo.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
@@ -36,6 +35,8 @@
 #include <algorithm>
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-cp-islands"
+
 STATISTIC(NumCPEs,       "Number of constpool entries");
 STATISTIC(NumSplit,      "Number of uncond branches inserted");
 STATISTIC(NumCBrFixed,   "Number of cond branches fixed");
@@ -593,7 +594,7 @@ ARMConstantIslands::CPEntry
     if (CPEs[i].CPEMI == CPEMI)
       return &CPEs[i];
   }
-  return NULL;
+  return nullptr;
 }
 
 /// getCPELogAlign - Returns the required alignment of the constant pool entry
@@ -1102,7 +1103,7 @@ bool ARMConstantIslands::decrementCPEReferenceCount(unsigned CPI,
   assert(CPE && "Unexpected!");
   if (--CPE->RefCount == 0) {
     removeDeadCPEMI(CPEMI);
-    CPE->CPEMI = NULL;
+    CPE->CPEMI = nullptr;
     --NumCPEs;
     return true;
   }
@@ -1135,7 +1136,7 @@ int ARMConstantIslands::findInRangeCPEntry(CPUser& U, unsigned UserOffset)
     if (CPEs[i].CPEMI == CPEMI)
       continue;
     // Removing CPEs can leave empty entries, skip
-    if (CPEs[i].CPEMI == NULL)
+    if (CPEs[i].CPEMI == nullptr)
       continue;
     if (isCPEntryInRange(UserMI, UserOffset, CPEs[i].CPEMI, U.getMaxDisp(),
                      U.NegOk)) {
@@ -1317,7 +1318,7 @@ void ARMConstantIslands::createNewWater(unsigned CPUserIndex,
   ++MI;
   unsigned CPUIndex = CPUserIndex+1;
   unsigned NumCPUsers = CPUsers.size();
-  MachineInstr *LastIT = 0;
+  MachineInstr *LastIT = nullptr;
   for (unsigned Offset = UserOffset+TII->GetInstSizeInBytes(UserMI);
        Offset < BaseInsertOffset;
        Offset += TII->GetInstSizeInBytes(MI), MI = std::next(MI)) {
@@ -1491,7 +1492,7 @@ bool ARMConstantIslands::removeUnusedCPEntries() {
       for (unsigned j = 0, ee = CPEs.size(); j != ee; ++j) {
         if (CPEs[j].RefCount == 0 && CPEs[j].CPEMI) {
           removeDeadCPEMI(CPEs[j].CPEMI);
-          CPEs[j].CPEMI = NULL;
+          CPEs[j].CPEMI = nullptr;
           MadeChange = true;
         }
       }
@@ -1844,7 +1845,7 @@ bool ARMConstantIslands::optimizeThumb2JumpTables() {
   // FIXME: After the tables are shrunk, can we get rid some of the
   // constantpool tables?
   MachineJumpTableInfo *MJTI = MF->getJumpTableInfo();
-  if (MJTI == 0) return false;
+  if (!MJTI) return false;
 
   const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();
   for (unsigned i = 0, e = T2JumpTables.size(); i != e; ++i) {
@@ -1970,7 +1971,7 @@ bool ARMConstantIslands::reorderThumb2JumpTables() {
   bool MadeChange = false;
 
   MachineJumpTableInfo *MJTI = MF->getJumpTableInfo();
-  if (MJTI == 0) return false;
+  if (!MJTI) return false;
 
   const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();
   for (unsigned i = 0, e = T2JumpTables.size(); i != e; ++i) {
@@ -2012,7 +2013,7 @@ adjustJTTargetBlockForward(MachineBasicBlock *BB, MachineBasicBlock *JTBB) {
   // try to move it; otherwise, create a new block following the jump
   // table that branches back to the actual target. This is a very simple
   // heuristic. FIXME: We can definitely improve it.
-  MachineBasicBlock *TBB = 0, *FBB = 0;
+  MachineBasicBlock *TBB = nullptr, *FBB = nullptr;
   SmallVector<MachineOperand, 4> Cond;
   SmallVector<MachineOperand, 4> CondPrior;
   MachineFunction::iterator BBi = BB;
@@ -2032,7 +2033,7 @@ adjustJTTargetBlockForward(MachineBasicBlock *BB, MachineBasicBlock *JTBB) {
     // Update numbering to account for the block being moved.
     MF->RenumberBlocks();
     ++NumJTMoved;
-    return NULL;
+    return nullptr;
   }
 
   // Create a new MBB for the code after the jump BB.
diff --git a/lib/Target/ARM/ARMExpandPseudoInsts.cpp b/lib/Target/ARM/ARMExpandPseudoInsts.cpp
index bd4ee44..6045738 100644
--- a/lib/Target/ARM/ARMExpandPseudoInsts.cpp
+++ b/lib/Target/ARM/ARMExpandPseudoInsts.cpp
@@ -14,7 +14,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-pseudo"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMBaseRegisterInfo.h"
@@ -23,6 +22,7 @@
 #include "MCTargetDesc/ARMAddressingModes.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/MachineInstrBundle.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/IR/GlobalValue.h"
 #include "llvm/Support/CommandLine.h"
@@ -31,6 +31,8 @@
 #include "llvm/Target/TargetRegisterInfo.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-pseudo"
+
 static cl::opt<bool>
 VerifyARMPseudo("verify-arm-pseudo-expand", cl::Hidden,
                 cl::desc("Verify machine code after expanding ARM pseudos"));
@@ -345,7 +347,7 @@ static const NEONLdStTableEntry *LookupNEONLdSt(unsigned Opcode) {
     std::lower_bound(NEONLdStTable, NEONLdStTable + NumEntries, Opcode);
   if (I != NEONLdStTable + NumEntries && I->PseudoOpc == Opcode)
     return I;
-  return NULL;
+  return nullptr;
 }
 
 /// GetDSubRegs - Get 4 D subregisters of a Q, QQ, or QQQQ register,
@@ -614,6 +616,39 @@ void ARMExpandPseudo::ExpandVTBL(MachineBasicBlock::iterator &MBBI,
   MI.eraseFromParent();
 }
 
+static bool IsAnAddressOperand(const MachineOperand &MO) {
+  // This check is overly conservative.  Unless we are certain that the machine
+  // operand is not a symbol reference, we return that it is a symbol reference.
+  // This is important as the load pair may not be split up Windows.
+  switch (MO.getType()) {
+  case MachineOperand::MO_Register:
+  case MachineOperand::MO_Immediate:
+  case MachineOperand::MO_CImmediate:
+  case MachineOperand::MO_FPImmediate:
+    return false;
+  case MachineOperand::MO_MachineBasicBlock:
+    return true;
+  case MachineOperand::MO_FrameIndex:
+    return false;
+  case MachineOperand::MO_ConstantPoolIndex:
+  case MachineOperand::MO_TargetIndex:
+  case MachineOperand::MO_JumpTableIndex:
+  case MachineOperand::MO_ExternalSymbol:
+  case MachineOperand::MO_GlobalAddress:
+  case MachineOperand::MO_BlockAddress:
+    return true;
+  case MachineOperand::MO_RegisterMask:
+  case MachineOperand::MO_RegisterLiveOut:
+    return false;
+  case MachineOperand::MO_Metadata:
+  case MachineOperand::MO_MCSymbol:
+    return true;
+  case MachineOperand::MO_CFIIndex:
+    return false;
+  }
+  llvm_unreachable("unhandled machine operand type");
+}
+
 void ARMExpandPseudo::ExpandMOV32BitImm(MachineBasicBlock &MBB,
                                         MachineBasicBlock::iterator &MBBI) {
   MachineInstr &MI = *MBBI;
@@ -624,10 +659,14 @@ void ARMExpandPseudo::ExpandMOV32BitImm(MachineBasicBlock &MBB,
   bool DstIsDead = MI.getOperand(0).isDead();
   bool isCC = Opcode == ARM::MOVCCi32imm || Opcode == ARM::t2MOVCCi32imm;
   const MachineOperand &MO = MI.getOperand(isCC ? 2 : 1);
+  bool RequiresBundling = STI->isTargetWindows() && IsAnAddressOperand(MO);
   MachineInstrBuilder LO16, HI16;
 
   if (!STI->hasV6T2Ops() &&
       (Opcode == ARM::MOVi32imm || Opcode == ARM::MOVCCi32imm)) {
+    // FIXME Windows CE supports older ARM CPUs
+    assert(!STI->isTargetWindows() && "Windows on ARM requires ARMv7+");
+
     // Expand into a movi + orr.
     LO16 = BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(ARM::MOVi), DstReg);
     HI16 = BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(ARM::ORRri))
@@ -664,17 +703,29 @@ void ARMExpandPseudo::ExpandMOV32BitImm(MachineBasicBlock &MBB,
     .addReg(DstReg, RegState::Define | getDeadRegState(DstIsDead))
     .addReg(DstReg);
 
-  if (MO.isImm()) {
+  switch (MO.getType()) {
+  case MachineOperand::MO_Immediate: {
     unsigned Imm = MO.getImm();
     unsigned Lo16 = Imm & 0xffff;
     unsigned Hi16 = (Imm >> 16) & 0xffff;
     LO16 = LO16.addImm(Lo16);
     HI16 = HI16.addImm(Hi16);
-  } else {
+    break;
+  }
+  case MachineOperand::MO_ExternalSymbol: {
+    const char *ES = MO.getSymbolName();
+    unsigned TF = MO.getTargetFlags();
+    LO16 = LO16.addExternalSymbol(ES, TF | ARMII::MO_LO16);
+    HI16 = HI16.addExternalSymbol(ES, TF | ARMII::MO_HI16);
+    break;
+  }
+  default: {
     const GlobalValue *GV = MO.getGlobal();
     unsigned TF = MO.getTargetFlags();
     LO16 = LO16.addGlobalAddress(GV, MO.getOffset(), TF | ARMII::MO_LO16);
     HI16 = HI16.addGlobalAddress(GV, MO.getOffset(), TF | ARMII::MO_HI16);
+    break;
+  }
   }
 
   LO16->setMemRefs(MI.memoperands_begin(), MI.memoperands_end());
@@ -682,6 +733,9 @@ void ARMExpandPseudo::ExpandMOV32BitImm(MachineBasicBlock &MBB,
   LO16.addImm(Pred).addReg(PredReg);
   HI16.addImm(Pred).addReg(PredReg);
 
+  if (RequiresBundling)
+    finalizeBundle(MBB, &*LO16, &*MBBI);
+
   TransferImpOps(MI, LO16, HI16);
   MI.eraseFromParent();
 }
diff --git a/lib/Target/ARM/ARMFastISel.cpp b/lib/Target/ARM/ARMFastISel.cpp
index c442444..6f8fb1a 100644
--- a/lib/Target/ARM/ARMFastISel.cpp
+++ b/lib/Target/ARM/ARMFastISel.cpp
@@ -166,8 +166,6 @@ class ARMFastISel final : public FastISel {
 
     // Utility routines.
   private:
-    unsigned constrainOperandRegClass(const MCInstrDesc &II, unsigned OpNum,
-                                      unsigned Op);
     bool isTypeLegal(Type *Ty, MVT &VT);
     bool isLoadTypeLegal(Type *Ty, MVT &VT);
     bool ARMEmitCmp(const Value *Src1Value, const Value *Src2Value,
@@ -191,6 +189,8 @@ class ARMFastISel final : public FastISel {
     unsigned ARMSelectCallOp(bool UseReg);
     unsigned ARMLowerPICELF(const GlobalValue *GV, unsigned Align, MVT VT);
 
+    const TargetLowering *getTargetLowering() { return TM.getTargetLowering(); }
+
     // Call handling routines.
   private:
     CCAssignFn *CCAssignFnForCall(CallingConv::ID CC,
@@ -283,23 +283,6 @@ ARMFastISel::AddOptionalDefs(const MachineInstrBuilder &MIB) {
   return MIB;
 }
 
-unsigned ARMFastISel::constrainOperandRegClass(const MCInstrDesc &II,
-                                               unsigned Op, unsigned OpNum) {
-  if (TargetRegisterInfo::isVirtualRegister(Op)) {
-    const TargetRegisterClass *RegClass =
-        TII.getRegClass(II, OpNum, &TRI, *FuncInfo.MF);
-    if (!MRI.constrainRegClass(Op, RegClass)) {
-      // If it's not legal to COPY between the register classes, something
-      // has gone very wrong before we got here.
-      unsigned NewOp = createResultReg(RegClass);
-      AddOptionalDefs(BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
-                              TII.get(TargetOpcode::COPY), NewOp).addReg(Op));
-      return NewOp;
-    }
-  }
-  return Op;
-}
-
 unsigned ARMFastISel::FastEmitInst_r(unsigned MachineInstOpcode,
                                      const TargetRegisterClass *RC,
                                      unsigned Op0, bool Op0IsKill) {
@@ -769,7 +752,7 @@ bool ARMFastISel::isLoadTypeLegal(Type *Ty, MVT &VT) {
 // Computes the address to get to an object.
 bool ARMFastISel::ARMComputeAddress(const Value *Obj, Address &Addr) {
   // Some boilerplate from the X86 FastISel.
-  const User *U = NULL;
+  const User *U = nullptr;
   unsigned Opcode = Instruction::UserOp1;
   if (const Instruction *I = dyn_cast<Instruction>(Obj)) {
     // Don't walk into other basic blocks unless the object is an alloca from
@@ -1400,7 +1383,7 @@ bool ARMFastISel::ARMEmitCmp(const Value *Src1Value, const Value *Src2Value,
       const APInt &CIVal = ConstInt->getValue();
       Imm = (isZExt) ? (int)CIVal.getZExtValue() : (int)CIVal.getSExtValue();
       // For INT_MIN/LONG_MIN (i.e., 0x80000000) we need to use a cmp, rather
-      // then a cmn, because there is no way to represent 2147483648 as a 
+      // then a cmn, because there is no way to represent 2147483648 as a
       // signed 32-bit int.
       if (Imm < 0 && Imm != (int)0x80000000) {
         isNegativeImm = true;
@@ -2182,7 +2165,8 @@ unsigned ARMFastISel::getLibcallReg(const Twine &Name) {
   if (!LCREVT.isSimple()) return 0;
 
   GlobalValue *GV = new GlobalVariable(M, Type::getInt32Ty(*Context), false,
-                                       GlobalValue::ExternalLinkage, 0, Name);
+                                       GlobalValue::ExternalLinkage, nullptr,
+                                       Name);
   assert(GV->getType() == GVTy && "We miscomputed the type for the global!");
   return ARMMaterializeGV(GV, LCREVT.getSimpleVT());
 }
@@ -2286,7 +2270,7 @@ bool ARMFastISel::ARMEmitLibcall(const Instruction *I, RTLIB::Libcall Call) {
 }
 
 bool ARMFastISel::SelectCall(const Instruction *I,
-                             const char *IntrMemName = 0) {
+                             const char *IntrMemName = nullptr) {
   const CallInst *CI = cast<CallInst>(I);
   const Value *Callee = CI->getCalledValue();
 
@@ -3092,6 +3076,6 @@ namespace llvm {
       TM.Options.NoFramePointerElim = true;
       return new ARMFastISel(funcInfo, libInfo);
     }
-    return 0;
+    return nullptr;
   }
 }
diff --git a/lib/Target/ARM/ARMFeatures.h b/lib/Target/ARM/ARMFeatures.h
index a30f4cd..e191a3c 100644
--- a/lib/Target/ARM/ARMFeatures.h
+++ b/lib/Target/ARM/ARMFeatures.h
@@ -1,4 +1,4 @@
-//===-- ARMFeatures.h - Checks for ARM instruction features ------*- C++ -*-===//
+//===-- ARMFeatures.h - Checks for ARM instruction features -----*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -16,11 +16,11 @@
 
 #include "MCTargetDesc/ARMMCTargetDesc.h"
 
+namespace llvm {
+
 template<typename InstrType> // could be MachineInstr or MCInst
 bool IsCPSRDead(InstrType *Instr);
 
-namespace llvm {
-
 template<typename InstrType> // could be MachineInstr or MCInst
 inline bool isV8EligibleForIT(InstrType *Instr) {
   switch (Instr->getOpcode()) {
diff --git a/lib/Target/ARM/ARMFrameLowering.cpp b/lib/Target/ARM/ARMFrameLowering.cpp
index 36ecfca..0caf4bf 100644
--- a/lib/Target/ARM/ARMFrameLowering.cpp
+++ b/lib/Target/ARM/ARMFrameLowering.cpp
@@ -87,7 +87,7 @@ ARMFrameLowering::canSimplifyCallFramePseudos(const MachineFunction &MF) const {
 
 static bool isCSRestore(MachineInstr *MI,
                         const ARMBaseInstrInfo &TII,
-                        const uint16_t *CSRegs) {
+                        const MCPhysReg *CSRegs) {
   // Integer spill area is handled with "pop".
   if (isPopOpcode(MI->getOpcode())) {
     // The first two operands are predicates. The last two are
@@ -142,6 +142,14 @@ static int sizeOfSPAdjustment(const MachineInstr *MI) {
   return count;
 }
 
+static bool WindowsRequiresStackProbe(const MachineFunction &MF,
+                                      size_t StackSizeInBytes) {
+  const MachineFrameInfo *MFI = MF.getFrameInfo();
+  if (MFI->getStackProtectorIndex() > 0)
+    return StackSizeInBytes >= 4080;
+  return StackSizeInBytes >= 4096;
+}
+
 void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
   MachineBasicBlock &MBB = MF.front();
   MachineBasicBlock::iterator MBBI = MBB.begin();
@@ -149,15 +157,16 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
   ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
   MachineModuleInfo &MMI = MF.getMMI();
   MCContext &Context = MMI.getContext();
+  const TargetMachine &TM = MF.getTarget();
   const MCRegisterInfo *MRI = Context.getRegisterInfo();
   const ARMBaseRegisterInfo *RegInfo =
-    static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
+    static_cast<const ARMBaseRegisterInfo*>(TM.getRegisterInfo());
   const ARMBaseInstrInfo &TII =
-    *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
+    *static_cast<const ARMBaseInstrInfo*>(TM.getInstrInfo());
   assert(!AFI->isThumb1OnlyFunction() &&
          "This emitPrologue does not support Thumb1!");
   bool isARM = !AFI->isThumbFunction();
-  unsigned Align = MF.getTarget().getFrameLowering()->getStackAlignment();
+  unsigned Align = TM.getFrameLowering()->getStackAlignment();
   unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize(Align);
   unsigned NumBytes = MFI->getStackSize();
   const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
@@ -187,7 +196,8 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
         .addCFIIndex(CFIIndex);
   }
 
-  if (!AFI->hasStackFrame()) {
+  if (!AFI->hasStackFrame() &&
+      (!STI.isTargetWindows() || !WindowsRequiresStackProbe(MF, NumBytes))) {
     if (NumBytes - ArgRegsSaveSize != 0) {
       emitSPUpdate(isARM, MBB, MBBI, dl, TII, -(NumBytes - ArgRegsSaveSize),
                    MachineInstr::FrameSetup);
@@ -284,6 +294,51 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
   } else
     NumBytes = DPRCSOffset;
 
+  if (STI.isTargetWindows() && WindowsRequiresStackProbe(MF, NumBytes)) {
+    uint32_t NumWords = NumBytes >> 2;
+
+    if (NumWords < 65536)
+      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi16), ARM::R4)
+                     .addImm(NumWords)
+                     .setMIFlags(MachineInstr::FrameSetup));
+    else
+      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R4)
+        .addImm(NumWords)
+        .setMIFlags(MachineInstr::FrameSetup);
+
+    switch (TM.getCodeModel()) {
+    case CodeModel::Small:
+    case CodeModel::Medium:
+    case CodeModel::Default:
+    case CodeModel::Kernel:
+      BuildMI(MBB, MBBI, dl, TII.get(ARM::tBL))
+        .addImm((unsigned)ARMCC::AL).addReg(0)
+        .addExternalSymbol("__chkstk")
+        .addReg(ARM::R4, RegState::Implicit)
+        .setMIFlags(MachineInstr::FrameSetup);
+      break;
+    case CodeModel::Large:
+    case CodeModel::JITDefault:
+      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R12)
+        .addExternalSymbol("__chkstk")
+        .setMIFlags(MachineInstr::FrameSetup);
+
+      BuildMI(MBB, MBBI, dl, TII.get(ARM::tBLXr))
+        .addImm((unsigned)ARMCC::AL).addReg(0)
+        .addReg(ARM::R12, RegState::Kill)
+        .addReg(ARM::R4, RegState::Implicit)
+        .setMIFlags(MachineInstr::FrameSetup);
+      break;
+    }
+
+    AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::t2SUBrr),
+                                        ARM::SP)
+                                .addReg(ARM::SP, RegState::Define)
+                                .addReg(ARM::R4, RegState::Kill)
+                                .setMIFlags(MachineInstr::FrameSetup)));
+    NumBytes = 0;
+  }
+
   unsigned adjustedGPRCS1Size = GPRCS1Size;
   if (NumBytes) {
     // Adjust SP after all the callee-save spills.
@@ -316,10 +371,9 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
     MachineBasicBlock::iterator Pos = ++GPRCS1Push;
     BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
         .addCFIIndex(CFIIndex);
-    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
-           E = CSI.end(); I != E; ++I) {
-      unsigned Reg = I->getReg();
-      int FI = I->getFrameIdx();
+    for (const auto &Entry : CSI) {
+      unsigned Reg = Entry.getReg();
+      int FI = Entry.getFrameIdx();
       switch (Reg) {
       case ARM::R8:
       case ARM::R9:
@@ -382,10 +436,9 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
       BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
           .addCFIIndex(CFIIndex);
     }
-    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
-           E = CSI.end(); I != E; ++I) {
-      unsigned Reg = I->getReg();
-      int FI = I->getFrameIdx();
+    for (const auto &Entry : CSI) {
+      unsigned Reg = Entry.getReg();
+      int FI = Entry.getFrameIdx();
       switch (Reg) {
       case ARM::R8:
       case ARM::R9:
@@ -411,7 +464,7 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
     do {
       MachineBasicBlock::iterator Push = DPRCSPush++;
       if (!HasFP) {
-        CFAOffset -= sizeOfSPAdjustment(Push);;
+        CFAOffset -= sizeOfSPAdjustment(Push);
         unsigned CFIIndex = MMI.addFrameInst(
             MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
         BuildMI(MBB, DPRCSPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
@@ -419,10 +472,9 @@ void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
       }
     } while (DPRCSPush->getOpcode() == ARM::VSTMDDB_UPD);
 
-    for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
-           E = CSI.end(); I != E; ++I) {
-      unsigned Reg = I->getReg();
-      int FI = I->getFrameIdx();
+    for (const auto &Entry : CSI) {
+      unsigned Reg = Entry.getReg();
+      int FI = Entry.getFrameIdx();
       if ((Reg >= ARM::D0 && Reg <= ARM::D31) &&
           (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())) {
         unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true);
@@ -540,7 +592,7 @@ void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
       emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes - ArgRegsSaveSize);
   } else {
     // Unwind MBBI to point to first LDR / VLDRD.
-    const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
+    const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
     if (MBBI != MBB.begin()) {
       do {
         --MBBI;
@@ -1205,12 +1257,9 @@ bool ARMFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
 static unsigned GetFunctionSizeInBytes(const MachineFunction &MF,
                                        const ARMBaseInstrInfo &TII) {
   unsigned FnSize = 0;
-  for (MachineFunction::const_iterator MBBI = MF.begin(), E = MF.end();
-       MBBI != E; ++MBBI) {
-    const MachineBasicBlock &MBB = *MBBI;
-    for (MachineBasicBlock::const_iterator I = MBB.begin(),E = MBB.end();
-         I != E; ++I)
-      FnSize += TII.GetInstSizeInBytes(I);
+  for (auto &MBB : MF) {
+    for (auto &MI : MBB)
+      FnSize += TII.GetInstSizeInBytes(&MI);
   }
   return FnSize;
 }
@@ -1223,21 +1272,21 @@ static unsigned estimateRSStackSizeLimit(MachineFunction &MF,
                                          const TargetFrameLowering *TFI) {
   const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
   unsigned Limit = (1 << 12) - 1;
-  for (MachineFunction::iterator BB = MF.begin(),E = MF.end(); BB != E; ++BB) {
-    for (MachineBasicBlock::iterator I = BB->begin(), E = BB->end();
-         I != E; ++I) {
-      for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
-        if (!I->getOperand(i).isFI()) continue;
+  for (auto &MBB : MF) {
+    for (auto &MI : MBB) {
+      for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+        if (!MI.getOperand(i).isFI())
+          continue;
 
         // When using ADDri to get the address of a stack object, 255 is the
         // largest offset guaranteed to fit in the immediate offset.
-        if (I->getOpcode() == ARM::ADDri) {
+        if (MI.getOpcode() == ARM::ADDri) {
           Limit = std::min(Limit, (1U << 8) - 1);
           break;
         }
 
         // Otherwise check the addressing mode.
-        switch (I->getDesc().TSFlags & ARMII::AddrModeMask) {
+        switch (MI.getDesc().TSFlags & ARMII::AddrModeMask) {
         case ARMII::AddrMode3:
         case ARMII::AddrModeT2_i8:
           Limit = std::min(Limit, (1U << 8) - 1);
@@ -1374,7 +1423,7 @@ ARMFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
 
   // Don't spill FP if the frame can be eliminated. This is determined
   // by scanning the callee-save registers to see if any is used.
-  const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
+  const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
   for (unsigned i = 0; CSRegs[i]; ++i) {
     unsigned Reg = CSRegs[i];
     bool Spilled = false;
@@ -1486,6 +1535,10 @@ ARMFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
 
     if (hasFP(MF)) {
       MRI.setPhysRegUsed(FramePtr);
+      auto FPPos = std::find(UnspilledCS1GPRs.begin(), UnspilledCS1GPRs.end(),
+                             FramePtr);
+      if (FPPos != UnspilledCS1GPRs.end())
+        UnspilledCS1GPRs.erase(FPPos);
       NumGPRSpills++;
     }
 
@@ -1681,7 +1734,7 @@ void ARMFrameLowering::adjustForSegmentedStacks(MachineFunction &MF) const {
   if (MF.getFunction()->isVarArg())
     report_fatal_error("Segmented stacks do not support vararg functions.");
   if (!ST->isTargetAndroid() && !ST->isTargetLinux())
-    report_fatal_error("Segmented stacks not supported on this platfrom.");
+    report_fatal_error("Segmented stacks not supported on this platform.");
 
   MachineBasicBlock &prologueMBB = MF.front();
   MachineFrameInfo *MFI = MF.getFrameInfo();
@@ -1693,6 +1746,12 @@ void ARMFrameLowering::adjustForSegmentedStacks(MachineFunction &MF) const {
   ARMFunctionInfo *ARMFI = MF.getInfo<ARMFunctionInfo>();
   DebugLoc DL;
 
+  uint64_t StackSize = MFI->getStackSize();
+
+  // Do not generate a prologue for functions with a stack of size zero
+  if (StackSize == 0)
+    return;
+
   // Use R4 and R5 as scratch registers.
   // We save R4 and R5 before use and restore them before leaving the function.
   unsigned ScratchReg0 = ARM::R4;
@@ -1722,8 +1781,6 @@ void ARMFrameLowering::adjustForSegmentedStacks(MachineFunction &MF) const {
   MF.push_front(PrevStackMBB);
 
   // The required stack size that is aligned to ARM constant criterion.
-  uint64_t StackSize = MFI->getStackSize();
-
   AlignedStackSize = alignToARMConstant(StackSize);
 
   // When the frame size is less than 256 we just compare the stack
diff --git a/lib/Target/ARM/ARMFrameLowering.h b/lib/Target/ARM/ARMFrameLowering.h
index 524ee36..981d320 100644
--- a/lib/Target/ARM/ARMFrameLowering.h
+++ b/lib/Target/ARM/ARMFrameLowering.h
@@ -57,7 +57,7 @@ public:
   void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
                                             RegScavenger *RS) const override;
 
-  void adjustForSegmentedStacks(MachineFunction &MF) const;
+  void adjustForSegmentedStacks(MachineFunction &MF) const override;
 
  private:
   void emitPushInst(MachineBasicBlock &MBB, MachineBasicBlock::iterator MI,
diff --git a/lib/Target/ARM/ARMHazardRecognizer.cpp b/lib/Target/ARM/ARMHazardRecognizer.cpp
index 61d4e12..0885c4e 100644
--- a/lib/Target/ARM/ARMHazardRecognizer.cpp
+++ b/lib/Target/ARM/ARMHazardRecognizer.cpp
@@ -77,7 +77,7 @@ ARMHazardRecognizer::getHazardType(SUnit *SU, int Stalls) {
 }
 
 void ARMHazardRecognizer::Reset() {
-  LastMI = 0;
+  LastMI = nullptr;
   FpMLxStalls = 0;
   ScoreboardHazardRecognizer::Reset();
 }
@@ -95,7 +95,7 @@ void ARMHazardRecognizer::EmitInstruction(SUnit *SU) {
 void ARMHazardRecognizer::AdvanceCycle() {
   if (FpMLxStalls && --FpMLxStalls == 0)
     // Stalled for 4 cycles but still can't schedule any other instructions.
-    LastMI = 0;
+    LastMI = nullptr;
   ScoreboardHazardRecognizer::AdvanceCycle();
 }
 
diff --git a/lib/Target/ARM/ARMHazardRecognizer.h b/lib/Target/ARM/ARMHazardRecognizer.h
index e88cd0d..a8198e2 100644
--- a/lib/Target/ARM/ARMHazardRecognizer.h
+++ b/lib/Target/ARM/ARMHazardRecognizer.h
@@ -35,7 +35,7 @@ public:
   ARMHazardRecognizer(const InstrItineraryData *ItinData,
                       const ScheduleDAG *DAG)
     : ScoreboardHazardRecognizer(ItinData, DAG, "post-RA-sched"),
-      LastMI(0) {}
+      LastMI(nullptr) {}
 
   HazardType getHazardType(SUnit *SU, int Stalls) override;
   void Reset() override;
diff --git a/lib/Target/ARM/ARMISelDAGToDAG.cpp b/lib/Target/ARM/ARMISelDAGToDAG.cpp
index 70e11c5..08d598d 100644
--- a/lib/Target/ARM/ARMISelDAGToDAG.cpp
+++ b/lib/Target/ARM/ARMISelDAGToDAG.cpp
@@ -11,7 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-isel"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMTargetMachine.h"
@@ -37,6 +36,8 @@
 
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-isel"
+
 static cl::opt<bool>
 DisableShifterOp("disable-shifter-op", cl::Hidden,
   cl::desc("Disable isel of shifter-op"),
@@ -72,6 +73,13 @@ public:
       Subtarget(&TM.getSubtarget<ARMSubtarget>()) {
   }
 
+  bool runOnMachineFunction(MachineFunction &MF) override {
+    // Reset the subtarget each time through.
+    Subtarget = &TM.getSubtarget<ARMSubtarget>();
+    SelectionDAGISel::runOnMachineFunction(MF);
+    return true;
+  }
+
   const char *getPassName() const override {
     return "ARM Instruction Selection";
   }
@@ -397,7 +405,7 @@ void ARMDAGToDAGISel::PreprocessISelDAG() {
     N1 = CurDAG->getNode(ISD::SHL, SDLoc(N1), MVT::i32,
                          N1, CurDAG->getConstant(TZ, MVT::i32));
     CurDAG->UpdateNodeOperands(N, N0, N1);
-  }  
+  }
 }
 
 /// hasNoVMLxHazardUse - Return true if it's desirable to select a FP MLA / MLS
@@ -1440,7 +1448,7 @@ SDNode *ARMDAGToDAGISel::SelectARMIndexedLoad(SDNode *N) {
   LoadSDNode *LD = cast<LoadSDNode>(N);
   ISD::MemIndexedMode AM = LD->getAddressingMode();
   if (AM == ISD::UNINDEXED)
-    return NULL;
+    return nullptr;
 
   EVT LoadedVT = LD->getMemoryVT();
   SDValue Offset, AMOpc;
@@ -1506,14 +1514,14 @@ SDNode *ARMDAGToDAGISel::SelectARMIndexedLoad(SDNode *N) {
     }
   }
 
-  return NULL;
+  return nullptr;
 }
 
 SDNode *ARMDAGToDAGISel::SelectT2IndexedLoad(SDNode *N) {
   LoadSDNode *LD = cast<LoadSDNode>(N);
   ISD::MemIndexedMode AM = LD->getAddressingMode();
   if (AM == ISD::UNINDEXED)
-    return NULL;
+    return nullptr;
 
   EVT LoadedVT = LD->getMemoryVT();
   bool isSExtLd = LD->getExtensionType() == ISD::SEXTLOAD;
@@ -1540,7 +1548,7 @@ SDNode *ARMDAGToDAGISel::SelectT2IndexedLoad(SDNode *N) {
         Opcode = isPre ? ARM::t2LDRB_PRE : ARM::t2LDRB_POST;
       break;
     default:
-      return NULL;
+      return nullptr;
     }
     Match = true;
   }
@@ -1554,7 +1562,7 @@ SDNode *ARMDAGToDAGISel::SelectT2IndexedLoad(SDNode *N) {
                                   MVT::Other, Ops);
   }
 
-  return NULL;
+  return nullptr;
 }
 
 /// \brief Form a GPRPair pseudo register from a pair of GPR regs.
@@ -1699,10 +1707,10 @@ static bool isVSTfixed(unsigned Opc)
   case ARM::VST1d16wb_fixed : return true;
   case ARM::VST1d32wb_fixed : return true;
   case ARM::VST1d64wb_fixed : return true;
-  case ARM::VST1q8wb_fixed : return true; 
-  case ARM::VST1q16wb_fixed : return true; 
-  case ARM::VST1q32wb_fixed : return true; 
-  case ARM::VST1q64wb_fixed : return true; 
+  case ARM::VST1q8wb_fixed : return true;
+  case ARM::VST1q16wb_fixed : return true;
+  case ARM::VST1q32wb_fixed : return true;
+  case ARM::VST1q64wb_fixed : return true;
   case ARM::VST1d64TPseudoWB_fixed : return true;
   case ARM::VST1d64QPseudoWB_fixed : return true;
   case ARM::VST2d8wb_fixed : return true;
@@ -1776,7 +1784,7 @@ SDNode *ARMDAGToDAGISel::SelectVLD(SDNode *N, bool isUpdating, unsigned NumVecs,
   SDValue MemAddr, Align;
   unsigned AddrOpIdx = isUpdating ? 1 : 2;
   if (!SelectAddrMode6(N, N->getOperand(AddrOpIdx), MemAddr, Align))
-    return NULL;
+    return nullptr;
 
   SDValue Chain = N->getOperand(0);
   EVT VT = N->getValueType(0);
@@ -1895,7 +1903,7 @@ SDNode *ARMDAGToDAGISel::SelectVLD(SDNode *N, bool isUpdating, unsigned NumVecs,
   ReplaceUses(SDValue(N, NumVecs), SDValue(VLd, 1));
   if (isUpdating)
     ReplaceUses(SDValue(N, NumVecs + 1), SDValue(VLd, 2));
-  return NULL;
+  return nullptr;
 }
 
 SDNode *ARMDAGToDAGISel::SelectVST(SDNode *N, bool isUpdating, unsigned NumVecs,
@@ -1909,7 +1917,7 @@ SDNode *ARMDAGToDAGISel::SelectVST(SDNode *N, bool isUpdating, unsigned NumVecs,
   unsigned AddrOpIdx = isUpdating ? 1 : 2;
   unsigned Vec0Idx = 3; // AddrOpIdx + (isUpdating ? 2 : 1)
   if (!SelectAddrMode6(N, N->getOperand(AddrOpIdx), MemAddr, Align))
-    return NULL;
+    return nullptr;
 
   MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
   MemOp[0] = cast<MemIntrinsicSDNode>(N)->getMemOperand();
@@ -2055,7 +2063,7 @@ SDNode *ARMDAGToDAGISel::SelectVLDSTLane(SDNode *N, bool IsLoad,
   unsigned AddrOpIdx = isUpdating ? 1 : 2;
   unsigned Vec0Idx = 3; // AddrOpIdx + (isUpdating ? 2 : 1)
   if (!SelectAddrMode6(N, N->getOperand(AddrOpIdx), MemAddr, Align))
-    return NULL;
+    return nullptr;
 
   MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
   MemOp[0] = cast<MemIntrinsicSDNode>(N)->getMemOperand();
@@ -2160,7 +2168,7 @@ SDNode *ARMDAGToDAGISel::SelectVLDSTLane(SDNode *N, bool IsLoad,
   ReplaceUses(SDValue(N, NumVecs), SDValue(VLdLn, 1));
   if (isUpdating)
     ReplaceUses(SDValue(N, NumVecs + 1), SDValue(VLdLn, 2));
-  return NULL;
+  return nullptr;
 }
 
 SDNode *ARMDAGToDAGISel::SelectVLDDup(SDNode *N, bool isUpdating,
@@ -2171,7 +2179,7 @@ SDNode *ARMDAGToDAGISel::SelectVLDDup(SDNode *N, bool isUpdating,
 
   SDValue MemAddr, Align;
   if (!SelectAddrMode6(N, N->getOperand(1), MemAddr, Align))
-    return NULL;
+    return nullptr;
 
   MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
   MemOp[0] = cast<MemIntrinsicSDNode>(N)->getMemOperand();
@@ -2243,7 +2251,7 @@ SDNode *ARMDAGToDAGISel::SelectVLDDup(SDNode *N, bool isUpdating,
   ReplaceUses(SDValue(N, NumVecs), SDValue(VLdDup, 1));
   if (isUpdating)
     ReplaceUses(SDValue(N, NumVecs + 1), SDValue(VLdDup, 2));
-  return NULL;
+  return nullptr;
 }
 
 SDNode *ARMDAGToDAGISel::SelectVTBL(SDNode *N, bool IsExt, unsigned NumVecs,
@@ -2282,7 +2290,7 @@ SDNode *ARMDAGToDAGISel::SelectVTBL(SDNode *N, bool IsExt, unsigned NumVecs,
 SDNode *ARMDAGToDAGISel::SelectV6T2BitfieldExtractOp(SDNode *N,
                                                      bool isSigned) {
   if (!Subtarget->hasV6T2Ops())
-    return NULL;
+    return nullptr;
 
   unsigned Opc = isSigned
     ? (Subtarget->isThumb() ? ARM::t2SBFX : ARM::SBFX)
@@ -2295,7 +2303,7 @@ SDNode *ARMDAGToDAGISel::SelectV6T2BitfieldExtractOp(SDNode *N,
 
       // The immediate is a mask of the low bits iff imm & (imm+1) == 0
       if (And_imm & (And_imm + 1))
-        return NULL;
+        return nullptr;
 
       unsigned Srl_imm = 0;
       if (isOpcWithIntImmediate(N->getOperand(0).getNode(), ISD::SRL,
@@ -2315,7 +2323,7 @@ SDNode *ARMDAGToDAGISel::SelectV6T2BitfieldExtractOp(SDNode *N,
             SDValue Ops[] = { N->getOperand(0).getOperand(0),
                               CurDAG->getTargetConstant(LSB, MVT::i32),
                               getAL(CurDAG), Reg0, Reg0 };
-            return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops, 5);
+            return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops);
           }
 
           // ARM models shift instructions as MOVsi with shifter operand.
@@ -2325,17 +2333,17 @@ SDNode *ARMDAGToDAGISel::SelectV6T2BitfieldExtractOp(SDNode *N,
                                       MVT::i32);
           SDValue Ops[] = { N->getOperand(0).getOperand(0), ShOpc,
                             getAL(CurDAG), Reg0, Reg0 };
-          return CurDAG->SelectNodeTo(N, ARM::MOVsi, MVT::i32, Ops, 5);
+          return CurDAG->SelectNodeTo(N, ARM::MOVsi, MVT::i32, Ops);
         }
 
         SDValue Ops[] = { N->getOperand(0).getOperand(0),
                           CurDAG->getTargetConstant(LSB, MVT::i32),
                           CurDAG->getTargetConstant(Width, MVT::i32),
-          getAL(CurDAG), Reg0 };
-        return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops, 5);
+                          getAL(CurDAG), Reg0 };
+        return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops);
       }
     }
-    return NULL;
+    return nullptr;
   }
 
   // Otherwise, we're looking for a shift of a shift
@@ -2349,16 +2357,16 @@ SDNode *ARMDAGToDAGISel::SelectV6T2BitfieldExtractOp(SDNode *N,
       unsigned Width = 32 - Srl_imm - 1;
       int LSB = Srl_imm - Shl_imm;
       if (LSB < 0)
-        return NULL;
+        return nullptr;
       SDValue Reg0 = CurDAG->getRegister(0, MVT::i32);
       SDValue Ops[] = { N->getOperand(0).getOperand(0),
                         CurDAG->getTargetConstant(LSB, MVT::i32),
                         CurDAG->getTargetConstant(Width, MVT::i32),
                         getAL(CurDAG), Reg0 };
-      return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops, 5);
+      return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops);
     }
   }
-  return NULL;
+  return nullptr;
 }
 
 /// Target-specific DAG combining for ISD::XOR.
@@ -2377,10 +2385,10 @@ SDNode *ARMDAGToDAGISel::SelectABSOp(SDNode *N){
   EVT VT = N->getValueType(0);
 
   if (Subtarget->isThumb1Only())
-    return NULL;
+    return nullptr;
 
   if (XORSrc0.getOpcode() != ISD::ADD || XORSrc1.getOpcode() != ISD::SRA)
-    return NULL;
+    return nullptr;
 
   SDValue ADDSrc0 = XORSrc0.getOperand(0);
   SDValue ADDSrc1 = XORSrc0.getOperand(1);
@@ -2391,13 +2399,13 @@ SDNode *ARMDAGToDAGISel::SelectABSOp(SDNode *N){
   unsigned Size = XType.getSizeInBits() - 1;
 
   if (ADDSrc1 == XORSrc1 && ADDSrc0 == SRASrc0 &&
-      XType.isInteger() && SRAConstant != NULL &&
+      XType.isInteger() && SRAConstant != nullptr &&
       Size == SRAConstant->getZExtValue()) {
     unsigned Opcode = Subtarget->isThumb2() ? ARM::t2ABS : ARM::ABS;
     return CurDAG->SelectNodeTo(N, Opcode, VT, ADDSrc0);
   }
 
-  return NULL;
+  return nullptr;
 }
 
 SDNode *ARMDAGToDAGISel::SelectConcatVector(SDNode *N) {
@@ -2414,7 +2422,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
 
   if (N->isMachineOpcode()) {
     N->setNodeId(-1);
-    return NULL;   // Already selected.
+    return nullptr;   // Already selected.
   }
 
   switch (N->getOpcode()) {
@@ -2478,7 +2486,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                        Ops);
       }
       ReplaceUses(SDValue(N, 0), SDValue(ResNode, 0));
-      return NULL;
+      return nullptr;
     }
 
     // Other cases are autogenerated.
@@ -2492,14 +2500,14 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     if (Subtarget->isThumb1Only()) {
       SDValue Ops[] = { TFI, CurDAG->getTargetConstant(0, MVT::i32),
                         getAL(CurDAG), CurDAG->getRegister(0, MVT::i32) };
-      return CurDAG->SelectNodeTo(N, ARM::tADDrSPi, MVT::i32, Ops, 4);
+      return CurDAG->SelectNodeTo(N, ARM::tADDrSPi, MVT::i32, Ops);
     } else {
       unsigned Opc = ((Subtarget->isThumb() && Subtarget->hasThumb2()) ?
                       ARM::t2ADDri : ARM::ADDri);
       SDValue Ops[] = { TFI, CurDAG->getTargetConstant(0, MVT::i32),
                         getAL(CurDAG), CurDAG->getRegister(0, MVT::i32),
                         CurDAG->getRegister(0, MVT::i32) };
-      return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops, 5);
+      return CurDAG->SelectNodeTo(N, Opc, MVT::i32, Ops);
     }
   }
   case ISD::SRL:
@@ -2526,10 +2534,10 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
         SDValue Reg0 = CurDAG->getRegister(0, MVT::i32);
         if (Subtarget->isThumb()) {
           SDValue Ops[] = { V, V, ShImmOp, getAL(CurDAG), Reg0, Reg0 };
-          return CurDAG->SelectNodeTo(N, ARM::t2ADDrs, MVT::i32, Ops, 6);
+          return CurDAG->SelectNodeTo(N, ARM::t2ADDrs, MVT::i32, Ops);
         } else {
           SDValue Ops[] = { V, V, Reg0, ShImmOp, getAL(CurDAG), Reg0, Reg0 };
-          return CurDAG->SelectNodeTo(N, ARM::ADDrsi, MVT::i32, Ops, 7);
+          return CurDAG->SelectNodeTo(N, ARM::ADDrsi, MVT::i32, Ops);
         }
       }
       if (isPowerOf2_32(RHSV+1)) {  // 2^n-1?
@@ -2542,10 +2550,10 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
         SDValue Reg0 = CurDAG->getRegister(0, MVT::i32);
         if (Subtarget->isThumb()) {
           SDValue Ops[] = { V, V, ShImmOp, getAL(CurDAG), Reg0, Reg0 };
-          return CurDAG->SelectNodeTo(N, ARM::t2RSBrs, MVT::i32, Ops, 6);
+          return CurDAG->SelectNodeTo(N, ARM::t2RSBrs, MVT::i32, Ops);
         } else {
           SDValue Ops[] = { V, V, Reg0, ShImmOp, getAL(CurDAG), Reg0, Reg0 };
-          return CurDAG->SelectNodeTo(N, ARM::RSBrsi, MVT::i32, Ops, 7);
+          return CurDAG->SelectNodeTo(N, ARM::RSBrsi, MVT::i32, Ops);
         }
       }
     }
@@ -2660,7 +2668,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     }
   }
   case ISD::LOAD: {
-    SDNode *ResNode = 0;
+    SDNode *ResNode = nullptr;
     if (Subtarget->isThumb() && Subtarget->hasThumb2())
       ResNode = SelectT2IndexedLoad(N);
     else
@@ -2707,13 +2715,13 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     }
     ReplaceUses(SDValue(N, 0),
                 SDValue(Chain.getNode(), Chain.getResNo()));
-    return NULL;
+    return nullptr;
   }
   case ARMISD::VZIP: {
     unsigned Opc = 0;
     EVT VT = N->getValueType(0);
     switch (VT.getSimpleVT().SimpleTy) {
-    default: return NULL;
+    default: return nullptr;
     case MVT::v8i8:  Opc = ARM::VZIPd8; break;
     case MVT::v4i16: Opc = ARM::VZIPd16; break;
     case MVT::v2f32:
@@ -2733,7 +2741,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     unsigned Opc = 0;
     EVT VT = N->getValueType(0);
     switch (VT.getSimpleVT().SimpleTy) {
-    default: return NULL;
+    default: return nullptr;
     case MVT::v8i8:  Opc = ARM::VUZPd8; break;
     case MVT::v4i16: Opc = ARM::VUZPd16; break;
     case MVT::v2f32:
@@ -2753,7 +2761,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     unsigned Opc = 0;
     EVT VT = N->getValueType(0);
     switch (VT.getSimpleVT().SimpleTy) {
-    default: return NULL;
+    default: return nullptr;
     case MVT::v8i8:  Opc = ARM::VTRNd8; break;
     case MVT::v4i16: Opc = ARM::VTRNd16; break;
     case MVT::v2f32:
@@ -2834,7 +2842,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                          ARM::VLD1q16wb_fixed,
                                          ARM::VLD1q32wb_fixed,
                                          ARM::VLD1q64wb_fixed };
-    return SelectVLD(N, true, 1, DOpcodes, QOpcodes, 0);
+    return SelectVLD(N, true, 1, DOpcodes, QOpcodes, nullptr);
   }
 
   case ARMISD::VLD2_UPD: {
@@ -2845,7 +2853,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     static const uint16_t QOpcodes[] = { ARM::VLD2q8PseudoWB_fixed,
                                          ARM::VLD2q16PseudoWB_fixed,
                                          ARM::VLD2q32PseudoWB_fixed };
-    return SelectVLD(N, true, 2, DOpcodes, QOpcodes, 0);
+    return SelectVLD(N, true, 2, DOpcodes, QOpcodes, nullptr);
   }
 
   case ARMISD::VLD3_UPD: {
@@ -2912,7 +2920,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                          ARM::VST1q16wb_fixed,
                                          ARM::VST1q32wb_fixed,
                                          ARM::VST1q64wb_fixed };
-    return SelectVST(N, true, 1, DOpcodes, QOpcodes, 0);
+    return SelectVST(N, true, 1, DOpcodes, QOpcodes, nullptr);
   }
 
   case ARMISD::VST2_UPD: {
@@ -2923,7 +2931,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
     static const uint16_t QOpcodes[] = { ARM::VST2q8PseudoWB_fixed,
                                          ARM::VST2q16PseudoWB_fixed,
                                          ARM::VST2q32PseudoWB_fixed };
-    return SelectVST(N, true, 2, DOpcodes, QOpcodes, 0);
+    return SelectVST(N, true, 2, DOpcodes, QOpcodes, nullptr);
   }
 
   case ARMISD::VST3_UPD: {
@@ -3047,7 +3055,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
         ReplaceUses(SDValue(N, 1), Result);
       }
       ReplaceUses(SDValue(N, 2), OutChain);
-      return NULL;
+      return nullptr;
     }
     case Intrinsic::arm_stlexd:
     case Intrinsic::arm_strexd: {
@@ -3093,7 +3101,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                            ARM::VLD1d32, ARM::VLD1d64 };
       static const uint16_t QOpcodes[] = { ARM::VLD1q8, ARM::VLD1q16,
                                            ARM::VLD1q32, ARM::VLD1q64};
-      return SelectVLD(N, false, 1, DOpcodes, QOpcodes, 0);
+      return SelectVLD(N, false, 1, DOpcodes, QOpcodes, nullptr);
     }
 
     case Intrinsic::arm_neon_vld2: {
@@ -3101,7 +3109,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                            ARM::VLD2d32, ARM::VLD1q64 };
       static const uint16_t QOpcodes[] = { ARM::VLD2q8Pseudo, ARM::VLD2q16Pseudo,
                                            ARM::VLD2q32Pseudo };
-      return SelectVLD(N, false, 2, DOpcodes, QOpcodes, 0);
+      return SelectVLD(N, false, 2, DOpcodes, QOpcodes, nullptr);
     }
 
     case Intrinsic::arm_neon_vld3: {
@@ -3164,7 +3172,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                            ARM::VST1d32, ARM::VST1d64 };
       static const uint16_t QOpcodes[] = { ARM::VST1q8, ARM::VST1q16,
                                            ARM::VST1q32, ARM::VST1q64 };
-      return SelectVST(N, false, 1, DOpcodes, QOpcodes, 0);
+      return SelectVST(N, false, 1, DOpcodes, QOpcodes, nullptr);
     }
 
     case Intrinsic::arm_neon_vst2: {
@@ -3172,7 +3180,7 @@ SDNode *ARMDAGToDAGISel::Select(SDNode *N) {
                                            ARM::VST2d32, ARM::VST1q64 };
       static uint16_t QOpcodes[] = { ARM::VST2q8Pseudo, ARM::VST2q16Pseudo,
                                      ARM::VST2q32Pseudo };
-      return SelectVST(N, false, 2, DOpcodes, QOpcodes, 0);
+      return SelectVST(N, false, 2, DOpcodes, QOpcodes, nullptr);
     }
 
     case Intrinsic::arm_neon_vst3: {
@@ -3306,7 +3314,8 @@ SDNode *ARMDAGToDAGISel::SelectInlineAsm(SDNode *N){
   // them into a GPRPair.
 
   SDLoc dl(N);
-  SDValue Glue = N->getGluedNode() ? N->getOperand(NumOps-1) : SDValue(0,0);
+  SDValue Glue = N->getGluedNode() ? N->getOperand(NumOps-1)
+                                   : SDValue(nullptr,0);
 
   SmallVector<bool, 8> OpChanged;
   // Glue node will be appended late.
@@ -3388,7 +3397,7 @@ SDNode *ARMDAGToDAGISel::SelectInlineAsm(SDNode *N){
       // Update the original glue user.
       std::vector<SDValue> Ops(GU->op_begin(), GU->op_end()-1);
       Ops.push_back(T1.getValue(1));
-      CurDAG->UpdateNodeOperands(GU, &Ops[0], Ops.size());
+      CurDAG->UpdateNodeOperands(GU, Ops);
       GU = T1.getNode();
     }
     else {
@@ -3435,11 +3444,10 @@ SDNode *ARMDAGToDAGISel::SelectInlineAsm(SDNode *N){
   if (Glue.getNode())
     AsmNodeOperands.push_back(Glue);
   if (!Changed)
-    return NULL;
+    return nullptr;
 
   SDValue New = CurDAG->getNode(ISD::INLINEASM, SDLoc(N),
-      CurDAG->getVTList(MVT::Other, MVT::Glue), &AsmNodeOperands[0],
-                        AsmNodeOperands.size());
+      CurDAG->getVTList(MVT::Other, MVT::Glue), AsmNodeOperands);
   New->setNodeId(-1);
   return New.getNode();
 }
diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index 2ebad8e..00d07e8 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -12,7 +12,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-isel"
 #include "ARMISelLowering.h"
 #include "ARMCallingConv.h"
 #include "ARMConstantPoolValue.h"
@@ -37,18 +36,22 @@
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/Intrinsics.h"
 #include "llvm/IR/Type.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Target/TargetOptions.h"
 #include <utility>
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-isel"
+
 STATISTIC(NumTailCalls, "Number of tail calls");
 STATISTIC(NumMovwMovt, "Number of GAs materialized with movw + movt");
 STATISTIC(NumLoopByVals, "Number of loops generated for byval arguments");
@@ -79,7 +82,7 @@ namespace {
 }
 
 // The APCS parameter registers.
-static const uint16_t GPRArgRegs[] = {
+static const MCPhysReg GPRArgRegs[] = {
   ARM::R0, ARM::R1, ARM::R2, ARM::R3
 };
 
@@ -155,7 +158,8 @@ void ARMTargetLowering::addQRTypeForNEON(MVT VT) {
 static TargetLoweringObjectFile *createTLOF(TargetMachine &TM) {
   if (TM.getSubtarget<ARMSubtarget>().isTargetMachO())
     return new TargetLoweringObjectFileMachO();
-
+  if (TM.getSubtarget<ARMSubtarget>().isTargetWindows())
+    return new TargetLoweringObjectFileCOFF();
   return new ARMElfTargetObjectFile();
 }
 
@@ -170,7 +174,7 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
   if (Subtarget->isTargetMachO()) {
     // Uses VFP for Thumb libfuncs if available.
     if (Subtarget->isThumb() && Subtarget->hasVFP2() &&
-        Subtarget->hasARMOps()) {
+        Subtarget->hasARMOps() && !TM.Options.UseSoftFloat) {
       // Single-precision floating-point arithmetic.
       setLibcallName(RTLIB::ADD_F32, "__addsf3vfp");
       setLibcallName(RTLIB::SUB_F32, "__subsf3vfp");
@@ -246,173 +250,134 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
   }
 
   // These libcalls are not available in 32-bit.
-  setLibcallName(RTLIB::SHL_I128, 0);
-  setLibcallName(RTLIB::SRL_I128, 0);
-  setLibcallName(RTLIB::SRA_I128, 0);
+  setLibcallName(RTLIB::SHL_I128, nullptr);
+  setLibcallName(RTLIB::SRL_I128, nullptr);
+  setLibcallName(RTLIB::SRA_I128, nullptr);
 
   if (Subtarget->isAAPCS_ABI() && !Subtarget->isTargetMachO() &&
       !Subtarget->isTargetWindows()) {
-    // Double-precision floating-point arithmetic helper functions
-    // RTABI chapter 4.1.2, Table 2
-    setLibcallName(RTLIB::ADD_F64, "__aeabi_dadd");
-    setLibcallName(RTLIB::DIV_F64, "__aeabi_ddiv");
-    setLibcallName(RTLIB::MUL_F64, "__aeabi_dmul");
-    setLibcallName(RTLIB::SUB_F64, "__aeabi_dsub");
-    setLibcallCallingConv(RTLIB::ADD_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::DIV_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::MUL_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SUB_F64, CallingConv::ARM_AAPCS);
-
-    // Double-precision floating-point comparison helper functions
-    // RTABI chapter 4.1.2, Table 3
-    setLibcallName(RTLIB::OEQ_F64, "__aeabi_dcmpeq");
-    setCmpLibcallCC(RTLIB::OEQ_F64, ISD::SETNE);
-    setLibcallName(RTLIB::UNE_F64, "__aeabi_dcmpeq");
-    setCmpLibcallCC(RTLIB::UNE_F64, ISD::SETEQ);
-    setLibcallName(RTLIB::OLT_F64, "__aeabi_dcmplt");
-    setCmpLibcallCC(RTLIB::OLT_F64, ISD::SETNE);
-    setLibcallName(RTLIB::OLE_F64, "__aeabi_dcmple");
-    setCmpLibcallCC(RTLIB::OLE_F64, ISD::SETNE);
-    setLibcallName(RTLIB::OGE_F64, "__aeabi_dcmpge");
-    setCmpLibcallCC(RTLIB::OGE_F64, ISD::SETNE);
-    setLibcallName(RTLIB::OGT_F64, "__aeabi_dcmpgt");
-    setCmpLibcallCC(RTLIB::OGT_F64, ISD::SETNE);
-    setLibcallName(RTLIB::UO_F64,  "__aeabi_dcmpun");
-    setCmpLibcallCC(RTLIB::UO_F64,  ISD::SETNE);
-    setLibcallName(RTLIB::O_F64,   "__aeabi_dcmpun");
-    setCmpLibcallCC(RTLIB::O_F64,   ISD::SETEQ);
-    setLibcallCallingConv(RTLIB::OEQ_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UNE_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OLT_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OLE_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OGE_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OGT_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UO_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::O_F64, CallingConv::ARM_AAPCS);
-
-    // Single-precision floating-point arithmetic helper functions
-    // RTABI chapter 4.1.2, Table 4
-    setLibcallName(RTLIB::ADD_F32, "__aeabi_fadd");
-    setLibcallName(RTLIB::DIV_F32, "__aeabi_fdiv");
-    setLibcallName(RTLIB::MUL_F32, "__aeabi_fmul");
-    setLibcallName(RTLIB::SUB_F32, "__aeabi_fsub");
-    setLibcallCallingConv(RTLIB::ADD_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::DIV_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::MUL_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SUB_F32, CallingConv::ARM_AAPCS);
-
-    // Single-precision floating-point comparison helper functions
-    // RTABI chapter 4.1.2, Table 5
-    setLibcallName(RTLIB::OEQ_F32, "__aeabi_fcmpeq");
-    setCmpLibcallCC(RTLIB::OEQ_F32, ISD::SETNE);
-    setLibcallName(RTLIB::UNE_F32, "__aeabi_fcmpeq");
-    setCmpLibcallCC(RTLIB::UNE_F32, ISD::SETEQ);
-    setLibcallName(RTLIB::OLT_F32, "__aeabi_fcmplt");
-    setCmpLibcallCC(RTLIB::OLT_F32, ISD::SETNE);
-    setLibcallName(RTLIB::OLE_F32, "__aeabi_fcmple");
-    setCmpLibcallCC(RTLIB::OLE_F32, ISD::SETNE);
-    setLibcallName(RTLIB::OGE_F32, "__aeabi_fcmpge");
-    setCmpLibcallCC(RTLIB::OGE_F32, ISD::SETNE);
-    setLibcallName(RTLIB::OGT_F32, "__aeabi_fcmpgt");
-    setCmpLibcallCC(RTLIB::OGT_F32, ISD::SETNE);
-    setLibcallName(RTLIB::UO_F32,  "__aeabi_fcmpun");
-    setCmpLibcallCC(RTLIB::UO_F32,  ISD::SETNE);
-    setLibcallName(RTLIB::O_F32,   "__aeabi_fcmpun");
-    setCmpLibcallCC(RTLIB::O_F32,   ISD::SETEQ);
-    setLibcallCallingConv(RTLIB::OEQ_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UNE_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OLT_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OLE_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OGE_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::OGT_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UO_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::O_F32, CallingConv::ARM_AAPCS);
-
-    // Floating-point to integer conversions.
-    // RTABI chapter 4.1.2, Table 6
-    setLibcallName(RTLIB::FPTOSINT_F64_I32, "__aeabi_d2iz");
-    setLibcallName(RTLIB::FPTOUINT_F64_I32, "__aeabi_d2uiz");
-    setLibcallName(RTLIB::FPTOSINT_F64_I64, "__aeabi_d2lz");
-    setLibcallName(RTLIB::FPTOUINT_F64_I64, "__aeabi_d2ulz");
-    setLibcallName(RTLIB::FPTOSINT_F32_I32, "__aeabi_f2iz");
-    setLibcallName(RTLIB::FPTOUINT_F32_I32, "__aeabi_f2uiz");
-    setLibcallName(RTLIB::FPTOSINT_F32_I64, "__aeabi_f2lz");
-    setLibcallName(RTLIB::FPTOUINT_F32_I64, "__aeabi_f2ulz");
-    setLibcallCallingConv(RTLIB::FPTOSINT_F64_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOUINT_F64_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOSINT_F64_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOUINT_F64_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOSINT_F32_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOUINT_F32_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOSINT_F32_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPTOUINT_F32_I64, CallingConv::ARM_AAPCS);
-
-    // Conversions between floating types.
-    // RTABI chapter 4.1.2, Table 7
-    setLibcallName(RTLIB::FPROUND_F64_F32, "__aeabi_d2f");
-    setLibcallName(RTLIB::FPEXT_F32_F64,   "__aeabi_f2d");
-    setLibcallCallingConv(RTLIB::FPROUND_F64_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::FPEXT_F32_F64, CallingConv::ARM_AAPCS);
-
-    // Integer to floating-point conversions.
-    // RTABI chapter 4.1.2, Table 8
-    setLibcallName(RTLIB::SINTTOFP_I32_F64, "__aeabi_i2d");
-    setLibcallName(RTLIB::UINTTOFP_I32_F64, "__aeabi_ui2d");
-    setLibcallName(RTLIB::SINTTOFP_I64_F64, "__aeabi_l2d");
-    setLibcallName(RTLIB::UINTTOFP_I64_F64, "__aeabi_ul2d");
-    setLibcallName(RTLIB::SINTTOFP_I32_F32, "__aeabi_i2f");
-    setLibcallName(RTLIB::UINTTOFP_I32_F32, "__aeabi_ui2f");
-    setLibcallName(RTLIB::SINTTOFP_I64_F32, "__aeabi_l2f");
-    setLibcallName(RTLIB::UINTTOFP_I64_F32, "__aeabi_ul2f");
-    setLibcallCallingConv(RTLIB::SINTTOFP_I32_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UINTTOFP_I32_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SINTTOFP_I64_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UINTTOFP_I64_F64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SINTTOFP_I32_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UINTTOFP_I32_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SINTTOFP_I64_F32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UINTTOFP_I64_F32, CallingConv::ARM_AAPCS);
-
-    // Long long helper functions
-    // RTABI chapter 4.2, Table 9
-    setLibcallName(RTLIB::MUL_I64,  "__aeabi_lmul");
-    setLibcallName(RTLIB::SHL_I64, "__aeabi_llsl");
-    setLibcallName(RTLIB::SRL_I64, "__aeabi_llsr");
-    setLibcallName(RTLIB::SRA_I64, "__aeabi_lasr");
-    setLibcallCallingConv(RTLIB::MUL_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SDIV_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UDIV_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SHL_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SRL_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SRA_I64, CallingConv::ARM_AAPCS);
-
-    // Integer division functions
-    // RTABI chapter 4.3.1
-    setLibcallName(RTLIB::SDIV_I8,  "__aeabi_idiv");
-    setLibcallName(RTLIB::SDIV_I16, "__aeabi_idiv");
-    setLibcallName(RTLIB::SDIV_I32, "__aeabi_idiv");
-    setLibcallName(RTLIB::SDIV_I64, "__aeabi_ldivmod");
-    setLibcallName(RTLIB::UDIV_I8,  "__aeabi_uidiv");
-    setLibcallName(RTLIB::UDIV_I16, "__aeabi_uidiv");
-    setLibcallName(RTLIB::UDIV_I32, "__aeabi_uidiv");
-    setLibcallName(RTLIB::UDIV_I64, "__aeabi_uldivmod");
-    setLibcallCallingConv(RTLIB::SDIV_I8, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SDIV_I16, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SDIV_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::SDIV_I64, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UDIV_I8, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UDIV_I16, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UDIV_I32, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::UDIV_I64, CallingConv::ARM_AAPCS);
-
-    // Memory operations
-    // RTABI chapter 4.3.4
-    setLibcallName(RTLIB::MEMCPY,  "__aeabi_memcpy");
-    setLibcallName(RTLIB::MEMMOVE, "__aeabi_memmove");
-    setLibcallName(RTLIB::MEMSET,  "__aeabi_memset");
-    setLibcallCallingConv(RTLIB::MEMCPY, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::MEMMOVE, CallingConv::ARM_AAPCS);
-    setLibcallCallingConv(RTLIB::MEMSET, CallingConv::ARM_AAPCS);
+    static const struct {
+      const RTLIB::Libcall Op;
+      const char * const Name;
+      const CallingConv::ID CC;
+      const ISD::CondCode Cond;
+    } LibraryCalls[] = {
+      // Double-precision floating-point arithmetic helper functions
+      // RTABI chapter 4.1.2, Table 2
+      { RTLIB::ADD_F64, "__aeabi_dadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::DIV_F64, "__aeabi_ddiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::MUL_F64, "__aeabi_dmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SUB_F64, "__aeabi_dsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Double-precision floating-point comparison helper functions
+      // RTABI chapter 4.1.2, Table 3
+      { RTLIB::OEQ_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::UNE_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
+      { RTLIB::OLT_F64, "__aeabi_dcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OLE_F64, "__aeabi_dcmple", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OGE_F64, "__aeabi_dcmpge", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OGT_F64, "__aeabi_dcmpgt", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::UO_F64,  "__aeabi_dcmpun", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::O_F64,   "__aeabi_dcmpun", CallingConv::ARM_AAPCS, ISD::SETEQ },
+
+      // Single-precision floating-point arithmetic helper functions
+      // RTABI chapter 4.1.2, Table 4
+      { RTLIB::ADD_F32, "__aeabi_fadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::DIV_F32, "__aeabi_fdiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::MUL_F32, "__aeabi_fmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SUB_F32, "__aeabi_fsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Single-precision floating-point comparison helper functions
+      // RTABI chapter 4.1.2, Table 5
+      { RTLIB::OEQ_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::UNE_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
+      { RTLIB::OLT_F32, "__aeabi_fcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OLE_F32, "__aeabi_fcmple", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OGE_F32, "__aeabi_fcmpge", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::OGT_F32, "__aeabi_fcmpgt", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::UO_F32,  "__aeabi_fcmpun", CallingConv::ARM_AAPCS, ISD::SETNE },
+      { RTLIB::O_F32,   "__aeabi_fcmpun", CallingConv::ARM_AAPCS, ISD::SETEQ },
+
+      // Floating-point to integer conversions.
+      // RTABI chapter 4.1.2, Table 6
+      { RTLIB::FPTOSINT_F64_I32, "__aeabi_d2iz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOUINT_F64_I32, "__aeabi_d2uiz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOSINT_F64_I64, "__aeabi_d2lz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOUINT_F64_I64, "__aeabi_d2ulz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOSINT_F32_I32, "__aeabi_f2iz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOUINT_F32_I32, "__aeabi_f2uiz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOSINT_F32_I64, "__aeabi_f2lz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPTOUINT_F32_I64, "__aeabi_f2ulz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Conversions between floating types.
+      // RTABI chapter 4.1.2, Table 7
+      { RTLIB::FPROUND_F64_F32, "__aeabi_d2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::FPEXT_F32_F64,   "__aeabi_f2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Integer to floating-point conversions.
+      // RTABI chapter 4.1.2, Table 8
+      { RTLIB::SINTTOFP_I32_F64, "__aeabi_i2d",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UINTTOFP_I32_F64, "__aeabi_ui2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SINTTOFP_I64_F64, "__aeabi_l2d",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UINTTOFP_I64_F64, "__aeabi_ul2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SINTTOFP_I32_F32, "__aeabi_i2f",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UINTTOFP_I32_F32, "__aeabi_ui2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SINTTOFP_I64_F32, "__aeabi_l2f",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UINTTOFP_I64_F32, "__aeabi_ul2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Long long helper functions
+      // RTABI chapter 4.2, Table 9
+      { RTLIB::MUL_I64, "__aeabi_lmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SHL_I64, "__aeabi_llsl", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SRL_I64, "__aeabi_llsr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SRA_I64, "__aeabi_lasr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Integer division functions
+      // RTABI chapter 4.3.1
+      { RTLIB::SDIV_I8,  "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SDIV_I16, "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SDIV_I32, "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::SDIV_I64, "__aeabi_ldivmod",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UDIV_I8,  "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UDIV_I16, "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UDIV_I32, "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::UDIV_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+
+      // Memory operations
+      // RTABI chapter 4.3.4
+      { RTLIB::MEMCPY,  "__aeabi_memcpy",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::MEMMOVE, "__aeabi_memmove", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+      { RTLIB::MEMSET,  "__aeabi_memset",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+    };
+
+    for (const auto &LC : LibraryCalls) {
+      setLibcallName(LC.Op, LC.Name);
+      setLibcallCallingConv(LC.Op, LC.CC);
+      if (LC.Cond != ISD::SETCC_INVALID)
+        setCmpLibcallCC(LC.Op, LC.Cond);
+    }
+  }
+
+  if (Subtarget->isTargetWindows()) {
+    static const struct {
+      const RTLIB::Libcall Op;
+      const char * const Name;
+      const CallingConv::ID CC;
+    } LibraryCalls[] = {
+      { RTLIB::FPTOSINT_F32_I64, "__stoi64", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::FPTOSINT_F64_I64, "__dtoi64", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::FPTOUINT_F32_I64, "__stou64", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::FPTOUINT_F64_I64, "__dtou64", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::SINTTOFP_I64_F32, "__i64tos", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::SINTTOFP_I64_F64, "__i64tod", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::UINTTOFP_I64_F32, "__u64tos", CallingConv::ARM_AAPCS_VFP },
+      { RTLIB::UINTTOFP_I64_F64, "__u64tod", CallingConv::ARM_AAPCS_VFP },
+    };
+
+    for (const auto &LC : LibraryCalls) {
+      setLibcallName(LC.Op, LC.Name);
+      setLibcallCallingConv(LC.Op, LC.CC);
+    }
   }
 
   // Use divmod compiler-rt calls for iOS 5.0 and later.
@@ -444,6 +409,13 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
     setLoadExtAction(ISD::SEXTLOAD, (MVT::SimpleValueType)VT, Expand);
     setLoadExtAction(ISD::ZEXTLOAD, (MVT::SimpleValueType)VT, Expand);
     setLoadExtAction(ISD::EXTLOAD, (MVT::SimpleValueType)VT, Expand);
+
+    setOperationAction(ISD::MULHS, (MVT::SimpleValueType)VT, Expand);
+    setOperationAction(ISD::SMUL_LOHI, (MVT::SimpleValueType)VT, Expand);
+    setOperationAction(ISD::MULHU, (MVT::SimpleValueType)VT, Expand);
+    setOperationAction(ISD::UMUL_LOHI, (MVT::SimpleValueType)VT, Expand);
+
+    setOperationAction(ISD::BSWAP, (MVT::SimpleValueType)VT, Expand);
   }
 
   setOperationAction(ISD::ConstantFP, MVT::f32, Custom);
@@ -631,6 +603,11 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
     }
   }
 
+  setOperationAction(ISD::SADDO, MVT::i32, Custom);
+  setOperationAction(ISD::UADDO, MVT::i32, Custom);
+  setOperationAction(ISD::SSUBO, MVT::i32, Custom);
+  setOperationAction(ISD::USUBO, MVT::i32, Custom);
+
   // i64 operation support.
   setOperationAction(ISD::MUL,     MVT::i64, Expand);
   setOperationAction(ISD::MULHU,   MVT::i32, Expand);
@@ -850,7 +827,7 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
       setOperationAction(ISD::FP32_TO_FP16, MVT::i32, Expand);
     }
   }
-      
+
   // Combine sin / cos into one node or libcall if possible.
   if (Subtarget->hasSinCos()) {
     setLibcallName(RTLIB::SINCOS_F32, "sincosf");
@@ -913,7 +890,7 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM)
 // and extractions.
 std::pair<const TargetRegisterClass*, uint8_t>
 ARMTargetLowering::findRepresentativeClass(MVT VT) const{
-  const TargetRegisterClass *RRC = 0;
+  const TargetRegisterClass *RRC = nullptr;
   uint8_t Cost = 1;
   switch (VT.SimpleTy) {
   default:
@@ -950,7 +927,7 @@ ARMTargetLowering::findRepresentativeClass(MVT VT) const{
 
 const char *ARMTargetLowering::getTargetNodeName(unsigned Opcode) const {
   switch (Opcode) {
-  default: return 0;
+  default: return nullptr;
   case ARMISD::Wrapper:       return "ARMISD::Wrapper";
   case ARMISD::WrapperPIC:    return "ARMISD::WrapperPIC";
   case ARMISD::WrapperJT:     return "ARMISD::WrapperJT";
@@ -1204,40 +1181,58 @@ static void FPCCToARMCC(ISD::CondCode CC, ARMCC::CondCodes &CondCode,
 
 #include "ARMGenCallingConv.inc"
 
-/// CCAssignFnForNode - Selects the correct CCAssignFn for a the
-/// given CallingConvention value.
-CCAssignFn *ARMTargetLowering::CCAssignFnForNode(CallingConv::ID CC,
-                                                 bool Return,
-                                                 bool isVarArg) const {
+/// getEffectiveCallingConv - Get the effective calling convention, taking into
+/// account presence of floating point hardware and calling convention
+/// limitations, such as support for variadic functions.
+CallingConv::ID
+ARMTargetLowering::getEffectiveCallingConv(CallingConv::ID CC,
+                                           bool isVarArg) const {
   switch (CC) {
   default:
     llvm_unreachable("Unsupported calling convention");
-  case CallingConv::Fast:
-    if (Subtarget->hasVFP2() && !isVarArg) {
-      if (!Subtarget->isAAPCS_ABI())
-        return (Return ? RetFastCC_ARM_APCS : FastCC_ARM_APCS);
-      // For AAPCS ABI targets, just use VFP variant of the calling convention.
-      return (Return ? RetCC_ARM_AAPCS_VFP : CC_ARM_AAPCS_VFP);
-    }
-    // Fallthrough
-  case CallingConv::C: {
-    // Use target triple & subtarget features to do actual dispatch.
+  case CallingConv::ARM_AAPCS:
+  case CallingConv::ARM_APCS:
+  case CallingConv::GHC:
+    return CC;
+  case CallingConv::ARM_AAPCS_VFP:
+    return isVarArg ? CallingConv::ARM_AAPCS : CallingConv::ARM_AAPCS_VFP;
+  case CallingConv::C:
     if (!Subtarget->isAAPCS_ABI())
-      return (Return ? RetCC_ARM_APCS : CC_ARM_APCS);
+      return CallingConv::ARM_APCS;
     else if (Subtarget->hasVFP2() &&
              getTargetMachine().Options.FloatABIType == FloatABI::Hard &&
              !isVarArg)
-      return (Return ? RetCC_ARM_AAPCS_VFP : CC_ARM_AAPCS_VFP);
-    return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
+      return CallingConv::ARM_AAPCS_VFP;
+    else
+      return CallingConv::ARM_AAPCS;
+  case CallingConv::Fast:
+    if (!Subtarget->isAAPCS_ABI()) {
+      if (Subtarget->hasVFP2() && !isVarArg)
+        return CallingConv::Fast;
+      return CallingConv::ARM_APCS;
+    } else if (Subtarget->hasVFP2() && !isVarArg)
+      return CallingConv::ARM_AAPCS_VFP;
+    else
+      return CallingConv::ARM_AAPCS;
   }
-  case CallingConv::ARM_AAPCS_VFP:
-    if (!isVarArg)
-      return (Return ? RetCC_ARM_AAPCS_VFP : CC_ARM_AAPCS_VFP);
-    // Fallthrough
-  case CallingConv::ARM_AAPCS:
-    return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
+}
+
+/// CCAssignFnForNode - Selects the correct CCAssignFn for the given
+/// CallingConvention.
+CCAssignFn *ARMTargetLowering::CCAssignFnForNode(CallingConv::ID CC,
+                                                 bool Return,
+                                                 bool isVarArg) const {
+  switch (getEffectiveCallingConv(CC, isVarArg)) {
+  default:
+    llvm_unreachable("Unsupported calling convention");
   case CallingConv::ARM_APCS:
     return (Return ? RetCC_ARM_APCS : CC_ARM_APCS);
+  case CallingConv::ARM_AAPCS:
+    return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
+  case CallingConv::ARM_AAPCS_VFP:
+    return (Return ? RetCC_ARM_AAPCS_VFP : CC_ARM_AAPCS_VFP);
+  case CallingConv::Fast:
+    return (Return ? RetFastCC_ARM_APCS : FastCC_ARM_APCS);
   case CallingConv::GHC:
     return (Return ? RetCC_ARM_APCS : CC_ARM_APCS_GHC);
   }
@@ -1286,6 +1281,8 @@ ARMTargetLowering::LowerCallResult(SDValue Chain, SDValue InFlag,
                                       InFlag);
       Chain = Hi.getValue(1);
       InFlag = Hi.getValue(2);
+      if (!Subtarget->isLittle())
+        std::swap (Lo, Hi);
       Val = DAG.getNode(ARMISD::VMOVDRR, dl, MVT::f64, Lo, Hi);
 
       if (VA.getLocVT() == MVT::v2f64) {
@@ -1301,6 +1298,8 @@ ARMTargetLowering::LowerCallResult(SDValue Chain, SDValue InFlag,
         Hi = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32, InFlag);
         Chain = Hi.getValue(1);
         InFlag = Hi.getValue(2);
+        if (!Subtarget->isLittle())
+          std::swap (Lo, Hi);
         Val = DAG.getNode(ARMISD::VMOVDRR, dl, MVT::f64, Lo, Hi);
         Val = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, MVT::v2f64, Vec, Val,
                           DAG.getConstant(1, MVT::i32));
@@ -1351,16 +1350,17 @@ void ARMTargetLowering::PassF64ArgInRegs(SDLoc dl, SelectionDAG &DAG,
 
   SDValue fmrrd = DAG.getNode(ARMISD::VMOVRRD, dl,
                               DAG.getVTList(MVT::i32, MVT::i32), Arg);
-  RegsToPass.push_back(std::make_pair(VA.getLocReg(), fmrrd));
+  unsigned id = Subtarget->isLittle() ? 0 : 1;
+  RegsToPass.push_back(std::make_pair(VA.getLocReg(), fmrrd.getValue(id)));
 
   if (NextVA.isRegLoc())
-    RegsToPass.push_back(std::make_pair(NextVA.getLocReg(), fmrrd.getValue(1)));
+    RegsToPass.push_back(std::make_pair(NextVA.getLocReg(), fmrrd.getValue(1-id)));
   else {
     assert(NextVA.isMemLoc());
-    if (StackPtr.getNode() == 0)
+    if (!StackPtr.getNode())
       StackPtr = DAG.getCopyFromReg(Chain, dl, ARM::SP, getPointerTy());
 
-    MemOpChains.push_back(LowerMemOpCallTo(Chain, StackPtr, fmrrd.getValue(1),
+    MemOpChains.push_back(LowerMemOpCallTo(Chain, StackPtr, fmrrd.getValue(1-id),
                                            dl, DAG, NextVA,
                                            Flags));
   }
@@ -1398,6 +1398,9 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
     isTailCall = IsEligibleForTailCallOptimization(Callee, CallConv,
                     isVarArg, isStructRet, MF.getFunction()->hasStructRetAttr(),
                                                    Outs, OutVals, Ins, DAG);
+    if (!isTailCall && CLI.CS && CLI.CS->isMustTailCall())
+      report_fatal_error("failed to perform tail call elimination on a call "
+                         "site marked musttail");
     // We don't support GuaranteedTailCallOpt for ARM, only automatically
     // detected sibcalls.
     if (isTailCall) {
@@ -1542,7 +1545,7 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
         SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue);
         SDValue Ops[] = { Chain, Dst, Src, SizeNode, AlignNode};
         MemOpChains.push_back(DAG.getNode(ARMISD::COPY_STRUCT_BYVAL, dl, VTs,
-                                          Ops, array_lengthof(Ops)));
+                                          Ops));
       }
     } else if (!isSibCall) {
       assert(VA.isMemLoc());
@@ -1553,8 +1556,7 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   }
 
   if (!MemOpChains.empty())
-    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
-                        &MemOpChains[0], MemOpChains.size());
+    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
 
   // Build a sequence of copy-to-reg nodes chained together with token chain
   // and flag operands which copy the outgoing args into the appropriate regs.
@@ -1741,10 +1743,10 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
 
   SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
   if (isTailCall)
-    return DAG.getNode(ARMISD::TC_RETURN, dl, NodeTys, &Ops[0], Ops.size());
+    return DAG.getNode(ARMISD::TC_RETURN, dl, NodeTys, Ops);
 
   // Returns a chain and a flag for retval copy to use.
-  Chain = DAG.getNode(CallOpc, dl, NodeTys, &Ops[0], Ops.size());
+  Chain = DAG.getNode(CallOpc, dl, NodeTys, Ops);
   InFlag = Chain.getValue(1);
 
   Chain = DAG.getCALLSEQ_END(Chain, DAG.getIntPtrConstant(NumBytes, true),
@@ -2049,8 +2051,7 @@ static SDValue LowerInterruptReturn(SmallVectorImpl<SDValue> &RetOps,
 
   RetOps.insert(RetOps.begin() + 1, DAG.getConstant(LROffset, MVT::i32, false));
 
-  return DAG.getNode(ARMISD::INTRET_FLAG, DL, MVT::Other,
-                     RetOps.data(), RetOps.size());
+  return DAG.getNode(ARMISD::INTRET_FLAG, DL, MVT::Other, RetOps);
 }
 
 SDValue
@@ -2074,6 +2075,7 @@ ARMTargetLowering::LowerReturn(SDValue Chain,
   SDValue Flag;
   SmallVector<SDValue, 4> RetOps;
   RetOps.push_back(Chain); // Operand #0 = Chain (updated below)
+  bool isLittleEndian = Subtarget->isLittle();
 
   // Copy the result values into the output registers.
   for (unsigned i = 0, realRVLocIdx = 0;
@@ -2100,12 +2102,15 @@ ARMTargetLowering::LowerReturn(SDValue Chain,
         SDValue HalfGPRs = DAG.getNode(ARMISD::VMOVRRD, dl,
                                        DAG.getVTList(MVT::i32, MVT::i32), Half);
 
-        Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), HalfGPRs, Flag);
+        Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
+                                 HalfGPRs.getValue(isLittleEndian ? 0 : 1),
+                                 Flag);
         Flag = Chain.getValue(1);
         RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
         VA = RVLocs[++i]; // skip ahead to next loc
         Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
-                                 HalfGPRs.getValue(1), Flag);
+                                 HalfGPRs.getValue(isLittleEndian ? 1 : 0),
+                                 Flag);
         Flag = Chain.getValue(1);
         RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
         VA = RVLocs[++i]; // skip ahead to next loc
@@ -2117,12 +2122,15 @@ ARMTargetLowering::LowerReturn(SDValue Chain,
       // Legalize ret f64 -> ret 2 x i32.  We always have fmrrd if f64 is
       // available.
       SDValue fmrrd = DAG.getNode(ARMISD::VMOVRRD, dl,
-                                  DAG.getVTList(MVT::i32, MVT::i32), &Arg, 1);
-      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), fmrrd, Flag);
+                                  DAG.getVTList(MVT::i32, MVT::i32), Arg);
+      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
+                               fmrrd.getValue(isLittleEndian ? 0 : 1),
+                               Flag);
       Flag = Chain.getValue(1);
       RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
       VA = RVLocs[++i]; // skip ahead to next loc
-      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), fmrrd.getValue(1),
+      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
+                               fmrrd.getValue(isLittleEndian ? 1 : 0),
                                Flag);
     } else
       Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), Arg, Flag);
@@ -2151,8 +2159,7 @@ ARMTargetLowering::LowerReturn(SDValue Chain,
     return LowerInterruptReturn(RetOps, dl, DAG);
   }
 
-  return DAG.getNode(ARMISD::RET_FLAG, dl, MVT::Other,
-                     RetOps.data(), RetOps.size());
+  return DAG.getNode(ARMISD::RET_FLAG, dl, MVT::Other, RetOps);
 }
 
 bool ARMTargetLowering::isUsedByReturnOnly(SDNode *N, SDValue &Chain) const {
@@ -2314,13 +2321,13 @@ ARMTargetLowering::LowerToTLSGeneralDynamicModel(GlobalAddressSDNode *GA,
   Entry.Node = Argument;
   Entry.Ty = (Type *) Type::getInt32Ty(*DAG.getContext());
   Args.push_back(Entry);
+
   // FIXME: is there useful debug info available here?
-  TargetLowering::CallLoweringInfo CLI(Chain,
-                (Type *) Type::getInt32Ty(*DAG.getContext()),
-                false, false, false, false,
-                0, CallingConv::C, /*isTailCall=*/false,
-                /*doesNotRet=*/false, /*isReturnValueUsed=*/true,
-                DAG.getExternalSymbol("__tls_get_addr", PtrVT), Args, DAG, dl);
+  TargetLowering::CallLoweringInfo CLI(DAG);
+  CLI.setDebugLoc(dl).setChain(Chain)
+    .setCallee(CallingConv::C, Type::getInt32Ty(*DAG.getContext()),
+               DAG.getExternalSymbol("__tls_get_addr", PtrVT), &Args, 0);
+
   std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
   return CallResult.first;
 }
@@ -2466,6 +2473,23 @@ SDValue ARMTargetLowering::LowerGlobalAddressDarwin(SDValue Op,
   return Result;
 }
 
+SDValue ARMTargetLowering::LowerGlobalAddressWindows(SDValue Op,
+                                                     SelectionDAG &DAG) const {
+  assert(Subtarget->isTargetWindows() && "non-Windows COFF is not supported");
+  assert(Subtarget->useMovt() && "Windows on ARM expects to use movw/movt");
+
+  const GlobalValue *GV = cast<GlobalAddressSDNode>(Op)->getGlobal();
+  EVT PtrVT = getPointerTy();
+  SDLoc DL(Op);
+
+  ++NumMovwMovt;
+
+  // FIXME: Once remat is capable of dealing with instructions with register
+  // operands, expand this into two nodes.
+  return DAG.getNode(ARMISD::Wrapper, DL, PtrVT,
+                     DAG.getTargetGlobalAddress(GV, DL, PtrVT));
+}
+
 SDValue ARMTargetLowering::LowerGLOBAL_OFFSET_TABLE(SDValue Op,
                                                     SelectionDAG &DAG) const {
   assert(Subtarget->isTargetELF() &&
@@ -2654,7 +2678,8 @@ ARMTargetLowering::GetF64FormalArgument(CCValAssign &VA, CCValAssign &NextVA,
     Reg = MF.addLiveIn(NextVA.getLocReg(), RC);
     ArgValue2 = DAG.getCopyFromReg(Root, dl, Reg, MVT::i32);
   }
-
+  if (!Subtarget->isLittle())
+    std::swap (ArgValue, ArgValue2);
   return DAG.getNode(ARMISD::VMOVDRR, dl, MVT::f64, ArgValue, ArgValue2);
 }
 
@@ -2803,8 +2828,7 @@ ARMTargetLowering::StoreByValRegs(CCState &CCInfo, SelectionDAG &DAG,
     AFI->setArgRegsSaveSize(ArgRegsSaveSize + AFI->getArgRegsSaveSize());
 
     if (!MemOps.empty())
-      Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
-                          &MemOps[0], MemOps.size());
+      Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOps);
     return FrameIndex;
   } else {
     if (ArgSize == 0) {
@@ -2834,8 +2858,9 @@ ARMTargetLowering::VarArgStyleRegisters(CCState &CCInfo, SelectionDAG &DAG,
   // If there is no regs to be stored, just point address after last
   // argument passed via stack.
   int FrameIndex =
-    StoreByValRegs(CCInfo, DAG, dl, Chain, 0, CCInfo.getInRegsParamsCount(),
-                   0, ArgOffset, 0, ForceMutable, 0, TotalArgRegsSaveSize);
+    StoreByValRegs(CCInfo, DAG, dl, Chain, nullptr,
+                   CCInfo.getInRegsParamsCount(), 0, ArgOffset, 0, ForceMutable,
+                   0, TotalArgRegsSaveSize);
 
   AFI->setVarArgsFrameIndex(FrameIndex);
 }
@@ -3166,11 +3191,96 @@ ARMTargetLowering::duplicateCmp(SDValue Cmp, SelectionDAG &DAG) const {
   return DAG.getNode(ARMISD::FMSTAT, DL, MVT::Glue, Cmp);
 }
 
+std::pair<SDValue, SDValue>
+ARMTargetLowering::getARMXALUOOp(SDValue Op, SelectionDAG &DAG,
+                                 SDValue &ARMcc) const {
+  assert(Op.getValueType() == MVT::i32 &&  "Unsupported value type");
+
+  SDValue Value, OverflowCmp;
+  SDValue LHS = Op.getOperand(0);
+  SDValue RHS = Op.getOperand(1);
+
+
+  // FIXME: We are currently always generating CMPs because we don't support
+  // generating CMN through the backend. This is not as good as the natural
+  // CMP case because it causes a register dependency and cannot be folded
+  // later.
+
+  switch (Op.getOpcode()) {
+  default:
+    llvm_unreachable("Unknown overflow instruction!");
+  case ISD::SADDO:
+    ARMcc = DAG.getConstant(ARMCC::VC, MVT::i32);
+    Value = DAG.getNode(ISD::ADD, SDLoc(Op), Op.getValueType(), LHS, RHS);
+    OverflowCmp = DAG.getNode(ARMISD::CMP, SDLoc(Op), MVT::Glue, Value, LHS);
+    break;
+  case ISD::UADDO:
+    ARMcc = DAG.getConstant(ARMCC::HS, MVT::i32);
+    Value = DAG.getNode(ISD::ADD, SDLoc(Op), Op.getValueType(), LHS, RHS);
+    OverflowCmp = DAG.getNode(ARMISD::CMP, SDLoc(Op), MVT::Glue, Value, LHS);
+    break;
+  case ISD::SSUBO:
+    ARMcc = DAG.getConstant(ARMCC::VC, MVT::i32);
+    Value = DAG.getNode(ISD::SUB, SDLoc(Op), Op.getValueType(), LHS, RHS);
+    OverflowCmp = DAG.getNode(ARMISD::CMP, SDLoc(Op), MVT::Glue, LHS, RHS);
+    break;
+  case ISD::USUBO:
+    ARMcc = DAG.getConstant(ARMCC::HS, MVT::i32);
+    Value = DAG.getNode(ISD::SUB, SDLoc(Op), Op.getValueType(), LHS, RHS);
+    OverflowCmp = DAG.getNode(ARMISD::CMP, SDLoc(Op), MVT::Glue, LHS, RHS);
+    break;
+  } // switch (...)
+
+  return std::make_pair(Value, OverflowCmp);
+}
+
+
+SDValue
+ARMTargetLowering::LowerXALUO(SDValue Op, SelectionDAG &DAG) const {
+  // Let legalize expand this if it isn't a legal type yet.
+  if (!DAG.getTargetLoweringInfo().isTypeLegal(Op.getValueType()))
+    return SDValue();
+
+  SDValue Value, OverflowCmp;
+  SDValue ARMcc;
+  std::tie(Value, OverflowCmp) = getARMXALUOOp(Op, DAG, ARMcc);
+  SDValue CCR = DAG.getRegister(ARM::CPSR, MVT::i32);
+  // We use 0 and 1 as false and true values.
+  SDValue TVal = DAG.getConstant(1, MVT::i32);
+  SDValue FVal = DAG.getConstant(0, MVT::i32);
+  EVT VT = Op.getValueType();
+
+  SDValue Overflow = DAG.getNode(ARMISD::CMOV, SDLoc(Op), VT, TVal, FVal,
+                                 ARMcc, CCR, OverflowCmp);
+
+  SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
+  return DAG.getNode(ISD::MERGE_VALUES, SDLoc(Op), VTs, Value, Overflow);
+}
+
+
 SDValue ARMTargetLowering::LowerSELECT(SDValue Op, SelectionDAG &DAG) const {
   SDValue Cond = Op.getOperand(0);
   SDValue SelectTrue = Op.getOperand(1);
   SDValue SelectFalse = Op.getOperand(2);
   SDLoc dl(Op);
+  unsigned Opc = Cond.getOpcode();
+
+  if (Cond.getResNo() == 1 &&
+      (Opc == ISD::SADDO || Opc == ISD::UADDO || Opc == ISD::SSUBO ||
+       Opc == ISD::USUBO)) {
+    if (!DAG.getTargetLoweringInfo().isTypeLegal(Cond->getValueType(0)))
+      return SDValue();
+
+    SDValue Value, OverflowCmp;
+    SDValue ARMcc;
+    std::tie(Value, OverflowCmp) = getARMXALUOOp(Cond, DAG, ARMcc);
+    SDValue CCR = DAG.getRegister(ARM::CPSR, MVT::i32);
+    EVT VT = Op.getValueType();
+
+    return DAG.getNode(ARMISD::CMOV, SDLoc(Op), VT, SelectTrue, SelectFalse,
+                       ARMcc, CCR, OverflowCmp);
+
+  }
 
   // Convert:
   //
@@ -3472,7 +3582,7 @@ ARMTargetLowering::OptimizeVFPBrcond(SDValue Op, SelectionDAG &DAG) const {
     ARMcc = DAG.getConstant(CondCode, MVT::i32);
     SDVTList VTList = DAG.getVTList(MVT::Other, MVT::Glue);
     SDValue Ops[] = { Chain, ARMcc, LHS1, LHS2, RHS1, RHS2, Dest };
-    return DAG.getNode(ARMISD::BCC_i64, dl, VTList, Ops, 7);
+    return DAG.getNode(ARMISD::BCC_i64, dl, VTList, Ops);
   }
 
   return SDValue();
@@ -3512,11 +3622,11 @@ SDValue ARMTargetLowering::LowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
   SDValue CCR = DAG.getRegister(ARM::CPSR, MVT::i32);
   SDVTList VTList = DAG.getVTList(MVT::Other, MVT::Glue);
   SDValue Ops[] = { Chain, Dest, ARMcc, CCR, Cmp };
-  SDValue Res = DAG.getNode(ARMISD::BRCOND, dl, VTList, Ops, 5);
+  SDValue Res = DAG.getNode(ARMISD::BRCOND, dl, VTList, Ops);
   if (CondCode2 != ARMCC::AL) {
     ARMcc = DAG.getConstant(CondCode2, MVT::i32);
     SDValue Ops[] = { Res, Dest, ARMcc, CCR, Res.getValue(1) };
-    Res = DAG.getNode(ARMISD::BRCOND, dl, VTList, Ops, 5);
+    Res = DAG.getNode(ARMISD::BRCOND, dl, VTList, Ops);
   }
   return Res;
 }
@@ -3713,7 +3823,7 @@ SDValue ARMTargetLowering::LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) const {
   // Bitcast operand 1 to i32.
   if (SrcVT == MVT::f64)
     Tmp1 = DAG.getNode(ARMISD::VMOVRRD, dl, DAG.getVTList(MVT::i32, MVT::i32),
-                       &Tmp1, 1).getValue(1);
+                       Tmp1).getValue(1);
   Tmp1 = DAG.getNode(ISD::BITCAST, dl, MVT::i32, Tmp1);
 
   // Or in the signbit with integer operations.
@@ -3729,7 +3839,7 @@ SDValue ARMTargetLowering::LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) const {
 
   // f64: Or the high part with signbit and then combine two parts.
   Tmp0 = DAG.getNode(ARMISD::VMOVRRD, dl, DAG.getVTList(MVT::i32, MVT::i32),
-                     &Tmp0, 1);
+                     Tmp0);
   SDValue Lo = Tmp0.getValue(0);
   SDValue Hi = DAG.getNode(ISD::AND, dl, MVT::i32, Tmp0.getValue(1), Mask2);
   Hi = DAG.getNode(ISD::OR, dl, MVT::i32, Hi, Tmp1);
@@ -3761,14 +3871,16 @@ SDValue ARMTargetLowering::LowerRETURNADDR(SDValue Op, SelectionDAG &DAG) const{
 }
 
 SDValue ARMTargetLowering::LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const {
-  MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
+  const ARMBaseRegisterInfo &ARI =
+    *static_cast<const ARMBaseRegisterInfo*>(RegInfo);
+  MachineFunction &MF = DAG.getMachineFunction();
+  MachineFrameInfo *MFI = MF.getFrameInfo();
   MFI->setFrameAddressIsTaken(true);
 
   EVT VT = Op.getValueType();
   SDLoc dl(Op);  // FIXME probably not meaningful
   unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
-  unsigned FrameReg = (Subtarget->isThumb() || Subtarget->isTargetMachO())
-    ? ARM::R7 : ARM::R11;
+  unsigned FrameReg = ARI.getFrameRegister(MF);
   SDValue FrameAddr = DAG.getCopyFromReg(DAG.getEntryNode(), dl, FrameReg, VT);
   while (Depth--)
     FrameAddr = DAG.getLoad(VT, dl, DAG.getEntryNode(), FrameAddr,
@@ -3777,6 +3889,18 @@ SDValue ARMTargetLowering::LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const {
   return FrameAddr;
 }
 
+// FIXME? Maybe this could be a TableGen attribute on some registers and
+// this table could be generated automatically from RegInfo.
+unsigned ARMTargetLowering::getRegisterByName(const char* RegName,
+                                              EVT VT) const {
+  unsigned Reg = StringSwitch<unsigned>(RegName)
+                       .Case("sp", ARM::SP)
+                       .Default(0);
+  if (Reg)
+    return Reg;
+  report_fatal_error("Invalid register name global variable");
+}
+
 /// ExpandBITCAST - If the target supports VFP, this function is called to
 /// expand a bit convert where either the source or destination type is i64 to
 /// use a VMOVDRR or VMOVRRD node.  This should not be done when the non-i64
@@ -3806,8 +3930,15 @@ static SDValue ExpandBITCAST(SDNode *N, SelectionDAG &DAG) {
 
   // Turn f64->i64 into VMOVRRD.
   if (DstVT == MVT::i64 && TLI.isTypeLegal(SrcVT)) {
-    SDValue Cvt = DAG.getNode(ARMISD::VMOVRRD, dl,
-                              DAG.getVTList(MVT::i32, MVT::i32), &Op, 1);
+    SDValue Cvt;
+    if (TLI.isBigEndian() && SrcVT.isVector() &&
+        SrcVT.getVectorNumElements() > 1)
+      Cvt = DAG.getNode(ARMISD::VMOVRRD, dl,
+                        DAG.getVTList(MVT::i32, MVT::i32),
+                        DAG.getNode(ARMISD::VREV64, dl, SrcVT, Op));
+    else
+      Cvt = DAG.getNode(ARMISD::VMOVRRD, dl,
+                        DAG.getVTList(MVT::i32, MVT::i32), Op);
     // Merge the pieces into a single i64 value.
     return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Cvt, Cvt.getValue(1));
   }
@@ -3863,7 +3994,7 @@ SDValue ARMTargetLowering::LowerShiftRightParts(SDValue Op,
                            CCR, Cmp);
 
   SDValue Ops[2] = { Lo, Hi };
-  return DAG.getMergeValues(Ops, 2, dl);
+  return DAG.getMergeValues(Ops, dl);
 }
 
 /// LowerShiftLeftParts - Lower SHL_PARTS, which returns two
@@ -3897,7 +4028,7 @@ SDValue ARMTargetLowering::LowerShiftLeftParts(SDValue Op,
                            CCR, Cmp);
 
   SDValue Ops[2] = { Lo, Hi };
-  return DAG.getMergeValues(Ops, 2, dl);
+  return DAG.getMergeValues(Ops, dl);
 }
 
 SDValue ARMTargetLowering::LowerFLT_ROUNDS_(SDValue Op,
@@ -4102,7 +4233,7 @@ static SDValue Expand64BitShift(SDNode *N, SelectionDAG &DAG,
   // First, build a SRA_FLAG/SRL_FLAG op, which shifts the top part by one and
   // captures the result into a carry flag.
   unsigned Opc = N->getOpcode() == ISD::SRL ? ARMISD::SRL_FLAG:ARMISD::SRA_FLAG;
-  Hi = DAG.getNode(Opc, dl, DAG.getVTList(MVT::i32, MVT::Glue), &Hi, 1);
+  Hi = DAG.getNode(Opc, dl, DAG.getVTList(MVT::i32, MVT::Glue), Hi);
 
   // The low part is an ARMISD::RRX operand, which shifts the carry in.
   Lo = DAG.getNode(ARMISD::RRX, dl, MVT::i32, Lo, Hi.getValue(1));
@@ -4859,7 +4990,7 @@ SDValue ARMTargetLowering::LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG,
           Ops.push_back(N);
           Ops.push_back(Op.getOperand(I));
           Ops.push_back(DAG.getConstant(I, MVT::i32));
-          N = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, VT, &Ops[0], 3);
+          N = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, VT, Ops);
         }
       }
       return N;
@@ -4870,7 +5001,7 @@ SDValue ARMTargetLowering::LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG,
         Ops.push_back(DAG.getNode(ISD::BITCAST, dl, MVT::i32,
                                   Op.getOperand(i)));
       EVT VecVT = EVT::getVectorVT(*DAG.getContext(), MVT::i32, NumElts);
-      SDValue Val = DAG.getNode(ISD::BUILD_VECTOR, dl, VecVT, &Ops[0], NumElts);
+      SDValue Val = DAG.getNode(ISD::BUILD_VECTOR, dl, VecVT, Ops);
       Val = LowerBUILD_VECTOR(Val, DAG, ST);
       if (Val.getNode())
         return DAG.getNode(ISD::BITCAST, dl, VT, Val);
@@ -4906,7 +5037,7 @@ SDValue ARMTargetLowering::LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG,
     SmallVector<SDValue, 8> Ops;
     for (unsigned i = 0; i < NumElts; ++i)
       Ops.push_back(DAG.getNode(ISD::BITCAST, dl, EltVT, Op.getOperand(i)));
-    SDValue Val = DAG.getNode(ARMISD::BUILD_VECTOR, dl, VecVT, &Ops[0],NumElts);
+    SDValue Val = DAG.getNode(ARMISD::BUILD_VECTOR, dl, VecVT, Ops);
     return DAG.getNode(ISD::BITCAST, dl, VT, Val);
   }
 
@@ -5213,12 +5344,10 @@ static SDValue LowerVECTOR_SHUFFLEv8i8(SDValue Op,
 
   if (V2.getNode()->getOpcode() == ISD::UNDEF)
     return DAG.getNode(ARMISD::VTBL1, DL, MVT::v8i8, V1,
-                       DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v8i8,
-                                   &VTBLMask[0], 8));
+                       DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v8i8, VTBLMask));
 
   return DAG.getNode(ARMISD::VTBL2, DL, MVT::v8i8, V1, V2,
-                     DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v8i8,
-                                 &VTBLMask[0], 8));
+                     DAG.getNode(ISD::BUILD_VECTOR, DL, MVT::v8i8, VTBLMask));
 }
 
 static SDValue LowerReverse_VECTOR_SHUFFLEv16i8_v8i16(SDValue Op,
@@ -5371,7 +5500,7 @@ static SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) {
                                   DAG.getConstant(ShuffleMask[i] & (NumElts-1),
                                                   MVT::i32)));
     }
-    SDValue Val = DAG.getNode(ARMISD::BUILD_VECTOR, dl, VecVT, &Ops[0],NumElts);
+    SDValue Val = DAG.getNode(ARMISD::BUILD_VECTOR, dl, VecVT, Ops);
     return DAG.getNode(ISD::BITCAST, dl, VT, Val);
   }
 
@@ -5608,7 +5737,7 @@ static SDValue SkipExtensionForVMULL(SDNode *N, SelectionDAG &DAG) {
     Ops.push_back(DAG.getConstant(CInt.zextOrTrunc(32), MVT::i32));
   }
   return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N),
-                     MVT::getVectorVT(TruncVT, NumElts), Ops.data(), NumElts);
+                     MVT::getVectorVT(TruncVT, NumElts), Ops);
 }
 
 static bool isAddSubSExt(SDNode *N, SelectionDAG &DAG) {
@@ -5946,12 +6075,12 @@ SDValue ARMTargetLowering::LowerFSINCOS(SDValue Op, SelectionDAG &DAG) const {
   ? "__sincos_stret" : "__sincosf_stret";
   SDValue Callee = DAG.getExternalSymbol(LibcallName, getPointerTy());
 
-  TargetLowering::
-  CallLoweringInfo CLI(DAG.getEntryNode(), Type::getVoidTy(*DAG.getContext()),
-                       false, false, false, false, 0,
-                       CallingConv::C, /*isTaillCall=*/false,
-                       /*doesNotRet=*/false, /*isReturnValueUsed*/false,
-                       Callee, Args, DAG, dl);
+  TargetLowering::CallLoweringInfo CLI(DAG);
+  CLI.setDebugLoc(dl).setChain(DAG.getEntryNode())
+    .setCallee(CallingConv::C, Type::getVoidTy(*DAG.getContext()), Callee,
+               &Args, 0)
+    .setDiscardResult();
+
   std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
 
   SDValue LoadSin = DAG.getLoad(ArgVT, dl, CallResult.second, SRet,
@@ -5998,8 +6127,7 @@ static void ReplaceREADCYCLECOUNTER(SDNode *N,
     };
 
     Cycles32 = DAG.getNode(ISD::INTRINSIC_W_CHAIN, DL,
-                           DAG.getVTList(MVT::i32, MVT::Other), &Ops[0],
-                           array_lengthof(Ops));
+                           DAG.getVTList(MVT::i32, MVT::Other), Ops);
     OutChain = Cycles32.getValue(1);
   } else {
     // Intrinsic is defined to return 0 on unsupported platforms. Technically
@@ -6022,8 +6150,15 @@ SDValue ARMTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
   case ISD::ConstantPool:  return LowerConstantPool(Op, DAG);
   case ISD::BlockAddress:  return LowerBlockAddress(Op, DAG);
   case ISD::GlobalAddress:
-    return Subtarget->isTargetMachO() ? LowerGlobalAddressDarwin(Op, DAG) :
-      LowerGlobalAddressELF(Op, DAG);
+    switch (Subtarget->getTargetTriple().getObjectFormat()) {
+    default: llvm_unreachable("unknown object format");
+    case Triple::COFF:
+      return LowerGlobalAddressWindows(Op, DAG);
+    case Triple::ELF:
+      return LowerGlobalAddressELF(Op, DAG);
+    case Triple::MachO:
+      return LowerGlobalAddressDarwin(Op, DAG);
+    }
   case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG);
   case ISD::SELECT:        return LowerSELECT(Op, DAG);
   case ISD::SELECT_CC:     return LowerSELECT_CC(Op, DAG);
@@ -6068,6 +6203,11 @@ SDValue ARMTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
   case ISD::ADDE:
   case ISD::SUBC:
   case ISD::SUBE:          return LowerADDC_ADDE_SUBC_SUBE(Op, DAG);
+  case ISD::SADDO:
+  case ISD::UADDO:
+  case ISD::SSUBO:
+  case ISD::USUBO:
+    return LowerXALUO(Op, DAG);
   case ISD::ATOMIC_LOAD:
   case ISD::ATOMIC_STORE:  return LowerAtomicLoadStore(Op, DAG);
   case ISD::FSINCOS:       return LowerFSINCOS(Op, DAG);
@@ -6558,7 +6698,7 @@ EmitSjLjDispatchBlock(MachineInstr *MI, MachineBasicBlock *MBB) const {
   }
 
   // N.B. the order the invoke BBs are processed in doesn't matter here.
-  const uint16_t *SavedRegs = RI.getCalleeSavedRegs(MF);
+  const MCPhysReg *SavedRegs = RI.getCalleeSavedRegs(MF);
   SmallVector<MachineBasicBlock*, 64> MBBLPads;
   for (SmallPtrSet<MachineBasicBlock*, 64>::iterator
          I = InvokeBBs.begin(), E = InvokeBBs.end(); I != E; ++I) {
@@ -6755,8 +6895,8 @@ ARMTargetLowering::EmitStructByval(MachineInstr *MI,
   MachineFunction *MF = BB->getParent();
   MachineRegisterInfo &MRI = MF->getRegInfo();
   unsigned UnitSize = 0;
-  const TargetRegisterClass *TRC = 0;
-  const TargetRegisterClass *VecTRC = 0;
+  const TargetRegisterClass *TRC = nullptr;
+  const TargetRegisterClass *VecTRC = nullptr;
 
   bool IsThumb1 = Subtarget->isThumb1Only();
   bool IsThumb2 = Subtarget->isThumb2();
@@ -6790,7 +6930,7 @@ ARMTargetLowering::EmitStructByval(MachineInstr *MI,
                  ? (const TargetRegisterClass *)&ARM::DPairRegClass
                  : UnitSize == 8
                        ? (const TargetRegisterClass *)&ARM::DPRRegClass
-                       : 0;
+                       : nullptr;
 
   unsigned BytesLeft = SizeVal % UnitSize;
   unsigned LoopSize = SizeVal - BytesLeft;
@@ -7520,8 +7660,7 @@ static SDValue AddCombineToVPADDL(SDNode *N, SDValue N0, SDValue N1,
       llvm_unreachable("Invalid vector element type for padd optimization.");
   }
 
-  SDValue tmp = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, SDLoc(N),
-                            widenType, &Ops[0], Ops.size());
+  SDValue tmp = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, SDLoc(N), widenType, Ops);
   unsigned ExtOp = VT.bitsGT(tmp.getValueType()) ? ISD::ANY_EXTEND : ISD::TRUNCATE;
   return DAG.getNode(ExtOp, SDLoc(N), VT, tmp);
 }
@@ -7581,7 +7720,7 @@ static SDValue AddCombineTo64bitMLAL(SDNode *AddcNode,
 
   // Look for the glued ADDE.
   SDNode* AddeNode = AddcNode->getGluedUser();
-  if (AddeNode == NULL)
+  if (!AddeNode)
     return SDValue();
 
   // Make sure it is really an ADDE.
@@ -7616,9 +7755,9 @@ static SDValue AddCombineTo64bitMLAL(SDNode *AddcNode,
 
   // Figure out the high and low input values to the MLAL node.
   SDValue* HiMul = &MULOp;
-  SDValue* HiAdd = NULL;
-  SDValue* LoMul = NULL;
-  SDValue* LowAdd = NULL;
+  SDValue* HiAdd = nullptr;
+  SDValue* LoMul = nullptr;
+  SDValue* LowAdd = nullptr;
 
   if (IsLeftOperandMUL)
     HiAdd = &AddeOp1;
@@ -7635,7 +7774,7 @@ static SDValue AddCombineTo64bitMLAL(SDNode *AddcNode,
     LowAdd = &AddcOp0;
   }
 
-  if (LoMul == NULL)
+  if (!LoMul)
     return SDValue();
 
   if (LoMul->getNode() != HiMul->getNode())
@@ -7652,8 +7791,7 @@ static SDValue AddCombineTo64bitMLAL(SDNode *AddcNode,
   Ops.push_back(*HiAdd);
 
   SDValue MLALNode =  DAG.getNode(FinalOpc, SDLoc(AddcNode),
-                                 DAG.getVTList(MVT::i32, MVT::i32),
-                                 &Ops[0], Ops.size());
+                                 DAG.getVTList(MVT::i32, MVT::i32), Ops);
 
   // Replace the ADDs' nodes uses by the MLA node's values.
   SDValue HiMLALResult(MLALNode.getNode(), 1);
@@ -8290,8 +8428,7 @@ static SDValue PerformSTORECombine(SDNode *N,
                             Increment);
       Chains.push_back(Ch);
     }
-    return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, &Chains[0],
-                       Chains.size());
+    return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Chains);
   }
 
   if (!ISD::isNormalStore(St))
@@ -8302,16 +8439,18 @@ static SDValue PerformSTORECombine(SDNode *N,
   if (StVal.getNode()->getOpcode() == ARMISD::VMOVDRR &&
       StVal.getNode()->hasOneUse()) {
     SelectionDAG  &DAG = DCI.DAG;
+    bool isBigEndian = DAG.getTargetLoweringInfo().isBigEndian();
     SDLoc DL(St);
     SDValue BasePtr = St->getBasePtr();
     SDValue NewST1 = DAG.getStore(St->getChain(), DL,
-                                  StVal.getNode()->getOperand(0), BasePtr,
-                                  St->getPointerInfo(), St->isVolatile(),
+                                  StVal.getNode()->getOperand(isBigEndian ? 1 : 0 ),
+                                  BasePtr, St->getPointerInfo(), St->isVolatile(),
                                   St->isNonTemporal(), St->getAlignment());
 
     SDValue OffsetPtr = DAG.getNode(ISD::ADD, DL, MVT::i32, BasePtr,
                                     DAG.getConstant(4, MVT::i32));
-    return DAG.getStore(NewST1.getValue(0), DL, StVal.getNode()->getOperand(1),
+    return DAG.getStore(NewST1.getValue(0), DL,
+                        StVal.getNode()->getOperand(isBigEndian ? 0 : 1),
                         OffsetPtr, St->getPointerInfo(), St->isVolatile(),
                         St->isNonTemporal(),
                         std::min(4U, St->getAlignment() / 2));
@@ -8387,7 +8526,7 @@ static SDValue PerformBUILD_VECTORCombine(SDNode *N,
     DCI.AddToWorklist(V.getNode());
   }
   EVT FloatVT = EVT::getVectorVT(*DAG.getContext(), MVT::f64, NumElts);
-  SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, dl, FloatVT, Ops.data(), NumElts);
+  SDValue BV = DAG.getNode(ISD::BUILD_VECTOR, dl, FloatVT, Ops);
   return DAG.getNode(ISD::BITCAST, dl, VT, BV);
 }
 
@@ -8470,7 +8609,7 @@ PerformARMBUILD_VECTORCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI) {
       // Fold obvious case.
       V = V.getOperand(0);
     else {
-      V = DAG.getNode(ISD::BITCAST, SDLoc(V), MVT::i32, V); 
+      V = DAG.getNode(ISD::BITCAST, SDLoc(V), MVT::i32, V);
       // Make the DAGCombiner fold the bitcasts.
       DCI.AddToWorklist(V.getNode());
     }
@@ -8666,7 +8805,7 @@ static SDValue CombineBaseUpdate(SDNode *N,
       Tys[n] = VecTy;
     Tys[n++] = MVT::i32;
     Tys[n] = MVT::Other;
-    SDVTList SDTys = DAG.getVTList(Tys, NumResultVecs+2);
+    SDVTList SDTys = DAG.getVTList(ArrayRef<EVT>(Tys, NumResultVecs+2));
     SmallVector<SDValue, 8> Ops;
     Ops.push_back(N->getOperand(0)); // incoming chain
     Ops.push_back(N->getOperand(AddrOpIdx));
@@ -8676,8 +8815,7 @@ static SDValue CombineBaseUpdate(SDNode *N,
     }
     MemIntrinsicSDNode *MemInt = cast<MemIntrinsicSDNode>(N);
     SDValue UpdN = DAG.getMemIntrinsicNode(NewOpc, SDLoc(N), SDTys,
-                                           Ops.data(), Ops.size(),
-                                           MemInt->getMemoryVT(),
+                                           Ops, MemInt->getMemoryVT(),
                                            MemInt->getMemOperand());
 
     // Update the uses.
@@ -8746,11 +8884,11 @@ static bool CombineVLDDUP(SDNode *N, TargetLowering::DAGCombinerInfo &DCI) {
   for (n = 0; n < NumVecs; ++n)
     Tys[n] = VT;
   Tys[n] = MVT::Other;
-  SDVTList SDTys = DAG.getVTList(Tys, NumVecs+1);
+  SDVTList SDTys = DAG.getVTList(ArrayRef<EVT>(Tys, NumVecs+1));
   SDValue Ops[] = { VLD->getOperand(0), VLD->getOperand(2) };
   MemIntrinsicSDNode *VLDMemInt = cast<MemIntrinsicSDNode>(VLD);
   SDValue VLDDup = DAG.getMemIntrinsicNode(NewOpc, SDLoc(VLD), SDTys,
-                                           Ops, 2, VLDMemInt->getMemoryVT(),
+                                           Ops, VLDMemInt->getMemoryVT(),
                                            VLDMemInt->getMemOperand());
 
   // Update the uses.
@@ -9348,7 +9486,7 @@ ARMTargetLowering::PerformCMOVCombine(SDNode *N, SelectionDAG &DAG) const {
 
   if (Res.getNode()) {
     APInt KnownZero, KnownOne;
-    DAG.ComputeMaskedBits(SDValue(N,0), KnownZero, KnownOne);
+    DAG.computeKnownBits(SDValue(N,0), KnownZero, KnownOne);
     // Capture demanded bits information that would be otherwise lost.
     if (KnownZero == 0xfffffffe)
       Res = DAG.getNode(ISD::AssertZext, dl, MVT::i32, Res,
@@ -9935,11 +10073,11 @@ bool ARMTargetLowering::getPostIndexedAddressParts(SDNode *N, SDNode *Op,
   return true;
 }
 
-void ARMTargetLowering::computeMaskedBitsForTargetNode(const SDValue Op,
-                                                       APInt &KnownZero,
-                                                       APInt &KnownOne,
-                                                       const SelectionDAG &DAG,
-                                                       unsigned Depth) const {
+void ARMTargetLowering::computeKnownBitsForTargetNode(const SDValue Op,
+                                                      APInt &KnownZero,
+                                                      APInt &KnownOne,
+                                                      const SelectionDAG &DAG,
+                                                      unsigned Depth) const {
   unsigned BitWidth = KnownOne.getBitWidth();
   KnownZero = KnownOne = APInt(BitWidth, 0);
   switch (Op.getOpcode()) {
@@ -9955,11 +10093,11 @@ void ARMTargetLowering::computeMaskedBitsForTargetNode(const SDValue Op,
     break;
   case ARMISD::CMOV: {
     // Bits are known zero/one if known on the LHS and RHS.
-    DAG.ComputeMaskedBits(Op.getOperand(0), KnownZero, KnownOne, Depth+1);
+    DAG.computeKnownBits(Op.getOperand(0), KnownZero, KnownOne, Depth+1);
     if (KnownZero == 0 && KnownOne == 0) return;
 
     APInt KnownZeroRHS, KnownOneRHS;
-    DAG.ComputeMaskedBits(Op.getOperand(1), KnownZeroRHS, KnownOneRHS, Depth+1);
+    DAG.computeKnownBits(Op.getOperand(1), KnownZeroRHS, KnownOneRHS, Depth+1);
     KnownZero &= KnownZeroRHS;
     KnownOne  &= KnownOneRHS;
     return;
@@ -10053,7 +10191,7 @@ ARMTargetLowering::getSingleConstraintMatchWeight(
   Value *CallOperandVal = info.CallOperandVal;
     // If we don't have a value, we can't do a match,
     // but allow it at the lowest weight.
-  if (CallOperandVal == NULL)
+  if (!CallOperandVal)
     return CW_Default;
   Type *type = CallOperandVal->getType();
   // Look at the constraint type.
@@ -10132,7 +10270,7 @@ void ARMTargetLowering::LowerAsmOperandForConstraint(SDValue Op,
                                                      std::string &Constraint,
                                                      std::vector<SDValue>&Ops,
                                                      SelectionDAG &DAG) const {
-  SDValue Result(0, 0);
+  SDValue Result;
 
   // Currently only support length 1 constraints.
   if (Constraint.length() != 1) return;
@@ -10331,13 +10469,12 @@ SDValue ARMTargetLowering::LowerDivRem(SDValue Op, SelectionDAG &DAG) const {
   Type *RetTy = (Type*)StructType::get(Ty, Ty, NULL);
 
   SDLoc dl(Op);
-  TargetLowering::
-  CallLoweringInfo CLI(InChain, RetTy, isSigned, !isSigned, false, true,
-                    0, getLibcallCallingConv(LC), /*isTailCall=*/false,
-                    /*doesNotReturn=*/false, /*isReturnValueUsed=*/true,
-                    Callee, Args, DAG, dl);
-  std::pair<SDValue, SDValue> CallInfo = LowerCallTo(CLI);
+  TargetLowering::CallLoweringInfo CLI(DAG);
+  CLI.setDebugLoc(dl).setChain(InChain)
+    .setCallee(getLibcallCallingConv(LC), RetTy, Callee, &Args, 0)
+    .setInRegister().setSExtResult(isSigned).setZExtResult(!isSigned);
 
+  std::pair<SDValue, SDValue> CallInfo = LowerCallTo(CLI);
   return CallInfo.first;
 }
 
@@ -10494,3 +10631,160 @@ bool ARMTargetLowering::shouldConvertConstantLoadToIntImm(const APInt &Imm,
     return false;
   return true;
 }
+
+bool ARMTargetLowering::shouldExpandAtomicInIR(Instruction *Inst) const {
+  // Loads and stores less than 64-bits are already atomic; ones above that
+  // are doomed anyway, so defer to the default libcall and blame the OS when
+  // things go wrong:
+  if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
+    return SI->getValueOperand()->getType()->getPrimitiveSizeInBits() == 64;
+  else if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
+    return LI->getType()->getPrimitiveSizeInBits() == 64;
+
+  // For the real atomic operations, we have ldrex/strex up to 64 bits.
+  return Inst->getType()->getPrimitiveSizeInBits() <= 64;
+}
+
+Value *ARMTargetLowering::emitLoadLinked(IRBuilder<> &Builder, Value *Addr,
+                                         AtomicOrdering Ord) const {
+  Module *M = Builder.GetInsertBlock()->getParent()->getParent();
+  Type *ValTy = cast<PointerType>(Addr->getType())->getElementType();
+  bool IsAcquire =
+      Ord == Acquire || Ord == AcquireRelease || Ord == SequentiallyConsistent;
+
+  // Since i64 isn't legal and intrinsics don't get type-lowered, the ldrexd
+  // intrinsic must return {i32, i32} and we have to recombine them into a
+  // single i64 here.
+  if (ValTy->getPrimitiveSizeInBits() == 64) {
+    Intrinsic::ID Int =
+        IsAcquire ? Intrinsic::arm_ldaexd : Intrinsic::arm_ldrexd;
+    Function *Ldrex = llvm::Intrinsic::getDeclaration(M, Int);
+
+    Addr = Builder.CreateBitCast(Addr, Type::getInt8PtrTy(M->getContext()));
+    Value *LoHi = Builder.CreateCall(Ldrex, Addr, "lohi");
+
+    Value *Lo = Builder.CreateExtractValue(LoHi, 0, "lo");
+    Value *Hi = Builder.CreateExtractValue(LoHi, 1, "hi");
+    if (!Subtarget->isLittle())
+      std::swap (Lo, Hi);
+    Lo = Builder.CreateZExt(Lo, ValTy, "lo64");
+    Hi = Builder.CreateZExt(Hi, ValTy, "hi64");
+    return Builder.CreateOr(
+        Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 32)), "val64");
+  }
+
+  Type *Tys[] = { Addr->getType() };
+  Intrinsic::ID Int = IsAcquire ? Intrinsic::arm_ldaex : Intrinsic::arm_ldrex;
+  Function *Ldrex = llvm::Intrinsic::getDeclaration(M, Int, Tys);
+
+  return Builder.CreateTruncOrBitCast(
+      Builder.CreateCall(Ldrex, Addr),
+      cast<PointerType>(Addr->getType())->getElementType());
+}
+
+Value *ARMTargetLowering::emitStoreConditional(IRBuilder<> &Builder, Value *Val,
+                                               Value *Addr,
+                                               AtomicOrdering Ord) const {
+  Module *M = Builder.GetInsertBlock()->getParent()->getParent();
+  bool IsRelease =
+      Ord == Release || Ord == AcquireRelease || Ord == SequentiallyConsistent;
+
+  // Since the intrinsics must have legal type, the i64 intrinsics take two
+  // parameters: "i32, i32". We must marshal Val into the appropriate form
+  // before the call.
+  if (Val->getType()->getPrimitiveSizeInBits() == 64) {
+    Intrinsic::ID Int =
+        IsRelease ? Intrinsic::arm_stlexd : Intrinsic::arm_strexd;
+    Function *Strex = Intrinsic::getDeclaration(M, Int);
+    Type *Int32Ty = Type::getInt32Ty(M->getContext());
+
+    Value *Lo = Builder.CreateTrunc(Val, Int32Ty, "lo");
+    Value *Hi = Builder.CreateTrunc(Builder.CreateLShr(Val, 32), Int32Ty, "hi");
+    if (!Subtarget->isLittle())
+      std::swap (Lo, Hi);
+    Addr = Builder.CreateBitCast(Addr, Type::getInt8PtrTy(M->getContext()));
+    return Builder.CreateCall3(Strex, Lo, Hi, Addr);
+  }
+
+  Intrinsic::ID Int = IsRelease ? Intrinsic::arm_stlex : Intrinsic::arm_strex;
+  Type *Tys[] = { Addr->getType() };
+  Function *Strex = Intrinsic::getDeclaration(M, Int, Tys);
+
+  return Builder.CreateCall2(
+      Strex, Builder.CreateZExtOrBitCast(
+                 Val, Strex->getFunctionType()->getParamType(0)),
+      Addr);
+}
+
+enum HABaseType {
+  HA_UNKNOWN = 0,
+  HA_FLOAT,
+  HA_DOUBLE,
+  HA_VECT64,
+  HA_VECT128
+};
+
+static bool isHomogeneousAggregate(Type *Ty, HABaseType &Base,
+                                   uint64_t &Members) {
+  if (const StructType *ST = dyn_cast<StructType>(Ty)) {
+    for (unsigned i = 0; i < ST->getNumElements(); ++i) {
+      uint64_t SubMembers = 0;
+      if (!isHomogeneousAggregate(ST->getElementType(i), Base, SubMembers))
+        return false;
+      Members += SubMembers;
+    }
+  } else if (const ArrayType *AT = dyn_cast<ArrayType>(Ty)) {
+    uint64_t SubMembers = 0;
+    if (!isHomogeneousAggregate(AT->getElementType(), Base, SubMembers))
+      return false;
+    Members += SubMembers * AT->getNumElements();
+  } else if (Ty->isFloatTy()) {
+    if (Base != HA_UNKNOWN && Base != HA_FLOAT)
+      return false;
+    Members = 1;
+    Base = HA_FLOAT;
+  } else if (Ty->isDoubleTy()) {
+    if (Base != HA_UNKNOWN && Base != HA_DOUBLE)
+      return false;
+    Members = 1;
+    Base = HA_DOUBLE;
+  } else if (const VectorType *VT = dyn_cast<VectorType>(Ty)) {
+    Members = 1;
+    switch (Base) {
+    case HA_FLOAT:
+    case HA_DOUBLE:
+      return false;
+    case HA_VECT64:
+      return VT->getBitWidth() == 64;
+    case HA_VECT128:
+      return VT->getBitWidth() == 128;
+    case HA_UNKNOWN:
+      switch (VT->getBitWidth()) {
+      case 64:
+        Base = HA_VECT64;
+        return true;
+      case 128:
+        Base = HA_VECT128;
+        return true;
+      default:
+        return false;
+      }
+    }
+  }
+
+  return (Members > 0 && Members <= 4);
+}
+
+/// \brief Return true if a type is an AAPCS-VFP homogeneous aggregate.
+bool ARMTargetLowering::functionArgumentNeedsConsecutiveRegisters(
+    Type *Ty, CallingConv::ID CallConv, bool isVarArg) const {
+  if (getEffectiveCallingConv(CallConv, isVarArg) !=
+      CallingConv::ARM_AAPCS_VFP)
+    return false;
+
+  HABaseType Base = HA_UNKNOWN;
+  uint64_t Members = 0;
+  bool result = isHomogeneousAggregate(Ty, Base, Members);
+  DEBUG(dbgs() << "isHA: " << result << " "; Ty->dump(); dbgs() << "\n");
+  return result;
+}
diff --git a/lib/Target/ARM/ARMISelLowering.h b/lib/Target/ARM/ARMISelLowering.h
index f33e6db..c15305c 100644
--- a/lib/Target/ARM/ARMISelLowering.h
+++ b/lib/Target/ARM/ARMISelLowering.h
@@ -313,10 +313,10 @@ namespace llvm {
                                     SDValue &Offset, ISD::MemIndexedMode &AM,
                                     SelectionDAG &DAG) const override;
 
-    void computeMaskedBitsForTargetNode(const SDValue Op, APInt &KnownZero,
-                                        APInt &KnownOne,
-                                        const SelectionDAG &DAG,
-                                        unsigned Depth) const override;
+    void computeKnownBitsForTargetNode(const SDValue Op, APInt &KnownZero,
+                                       APInt &KnownOne,
+                                       const SelectionDAG &DAG,
+                                       unsigned Depth) const override;
 
 
     bool ExpandInlineAsm(CallInst *CI) const override;
@@ -384,6 +384,18 @@ namespace llvm {
     bool shouldConvertConstantLoadToIntImm(const APInt &Imm,
                                            Type *Ty) const override;
 
+    /// \brief Returns true if an argument of type Ty needs to be passed in a
+    /// contiguous block of registers in calling convention CallConv.
+    bool functionArgumentNeedsConsecutiveRegisters(
+        Type *Ty, CallingConv::ID CallConv, bool isVarArg) const override;
+
+    Value *emitLoadLinked(IRBuilder<> &Builder, Value *Addr,
+                          AtomicOrdering Ord) const override;
+    Value *emitStoreConditional(IRBuilder<> &Builder, Value *Val,
+                                Value *Addr, AtomicOrdering Ord) const override;
+
+    bool shouldExpandAtomicInIR(Instruction *Inst) const override;
+
   protected:
     std::pair<const TargetRegisterClass*, uint8_t>
     findRepresentativeClass(MVT VT) const override;
@@ -404,6 +416,7 @@ namespace llvm {
     void addTypeForNEON(MVT VT, MVT PromotedLdStVT, MVT PromotedBitwiseVT);
     void addDRTypeForNEON(MVT VT);
     void addQRTypeForNEON(MVT VT);
+    std::pair<SDValue, SDValue> getARMXALUOOp(SDValue Op, SelectionDAG &DAG, SDValue &ARMcc) const;
 
     typedef SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPassVector;
     void PassF64ArgInRegs(SDLoc dl, SelectionDAG &DAG,
@@ -417,6 +430,8 @@ namespace llvm {
                                  SDValue &Root, SelectionDAG &DAG,
                                  SDLoc dl) const;
 
+    CallingConv::ID getEffectiveCallingConv(CallingConv::ID CC,
+                                            bool isVarArg) const;
     CCAssignFn *CCAssignFnForNode(CallingConv::ID CC, bool Return,
                                   bool isVarArg) const;
     SDValue LowerMemOpCallTo(SDValue Chain, SDValue StackPtr, SDValue Arg,
@@ -430,6 +445,7 @@ namespace llvm {
     SDValue LowerBlockAddress(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerGlobalAddressDarwin(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerGlobalAddressELF(SDValue Op, SelectionDAG &DAG) const;
+    SDValue LowerGlobalAddressWindows(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerToTLSGeneralDynamicModel(GlobalAddressSDNode *GA,
                                             SelectionDAG &DAG) const;
@@ -438,6 +454,7 @@ namespace llvm {
                                  TLSModel::Model model) const;
     SDValue LowerGLOBAL_OFFSET_TABLE(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerBR_JT(SDValue Op, SelectionDAG &DAG) const;
+    SDValue LowerXALUO(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerSELECT(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerBR_CC(SDValue Op, SelectionDAG &DAG) const;
@@ -454,6 +471,8 @@ namespace llvm {
     SDValue LowerFSINCOS(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerDivRem(SDValue Op, SelectionDAG &DAG) const;
 
+    unsigned getRegisterByName(const char* RegName, EVT VT) const override;
+
     /// isFMAFasterThanFMulAndFAdd - Return true if an FMA operation is faster
     /// than a pair of fmul and fadd instructions. fmuladd intrinsics will be
     /// expanded to FMAs when this method returns true, otherwise fmuladd is
@@ -567,7 +586,6 @@ namespace llvm {
     OtherModImm
   };
 
-
   namespace ARM {
     FastISel *createFastISel(FunctionLoweringInfo &funcInfo,
                              const TargetLibraryInfo *libInfo);
diff --git a/lib/Target/ARM/ARMInstrFormats.td b/lib/Target/ARM/ARMInstrFormats.td
index aafff98..59e9260 100644
--- a/lib/Target/ARM/ARMInstrFormats.td
+++ b/lib/Target/ARM/ARMInstrFormats.td
@@ -2029,7 +2029,7 @@ class N2V<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18, bits<2> op17_16,
 // Same as N2V but not predicated.
 class N2Vnp<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op7, bit op6,
             dag oops, dag iops, InstrItinClass itin, string OpcodeStr,
-            string Dt, ValueType ResTy, ValueType OpTy, list<dag> pattern>
+            string Dt, list<dag> pattern>
    : NeonInp<oops, iops, AddrModeNone, IndexModeNone, N2RegFrm, itin,
              OpcodeStr, Dt, "$Vd, $Vm", "", pattern> {
   bits<5> Vd;
@@ -2138,8 +2138,7 @@ class N3V<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op6, bit op4,
 
 class N3Vnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 bit op4, dag oops, dag iops,Format f, InstrItinClass itin,
-                string OpcodeStr, string Dt, ValueType ResTy, ValueType OpTy,
-                SDPatternOperator IntOp, bit Commutable, list<dag> pattern>
+                string OpcodeStr, string Dt, list<dag> pattern>
   : NeonInp<oops, iops, AddrModeNone, IndexModeNone, f, itin, OpcodeStr,
             Dt, "$Vd, $Vn, $Vm", "", pattern> {
   bits<5> Vd;
diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td
index 75a109e..718d5da 100644
--- a/lib/Target/ARM/ARMInstrInfo.td
+++ b/lib/Target/ARM/ARMInstrInfo.td
@@ -991,6 +991,81 @@ def addrmode6oneL32 : Operand<i32>,
   let EncoderMethod = "getAddrMode6OneLane32AddressOpValue";
 }
 
+// Base class for addrmode6 with specific alignment restrictions.
+class AddrMode6Align : Operand<i32>,
+                ComplexPattern<i32, 2, "SelectAddrMode6", [], [SDNPWantParent]>{
+  let PrintMethod = "printAddrMode6Operand";
+  let MIOperandInfo = (ops GPR:$addr, i32imm:$align);
+  let EncoderMethod = "getAddrMode6AddressOpValue";
+  let DecoderMethod = "DecodeAddrMode6Operand";
+}
+
+// Special version of addrmode6 to handle no allowed alignment encoding for
+// VLD/VST instructions and checking the alignment is not specified.
+def AddrMode6AlignNoneAsmOperand : AsmOperandClass {
+  let Name = "AlignedMemoryNone";
+  let DiagnosticType = "AlignedMemoryRequiresNone";
+}
+def addrmode6alignNone : AddrMode6Align {
+  // The alignment specifier can only be omitted.
+  let ParserMatchClass = AddrMode6AlignNoneAsmOperand;
+}
+
+// Special version of addrmode6 to handle 16-bit alignment encoding for
+// VLD/VST instructions and checking the alignment value.
+def AddrMode6Align16AsmOperand : AsmOperandClass {
+  let Name = "AlignedMemory16";
+  let DiagnosticType = "AlignedMemoryRequires16";
+}
+def addrmode6align16 : AddrMode6Align {
+  // The alignment specifier can only be 16 or omitted.
+  let ParserMatchClass = AddrMode6Align16AsmOperand;
+}
+
+// Special version of addrmode6 to handle 32-bit alignment encoding for
+// VLD/VST instructions and checking the alignment value.
+def AddrMode6Align32AsmOperand : AsmOperandClass {
+  let Name = "AlignedMemory32";
+  let DiagnosticType = "AlignedMemoryRequires32";
+}
+def addrmode6align32 : AddrMode6Align {
+  // The alignment specifier can only be 32 or omitted.
+  let ParserMatchClass = AddrMode6Align32AsmOperand;
+}
+
+// Special version of addrmode6 to handle 64-bit alignment encoding for
+// VLD/VST instructions and checking the alignment value.
+def AddrMode6Align64AsmOperand : AsmOperandClass {
+  let Name = "AlignedMemory64";
+  let DiagnosticType = "AlignedMemoryRequires64";
+}
+def addrmode6align64 : AddrMode6Align {
+  // The alignment specifier can only be 64 or omitted.
+  let ParserMatchClass = AddrMode6Align64AsmOperand;
+}
+
+// Special version of addrmode6 to handle 64-bit or 128-bit alignment encoding
+// for VLD/VST instructions and checking the alignment value.
+def AddrMode6Align64or128AsmOperand : AsmOperandClass {
+  let Name = "AlignedMemory64or128";
+  let DiagnosticType = "AlignedMemoryRequires64or128";
+}
+def addrmode6align64or128 : AddrMode6Align {
+  // The alignment specifier can only be 64, 128 or omitted.
+  let ParserMatchClass = AddrMode6Align64or128AsmOperand;
+}
+
+// Special version of addrmode6 to handle 64-bit, 128-bit or 256-bit alignment
+// encoding for VLD/VST instructions and checking the alignment value.
+def AddrMode6Align64or128or256AsmOperand : AsmOperandClass {
+  let Name = "AlignedMemory64or128or256";
+  let DiagnosticType = "AlignedMemoryRequires64or128or256";
+}
+def addrmode6align64or128or256 : AddrMode6Align {
+  // The alignment specifier can only be 64, 128, 256 or omitted.
+  let ParserMatchClass = AddrMode6Align64or128or256AsmOperand;
+}
+
 // Special version of addrmode6 to handle alignment encoding for VLD-dup
 // instructions, specifically VLD4-dup.
 def addrmode6dup : Operand<i32>,
@@ -1003,6 +1078,69 @@ def addrmode6dup : Operand<i32>,
   let ParserMatchClass = AddrMode6AsmOperand;
 }
 
+// Base class for addrmode6dup with specific alignment restrictions.
+class AddrMode6DupAlign : Operand<i32>,
+                ComplexPattern<i32, 2, "SelectAddrMode6", [], [SDNPWantParent]>{
+  let PrintMethod = "printAddrMode6Operand";
+  let MIOperandInfo = (ops GPR:$addr, i32imm);
+  let EncoderMethod = "getAddrMode6DupAddressOpValue";
+}
+
+// Special version of addrmode6 to handle no allowed alignment encoding for
+// VLD-dup instruction and checking the alignment is not specified.
+def AddrMode6dupAlignNoneAsmOperand : AsmOperandClass {
+  let Name = "DupAlignedMemoryNone";
+  let DiagnosticType = "DupAlignedMemoryRequiresNone";
+}
+def addrmode6dupalignNone : AddrMode6DupAlign {
+  // The alignment specifier can only be omitted.
+  let ParserMatchClass = AddrMode6dupAlignNoneAsmOperand;
+}
+
+// Special version of addrmode6 to handle 16-bit alignment encoding for VLD-dup
+// instruction and checking the alignment value.
+def AddrMode6dupAlign16AsmOperand : AsmOperandClass {
+  let Name = "DupAlignedMemory16";
+  let DiagnosticType = "DupAlignedMemoryRequires16";
+}
+def addrmode6dupalign16 : AddrMode6DupAlign {
+  // The alignment specifier can only be 16 or omitted.
+  let ParserMatchClass = AddrMode6dupAlign16AsmOperand;
+}
+
+// Special version of addrmode6 to handle 32-bit alignment encoding for VLD-dup
+// instruction and checking the alignment value.
+def AddrMode6dupAlign32AsmOperand : AsmOperandClass {
+  let Name = "DupAlignedMemory32";
+  let DiagnosticType = "DupAlignedMemoryRequires32";
+}
+def addrmode6dupalign32 : AddrMode6DupAlign {
+  // The alignment specifier can only be 32 or omitted.
+  let ParserMatchClass = AddrMode6dupAlign32AsmOperand;
+}
+
+// Special version of addrmode6 to handle 64-bit alignment encoding for VLD
+// instructions and checking the alignment value.
+def AddrMode6dupAlign64AsmOperand : AsmOperandClass {
+  let Name = "DupAlignedMemory64";
+  let DiagnosticType = "DupAlignedMemoryRequires64";
+}
+def addrmode6dupalign64 : AddrMode6DupAlign {
+  // The alignment specifier can only be 64 or omitted.
+  let ParserMatchClass = AddrMode6dupAlign64AsmOperand;
+}
+
+// Special version of addrmode6 to handle 64-bit or 128-bit alignment encoding
+// for VLD instructions and checking the alignment value.
+def AddrMode6dupAlign64or128AsmOperand : AsmOperandClass {
+  let Name = "DupAlignedMemory64or128";
+  let DiagnosticType = "DupAlignedMemoryRequires64or128";
+}
+def addrmode6dupalign64or128 : AddrMode6DupAlign {
+  // The alignment specifier can only be 64, 128 or omitted.
+  let ParserMatchClass = AddrMode6dupAlign64or128AsmOperand;
+}
+
 // addrmodepc := pc + reg
 //
 def addrmodepc : Operand<i32>,
@@ -1689,7 +1827,8 @@ PseudoInst<(outs), (ins i32imm:$amt, pred:$p), NoItinerary,
 }
 
 def HINT : AI<(outs), (ins imm0_239:$imm), MiscFrm, NoItinerary,
-              "hint", "\t$imm", []>, Requires<[IsARM, HasV6]> {
+              "hint", "\t$imm", [(int_arm_hint imm0_239:$imm)]>,
+           Requires<[IsARM, HasV6]> {
   bits<8> imm;
   let Inst{27-8} = 0b00110010000011110000;
   let Inst{7-0} = imm;
@@ -1702,8 +1841,6 @@ def : InstAlias<"wfi$p", (HINT 3, pred:$p)>, Requires<[IsARM, HasV6T2]>;
 def : InstAlias<"sev$p", (HINT 4, pred:$p)>, Requires<[IsARM, HasV6T2]>;
 def : InstAlias<"sevl$p", (HINT 5, pred:$p)>, Requires<[IsARM, HasV8]>;
 
-def : Pat<(int_arm_sevl), (HINT 5)>;
-
 def SEL : AI<(outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), DPFrm, NoItinerary, "sel",
              "\t$Rd, $Rn, $Rm", []>, Requires<[IsARM, HasV6]> {
   bits<4> Rd;
@@ -1830,6 +1967,18 @@ def DBG : AI<(outs), (ins imm0_15:$opt), MiscFrm, NoItinerary, "dbg", "\t$opt",
   let Inst{3-0} = opt;
 }
 
+// A8.8.247  UDF - Undefined (Encoding A1)
+def UDF : AInoP<(outs), (ins imm0_65535:$imm16), MiscFrm, NoItinerary,
+                "udf", "\t$imm16", [(int_arm_undefined imm0_65535:$imm16)]> {
+  bits<16> imm16;
+  let Inst{31-28} = 0b1110; // AL
+  let Inst{27-25} = 0b011;
+  let Inst{24-20} = 0b11111;
+  let Inst{19-8} = imm16{15-4};
+  let Inst{7-4} = 0b1111;
+  let Inst{3-0} = imm16{3-0};
+}
+
 /*
  * A5.4 Permanently UNDEFINED instructions.
  *
@@ -2282,12 +2431,6 @@ let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
   def LDRD : AI3ld<0b1101, 0, (outs GPR:$Rt, GPR:$Rt2), (ins addrmode3:$addr),
                    LdMiscFrm, IIC_iLoad_d_r, "ldrd", "\t$Rt, $Rt2, $addr", []>,
              Requires<[IsARM, HasV5TE]>;
-
-  // GNU Assembler extension (compatibility)
-  let isAsmParserOnly = 1 in
-    def LDRD_PAIR : AI3ld<0b1101, 0, (outs GPRPairOp:$Rt), (ins addrmode3:$addr),
-                          LdMiscFrm, IIC_iLoad_d_r, "ldrd", "\t$Rt, $addr", []>,
-                    Requires<[IsARM, HasV5TE]>;
 }
 
 def LDA : AIldracq<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr),
@@ -2557,14 +2700,6 @@ let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in {
              Requires<[IsARM, HasV5TE]> {
     let Inst{21} = 0;
   }
-
-  // GNU Assembler extension (compatibility)
-  let isAsmParserOnly = 1 in
-    def STRD_PAIR : AI3str<0b1111, (outs), (ins GPRPairOp:$Rt, addrmode3:$addr),
-                           StMiscFrm, IIC_iStore_d_r, "strd", "\t$Rt, $addr", []>,
-                    Requires<[IsARM, HasV5TE]> {
-      let Inst{21} = 0;
-    }
 }
 
 // Indexed stores
@@ -3999,6 +4134,11 @@ def REV16 : AMiscA1I<0b01101011, 0b1011, (outs GPR:$Rd), (ins GPR:$Rm),
                Requires<[IsARM, HasV6]>,
            Sched<[WriteALU]>;
 
+def : ARMV6Pat<(srl (bswap (extloadi16 addrmode3:$addr)), (i32 16)),
+              (REV16 (LDRH addrmode3:$addr))>;
+def : ARMV6Pat<(truncstorei16 (srl (bswap GPR:$Rn), (i32 16)), addrmode3:$addr),
+               (STRH (REV16 GPR:$Rn), addrmode3:$addr)>;
+
 let AddedComplexity = 5 in
 def REVSH : AMiscA1I<0b01101111, 0b1011, (outs GPR:$Rd), (ins GPR:$Rm),
                IIC_iUNAr, "revsh", "\t$Rd, $Rm",
@@ -4816,7 +4956,7 @@ def MCR2 : MovRCopro2<"mcr2", 0 /* from ARM core register to coprocessor */,
                       [(int_arm_mcr2 imm:$cop, imm:$opc1, GPR:$Rt, imm:$CRn,
                                      imm:$CRm, imm:$opc2)]>,
                       Requires<[PreV8]>;
-def : ARMInstAlias<"mcr2$ $cop, $opc1, $Rt, $CRn, $CRm",
+def : ARMInstAlias<"mcr2 $cop, $opc1, $Rt, $CRn, $CRm",
                    (MCR2 p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn,
                          c_imm:$CRm, 0)>;
 def MRC2 : MovRCopro2<"mrc2", 1 /* from coprocessor to ARM core register */,
@@ -4824,7 +4964,7 @@ def MRC2 : MovRCopro2<"mrc2", 1 /* from coprocessor to ARM core register */,
                       (ins p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm,
                            imm0_7:$opc2), []>,
                       Requires<[PreV8]>;
-def : ARMInstAlias<"mrc2$ $cop, $opc1, $Rt, $CRn, $CRm",
+def : ARMInstAlias<"mrc2 $cop, $opc1, $Rt, $CRn, $CRm",
                    (MRC2 GPRwithAPSR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn,
                          c_imm:$CRm, 0)>;
 
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index 0d46c49..b32b5d2 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -39,6 +39,49 @@ def nImmVMOVI32 : Operand<i32> {
   let PrintMethod = "printNEONModImmOperand";
   let ParserMatchClass = nImmVMOVI32AsmOperand;
 }
+
+def nImmVMOVI16AsmOperandByteReplicate :
+  AsmOperandClass {
+  let Name = "NEONi16vmovByteReplicate";
+  let PredicateMethod = "isNEONi16ByteReplicate";
+  let RenderMethod = "addNEONvmovByteReplicateOperands";
+}
+def nImmVMOVI32AsmOperandByteReplicate :
+  AsmOperandClass {
+  let Name = "NEONi32vmovByteReplicate";
+  let PredicateMethod = "isNEONi32ByteReplicate";
+  let RenderMethod = "addNEONvmovByteReplicateOperands";
+}
+def nImmVMVNI16AsmOperandByteReplicate :
+  AsmOperandClass {
+  let Name = "NEONi16invByteReplicate";
+  let PredicateMethod = "isNEONi16ByteReplicate";
+  let RenderMethod = "addNEONinvByteReplicateOperands";
+}
+def nImmVMVNI32AsmOperandByteReplicate :
+  AsmOperandClass {
+  let Name = "NEONi32invByteReplicate";
+  let PredicateMethod = "isNEONi32ByteReplicate";
+  let RenderMethod = "addNEONinvByteReplicateOperands";
+}
+
+def nImmVMOVI16ByteReplicate : Operand<i32> {
+  let PrintMethod = "printNEONModImmOperand";
+  let ParserMatchClass = nImmVMOVI16AsmOperandByteReplicate;
+}
+def nImmVMOVI32ByteReplicate : Operand<i32> {
+  let PrintMethod = "printNEONModImmOperand";
+  let ParserMatchClass = nImmVMOVI32AsmOperandByteReplicate;
+}
+def nImmVMVNI16ByteReplicate : Operand<i32> {
+  let PrintMethod = "printNEONModImmOperand";
+  let ParserMatchClass = nImmVMVNI16AsmOperandByteReplicate;
+}
+def nImmVMVNI32ByteReplicate : Operand<i32> {
+  let PrintMethod = "printNEONModImmOperand";
+  let ParserMatchClass = nImmVMVNI32AsmOperandByteReplicate;
+}
+
 def nImmVMOVI32NegAsmOperand : AsmOperandClass { let Name = "NEONi32vmovNeg"; }
 def nImmVMOVI32Neg : Operand<i32> {
   let PrintMethod = "printNEONModImmOperand";
@@ -617,37 +660,37 @@ class VLDQQQQWBPseudo<InstrItinClass itin>
 let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
 
 //   VLD1     : Vector Load (multiple single elements)
-class VLD1D<bits<4> op7_4, string Dt>
+class VLD1D<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0,0b10,0b0111,op7_4, (outs VecListOneD:$Vd),
-          (ins addrmode6:$Rn), IIC_VLD1,
+          (ins AddrMode:$Rn), IIC_VLD1,
           "vld1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-class VLD1Q<bits<4> op7_4, string Dt>
+class VLD1Q<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd),
-          (ins addrmode6:$Rn), IIC_VLD1x2,
+          (ins AddrMode:$Rn), IIC_VLD1x2,
           "vld1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
 
-def  VLD1d8   : VLD1D<{0,0,0,?}, "8">;
-def  VLD1d16  : VLD1D<{0,1,0,?}, "16">;
-def  VLD1d32  : VLD1D<{1,0,0,?}, "32">;
-def  VLD1d64  : VLD1D<{1,1,0,?}, "64">;
+def  VLD1d8   : VLD1D<{0,0,0,?}, "8",  addrmode6align64>;
+def  VLD1d16  : VLD1D<{0,1,0,?}, "16", addrmode6align64>;
+def  VLD1d32  : VLD1D<{1,0,0,?}, "32", addrmode6align64>;
+def  VLD1d64  : VLD1D<{1,1,0,?}, "64", addrmode6align64>;
 
-def  VLD1q8   : VLD1Q<{0,0,?,?}, "8">;
-def  VLD1q16  : VLD1Q<{0,1,?,?}, "16">;
-def  VLD1q32  : VLD1Q<{1,0,?,?}, "32">;
-def  VLD1q64  : VLD1Q<{1,1,?,?}, "64">;
+def  VLD1q8   : VLD1Q<{0,0,?,?}, "8",  addrmode6align64or128>;
+def  VLD1q16  : VLD1Q<{0,1,?,?}, "16", addrmode6align64or128>;
+def  VLD1q32  : VLD1Q<{1,0,?,?}, "32", addrmode6align64or128>;
+def  VLD1q64  : VLD1Q<{1,1,?,?}, "64", addrmode6align64or128>;
 
 // ...with address register writeback:
-multiclass VLD1DWB<bits<4> op7_4, string Dt> {
+multiclass VLD1DWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b10, 0b0111,op7_4, (outs VecListOneD:$Vd, GPR:$wb),
-                     (ins addrmode6:$Rn), IIC_VLD1u,
+                     (ins AddrMode:$Rn), IIC_VLD1u,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -655,16 +698,16 @@ multiclass VLD1DWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b10,0b0111,op7_4, (outs VecListOneD:$Vd, GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1u,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1u,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
 }
-multiclass VLD1QWB<bits<4> op7_4, string Dt> {
+multiclass VLD1QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd, GPR:$wb),
-                    (ins addrmode6:$Rn), IIC_VLD1x2u,
+                    (ins AddrMode:$Rn), IIC_VLD1x2u,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -672,7 +715,7 @@ multiclass VLD1QWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd, GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
@@ -680,27 +723,27 @@ multiclass VLD1QWB<bits<4> op7_4, string Dt> {
   }
 }
 
-defm VLD1d8wb  : VLD1DWB<{0,0,0,?}, "8">;
-defm VLD1d16wb : VLD1DWB<{0,1,0,?}, "16">;
-defm VLD1d32wb : VLD1DWB<{1,0,0,?}, "32">;
-defm VLD1d64wb : VLD1DWB<{1,1,0,?}, "64">;
-defm VLD1q8wb  : VLD1QWB<{0,0,?,?}, "8">;
-defm VLD1q16wb : VLD1QWB<{0,1,?,?}, "16">;
-defm VLD1q32wb : VLD1QWB<{1,0,?,?}, "32">;
-defm VLD1q64wb : VLD1QWB<{1,1,?,?}, "64">;
+defm VLD1d8wb  : VLD1DWB<{0,0,0,?}, "8",  addrmode6align64>;
+defm VLD1d16wb : VLD1DWB<{0,1,0,?}, "16", addrmode6align64>;
+defm VLD1d32wb : VLD1DWB<{1,0,0,?}, "32", addrmode6align64>;
+defm VLD1d64wb : VLD1DWB<{1,1,0,?}, "64", addrmode6align64>;
+defm VLD1q8wb  : VLD1QWB<{0,0,?,?}, "8",  addrmode6align64or128>;
+defm VLD1q16wb : VLD1QWB<{0,1,?,?}, "16", addrmode6align64or128>;
+defm VLD1q32wb : VLD1QWB<{1,0,?,?}, "32", addrmode6align64or128>;
+defm VLD1q64wb : VLD1QWB<{1,1,?,?}, "64", addrmode6align64or128>;
 
 // ...with 3 registers
-class VLD1D3<bits<4> op7_4, string Dt>
+class VLD1D3<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0,0b10,0b0110,op7_4, (outs VecListThreeD:$Vd),
-          (ins addrmode6:$Rn), IIC_VLD1x3, "vld1", Dt,
+          (ins AddrMode:$Rn), IIC_VLD1x3, "vld1", Dt,
           "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
+multiclass VLD1D3WB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b10,0b0110, op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
-                    (ins addrmode6:$Rn), IIC_VLD1x2u,
+                    (ins AddrMode:$Rn), IIC_VLD1x2u,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -708,7 +751,7 @@ multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b10,0b0110,op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
@@ -716,32 +759,32 @@ multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
   }
 }
 
-def VLD1d8T      : VLD1D3<{0,0,0,?}, "8">;
-def VLD1d16T     : VLD1D3<{0,1,0,?}, "16">;
-def VLD1d32T     : VLD1D3<{1,0,0,?}, "32">;
-def VLD1d64T     : VLD1D3<{1,1,0,?}, "64">;
+def VLD1d8T      : VLD1D3<{0,0,0,?}, "8",  addrmode6align64>;
+def VLD1d16T     : VLD1D3<{0,1,0,?}, "16", addrmode6align64>;
+def VLD1d32T     : VLD1D3<{1,0,0,?}, "32", addrmode6align64>;
+def VLD1d64T     : VLD1D3<{1,1,0,?}, "64", addrmode6align64>;
 
-defm VLD1d8Twb  : VLD1D3WB<{0,0,0,?}, "8">;
-defm VLD1d16Twb : VLD1D3WB<{0,1,0,?}, "16">;
-defm VLD1d32Twb : VLD1D3WB<{1,0,0,?}, "32">;
-defm VLD1d64Twb : VLD1D3WB<{1,1,0,?}, "64">;
+defm VLD1d8Twb  : VLD1D3WB<{0,0,0,?}, "8",  addrmode6align64>;
+defm VLD1d16Twb : VLD1D3WB<{0,1,0,?}, "16", addrmode6align64>;
+defm VLD1d32Twb : VLD1D3WB<{1,0,0,?}, "32", addrmode6align64>;
+defm VLD1d64Twb : VLD1D3WB<{1,1,0,?}, "64", addrmode6align64>;
 
 def VLD1d64TPseudo : VLDQQPseudo<IIC_VLD1x3>;
 def VLD1d64TPseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD1x3>;
 def VLD1d64TPseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD1x3>;
 
 // ...with 4 registers
-class VLD1D4<bits<4> op7_4, string Dt>
+class VLD1D4<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0, 0b10, 0b0010, op7_4, (outs VecListFourD:$Vd),
-          (ins addrmode6:$Rn), IIC_VLD1x4, "vld1", Dt,
+          (ins AddrMode:$Rn), IIC_VLD1x4, "vld1", Dt,
           "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
+multiclass VLD1D4WB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b10,0b0010, op7_4, (outs VecListFourD:$Vd, GPR:$wb),
-                    (ins addrmode6:$Rn), IIC_VLD1x2u,
+                    (ins AddrMode:$Rn), IIC_VLD1x2u,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -749,7 +792,7 @@ multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b10,0b0010,op7_4, (outs VecListFourD:$Vd, GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
@@ -757,15 +800,15 @@ multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
   }
 }
 
-def VLD1d8Q      : VLD1D4<{0,0,?,?}, "8">;
-def VLD1d16Q     : VLD1D4<{0,1,?,?}, "16">;
-def VLD1d32Q     : VLD1D4<{1,0,?,?}, "32">;
-def VLD1d64Q     : VLD1D4<{1,1,?,?}, "64">;
+def VLD1d8Q      : VLD1D4<{0,0,?,?}, "8",  addrmode6align64or128or256>;
+def VLD1d16Q     : VLD1D4<{0,1,?,?}, "16", addrmode6align64or128or256>;
+def VLD1d32Q     : VLD1D4<{1,0,?,?}, "32", addrmode6align64or128or256>;
+def VLD1d64Q     : VLD1D4<{1,1,?,?}, "64", addrmode6align64or128or256>;
 
-defm VLD1d8Qwb   : VLD1D4WB<{0,0,?,?}, "8">;
-defm VLD1d16Qwb  : VLD1D4WB<{0,1,?,?}, "16">;
-defm VLD1d32Qwb  : VLD1D4WB<{1,0,?,?}, "32">;
-defm VLD1d64Qwb  : VLD1D4WB<{1,1,?,?}, "64">;
+defm VLD1d8Qwb   : VLD1D4WB<{0,0,?,?}, "8",  addrmode6align64or128or256>;
+defm VLD1d16Qwb  : VLD1D4WB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VLD1d32Qwb  : VLD1D4WB<{1,0,?,?}, "32", addrmode6align64or128or256>;
+defm VLD1d64Qwb  : VLD1D4WB<{1,1,?,?}, "64", addrmode6align64or128or256>;
 
 def VLD1d64QPseudo : VLDQQPseudo<IIC_VLD1x4>;
 def VLD1d64QPseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD1x4>;
@@ -773,22 +816,28 @@ def VLD1d64QPseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD1x4>;
 
 //   VLD2     : Vector Load (multiple 2-element structures)
 class VLD2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
-           InstrItinClass itin>
+           InstrItinClass itin, Operand AddrMode>
   : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd),
-          (ins addrmode6:$Rn), itin,
+          (ins AddrMode:$Rn), itin,
           "vld2", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST2Instruction";
 }
 
-def  VLD2d8   : VLD2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2>;
-def  VLD2d16  : VLD2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2>;
-def  VLD2d32  : VLD2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2>;
+def  VLD2d8   : VLD2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2,
+                     addrmode6align64or128>;
+def  VLD2d16  : VLD2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2,
+                     addrmode6align64or128>;
+def  VLD2d32  : VLD2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2,
+                     addrmode6align64or128>;
 
-def  VLD2q8   : VLD2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2>;
-def  VLD2q16  : VLD2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2>;
-def  VLD2q32  : VLD2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2>;
+def  VLD2q8   : VLD2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2,
+                     addrmode6align64or128or256>;
+def  VLD2q16  : VLD2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2,
+                     addrmode6align64or128or256>;
+def  VLD2q32  : VLD2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2,
+                     addrmode6align64or128or256>;
 
 def  VLD2q8Pseudo  : VLDQQPseudo<IIC_VLD2x2>;
 def  VLD2q16Pseudo : VLDQQPseudo<IIC_VLD2x2>;
@@ -796,9 +845,9 @@ def  VLD2q32Pseudo : VLDQQPseudo<IIC_VLD2x2>;
 
 // ...with address register writeback:
 multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
-                  RegisterOperand VdTy, InstrItinClass itin> {
+                  RegisterOperand VdTy, InstrItinClass itin, Operand AddrMode> {
   def _fixed : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
-                     (ins addrmode6:$Rn), itin,
+                     (ins AddrMode:$Rn), itin,
                      "vld2", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -806,7 +855,7 @@ multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
     let DecoderMethod = "DecodeVLDST2Instruction";
   }
   def _register : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm), itin,
+                        (ins AddrMode:$Rn, rGPR:$Rm), itin,
                         "vld2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
@@ -814,13 +863,19 @@ multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
   }
 }
 
-defm VLD2d8wb  : VLD2WB<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2u>;
-defm VLD2d16wb : VLD2WB<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2u>;
-defm VLD2d32wb : VLD2WB<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2u>;
+defm VLD2d8wb  : VLD2WB<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2u,
+                        addrmode6align64or128>;
+defm VLD2d16wb : VLD2WB<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2u,
+                        addrmode6align64or128>;
+defm VLD2d32wb : VLD2WB<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2u,
+                        addrmode6align64or128>;
 
-defm VLD2q8wb  : VLD2WB<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2u>;
-defm VLD2q16wb : VLD2WB<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2u>;
-defm VLD2q32wb : VLD2WB<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2u>;
+defm VLD2q8wb  : VLD2WB<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2u,
+                        addrmode6align64or128or256>;
+defm VLD2q16wb : VLD2WB<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2u,
+                        addrmode6align64or128or256>;
+defm VLD2q32wb : VLD2WB<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2u,
+                        addrmode6align64or128or256>;
 
 def VLD2q8PseudoWB_fixed     : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
 def VLD2q16PseudoWB_fixed    : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
@@ -830,12 +885,18 @@ def VLD2q16PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
 def VLD2q32PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
 
 // ...with double-spaced registers
-def  VLD2b8    : VLD2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2>;
-def  VLD2b16   : VLD2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2>;
-def  VLD2b32   : VLD2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2>;
-defm VLD2b8wb  : VLD2WB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2u>;
-defm VLD2b16wb : VLD2WB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2u>;
-defm VLD2b32wb : VLD2WB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2u>;
+def  VLD2b8    : VLD2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2,
+                      addrmode6align64or128>;
+def  VLD2b16   : VLD2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2,
+                      addrmode6align64or128>;
+def  VLD2b32   : VLD2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2,
+                      addrmode6align64or128>;
+defm VLD2b8wb  : VLD2WB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2u,
+                        addrmode6align64or128>;
+defm VLD2b16wb : VLD2WB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2u,
+                        addrmode6align64or128>;
+defm VLD2b32wb : VLD2WB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2u,
+                        addrmode6align64or128>;
 
 //   VLD3     : Vector Load (multiple 3-element structures)
 class VLD3D<bits<4> op11_8, bits<4> op7_4, string Dt>
@@ -1293,47 +1354,55 @@ def VLD4LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4lnu>;
 } // mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1
 
 //   VLD1DUP  : Vector Load (single element to all lanes)
-class VLD1DUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp>
+class VLD1DUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp,
+              Operand AddrMode>
   : NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListOneDAllLanes:$Vd),
-          (ins addrmode6dup:$Rn),
+          (ins AddrMode:$Rn),
           IIC_VLD1dup, "vld1", Dt, "$Vd, $Rn", "",
           [(set VecListOneDAllLanes:$Vd,
-                (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
+                (Ty (NEONvdup (i32 (LoadOp AddrMode:$Rn)))))]> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLD1DupInstruction";
 }
-def VLD1DUPd8  : VLD1DUP<{0,0,0,?}, "8", v8i8, extloadi8>;
-def VLD1DUPd16 : VLD1DUP<{0,1,0,?}, "16", v4i16, extloadi16>;
-def VLD1DUPd32 : VLD1DUP<{1,0,0,?}, "32", v2i32, load>;
+def VLD1DUPd8  : VLD1DUP<{0,0,0,?}, "8", v8i8, extloadi8,
+                         addrmode6dupalignNone>;
+def VLD1DUPd16 : VLD1DUP<{0,1,0,?}, "16", v4i16, extloadi16,
+                         addrmode6dupalign16>;
+def VLD1DUPd32 : VLD1DUP<{1,0,0,?}, "32", v2i32, load,
+                         addrmode6dupalign32>;
 
 def : Pat<(v2f32 (NEONvdup (f32 (load addrmode6dup:$addr)))),
           (VLD1DUPd32 addrmode6:$addr)>;
 
-class VLD1QDUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp>
+class VLD1QDUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp,
+               Operand AddrMode>
   : NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListDPairAllLanes:$Vd),
-          (ins addrmode6dup:$Rn), IIC_VLD1dup,
+          (ins AddrMode:$Rn), IIC_VLD1dup,
           "vld1", Dt, "$Vd, $Rn", "",
           [(set VecListDPairAllLanes:$Vd,
-                (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
+                (Ty (NEONvdup (i32 (LoadOp AddrMode:$Rn)))))]> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLD1DupInstruction";
 }
 
-def VLD1DUPq8  : VLD1QDUP<{0,0,1,0}, "8", v16i8, extloadi8>;
-def VLD1DUPq16 : VLD1QDUP<{0,1,1,?}, "16", v8i16, extloadi16>;
-def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32", v4i32, load>;
+def VLD1DUPq8  : VLD1QDUP<{0,0,1,0}, "8", v16i8, extloadi8,
+                          addrmode6dupalignNone>;
+def VLD1DUPq16 : VLD1QDUP<{0,1,1,?}, "16", v8i16, extloadi16,
+                          addrmode6dupalign16>;
+def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32", v4i32, load,
+                          addrmode6dupalign32>;
 
 def : Pat<(v4f32 (NEONvdup (f32 (load addrmode6dup:$addr)))),
           (VLD1DUPq32 addrmode6:$addr)>;
 
 let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
 // ...with address register writeback:
-multiclass VLD1DUPWB<bits<4> op7_4, string Dt> {
+multiclass VLD1DUPWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
                      (outs VecListOneDAllLanes:$Vd, GPR:$wb),
-                     (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+                     (ins AddrMode:$Rn), IIC_VLD1dupu,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1342,17 +1411,17 @@ multiclass VLD1DUPWB<bits<4> op7_4, string Dt> {
   }
   def _register : NLdSt<1, 0b10, 0b1100, op7_4,
                         (outs VecListOneDAllLanes:$Vd, GPR:$wb),
-                        (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1dupu,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
     let DecoderMethod = "DecodeVLD1DupInstruction";
   }
 }
-multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
+multiclass VLD1QDUPWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
                      (outs VecListDPairAllLanes:$Vd, GPR:$wb),
-                     (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+                     (ins AddrMode:$Rn), IIC_VLD1dupu,
                      "vld1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1361,7 +1430,7 @@ multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
   }
   def _register : NLdSt<1, 0b10, 0b1100, op7_4,
                         (outs VecListDPairAllLanes:$Vd, GPR:$wb),
-                        (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1dupu,
                         "vld1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
@@ -1369,38 +1438,47 @@ multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
   }
 }
 
-defm VLD1DUPd8wb  : VLD1DUPWB<{0,0,0,0}, "8">;
-defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16">;
-defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32">;
+defm VLD1DUPd8wb  : VLD1DUPWB<{0,0,0,0}, "8", addrmode6dupalignNone>;
+defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16", addrmode6dupalign16>;
+defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32", addrmode6dupalign32>;
 
-defm VLD1DUPq8wb  : VLD1QDUPWB<{0,0,1,0}, "8">;
-defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16">;
-defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32">;
+defm VLD1DUPq8wb  : VLD1QDUPWB<{0,0,1,0}, "8", addrmode6dupalignNone>;
+defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16", addrmode6dupalign16>;
+defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32", addrmode6dupalign32>;
 
 //   VLD2DUP  : Vector Load (single 2-element structure to all lanes)
-class VLD2DUP<bits<4> op7_4, string Dt, RegisterOperand VdTy>
+class VLD2DUP<bits<4> op7_4, string Dt, RegisterOperand VdTy, Operand AddrMode>
   : NLdSt<1, 0b10, 0b1101, op7_4, (outs VdTy:$Vd),
-          (ins addrmode6dup:$Rn), IIC_VLD2dup,
+          (ins AddrMode:$Rn), IIC_VLD2dup,
           "vld2", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLD2DupInstruction";
 }
 
-def VLD2DUPd8  : VLD2DUP<{0,0,0,?}, "8",  VecListDPairAllLanes>;
-def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16", VecListDPairAllLanes>;
-def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32", VecListDPairAllLanes>;
+def VLD2DUPd8  : VLD2DUP<{0,0,0,?}, "8",  VecListDPairAllLanes,
+                         addrmode6dupalign16>;
+def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16", VecListDPairAllLanes,
+                         addrmode6dupalign32>;
+def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32", VecListDPairAllLanes,
+                         addrmode6dupalign64>;
 
+// HACK this one, VLD2DUPd8x2 must be changed at the same time with VLD2b8 or
+// "vld2.8 {d0[], d2[]}, [r4:32]" will become "vld2.8 {d0, d2}, [r4:32]".
 // ...with double-spaced registers
-def VLD2DUPd8x2  : VLD2DUP<{0,0,1,?}, "8",  VecListDPairSpacedAllLanes>;
-def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16", VecListDPairSpacedAllLanes>;
-def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32", VecListDPairSpacedAllLanes>;
+def VLD2DUPd8x2  : VLD2DUP<{0,0,1,?}, "8",  VecListDPairSpacedAllLanes,
+                           addrmode6dupalign16>;
+def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16", VecListDPairSpacedAllLanes,
+                           addrmode6dupalign32>;
+def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32", VecListDPairSpacedAllLanes,
+                           addrmode6dupalign64>;
 
 // ...with address register writeback:
-multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy> {
+multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy,
+                     Operand AddrMode> {
   def _fixed : NLdSt<1, 0b10, 0b1101, op7_4,
                      (outs VdTy:$Vd, GPR:$wb),
-                     (ins addrmode6dup:$Rn), IIC_VLD2dupu,
+                     (ins AddrMode:$Rn), IIC_VLD2dupu,
                      "vld2", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1409,7 +1487,7 @@ multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy> {
   }
   def _register : NLdSt<1, 0b10, 0b1101, op7_4,
                         (outs VdTy:$Vd, GPR:$wb),
-                        (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD2dupu,
+                        (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD2dupu,
                         "vld2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{4} = Rn{4};
@@ -1417,13 +1495,19 @@ multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy> {
   }
 }
 
-defm VLD2DUPd8wb    : VLD2DUPWB<{0,0,0,0}, "8",  VecListDPairAllLanes>;
-defm VLD2DUPd16wb   : VLD2DUPWB<{0,1,0,?}, "16", VecListDPairAllLanes>;
-defm VLD2DUPd32wb   : VLD2DUPWB<{1,0,0,?}, "32", VecListDPairAllLanes>;
+defm VLD2DUPd8wb    : VLD2DUPWB<{0,0,0,0}, "8",  VecListDPairAllLanes,
+                                addrmode6dupalign16>;
+defm VLD2DUPd16wb   : VLD2DUPWB<{0,1,0,?}, "16", VecListDPairAllLanes,
+                                addrmode6dupalign32>;
+defm VLD2DUPd32wb   : VLD2DUPWB<{1,0,0,?}, "32", VecListDPairAllLanes,
+                                addrmode6dupalign64>;
 
-defm VLD2DUPd8x2wb  : VLD2DUPWB<{0,0,1,0}, "8",  VecListDPairSpacedAllLanes>;
-defm VLD2DUPd16x2wb : VLD2DUPWB<{0,1,1,?}, "16", VecListDPairSpacedAllLanes>;
-defm VLD2DUPd32x2wb : VLD2DUPWB<{1,0,1,?}, "32", VecListDPairSpacedAllLanes>;
+defm VLD2DUPd8x2wb  : VLD2DUPWB<{0,0,1,0}, "8",  VecListDPairSpacedAllLanes,
+                                addrmode6dupalign16>;
+defm VLD2DUPd16x2wb : VLD2DUPWB<{0,1,1,?}, "16", VecListDPairSpacedAllLanes,
+                                addrmode6dupalign32>;
+defm VLD2DUPd32x2wb : VLD2DUPWB<{1,0,1,?}, "32", VecListDPairSpacedAllLanes,
+                                addrmode6dupalign64>;
 
 //   VLD3DUP  : Vector Load (single 3-element structure to all lanes)
 class VLD3DUP<bits<4> op7_4, string Dt>
@@ -1449,22 +1533,22 @@ def VLD3DUPq16 : VLD3DUP<{0,1,1,?}, "16">;
 def VLD3DUPq32 : VLD3DUP<{1,0,1,?}, "32">;
 
 // ...with address register writeback:
-class VLD3DUPWB<bits<4> op7_4, string Dt>
+class VLD3DUPWB<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<1, 0b10, 0b1110, op7_4, (outs DPR:$Vd, DPR:$dst2, DPR:$dst3, GPR:$wb),
-          (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD3dupu,
+          (ins AddrMode:$Rn, am6offset:$Rm), IIC_VLD3dupu,
           "vld3", Dt, "\\{$Vd[], $dst2[], $dst3[]\\}, $Rn$Rm",
           "$Rn.addr = $wb", []> {
   let Inst{4} = 0;
   let DecoderMethod = "DecodeVLD3DupInstruction";
 }
 
-def VLD3DUPd8_UPD  : VLD3DUPWB<{0,0,0,0}, "8">;
-def VLD3DUPd16_UPD : VLD3DUPWB<{0,1,0,?}, "16">;
-def VLD3DUPd32_UPD : VLD3DUPWB<{1,0,0,?}, "32">;
+def VLD3DUPd8_UPD  : VLD3DUPWB<{0,0,0,0}, "8",  addrmode6dupalign64>;
+def VLD3DUPd16_UPD : VLD3DUPWB<{0,1,0,?}, "16", addrmode6dupalign64>;
+def VLD3DUPd32_UPD : VLD3DUPWB<{1,0,0,?}, "32", addrmode6dupalign64>;
 
-def VLD3DUPq8_UPD  : VLD3DUPWB<{0,0,1,0}, "8">;
-def VLD3DUPq16_UPD : VLD3DUPWB<{0,1,1,?}, "16">;
-def VLD3DUPq32_UPD : VLD3DUPWB<{1,0,1,?}, "32">;
+def VLD3DUPq8_UPD  : VLD3DUPWB<{0,0,1,0}, "8",  addrmode6dupalign64>;
+def VLD3DUPq16_UPD : VLD3DUPWB<{0,1,1,?}, "16", addrmode6dupalign64>;
+def VLD3DUPq32_UPD : VLD3DUPWB<{1,0,1,?}, "32", addrmode6dupalign64>;
 
 def VLD3DUPd8Pseudo_UPD  : VLDQQWBPseudo<IIC_VLD3dupu>;
 def VLD3DUPd16Pseudo_UPD : VLDQQWBPseudo<IIC_VLD3dupu>;
@@ -1560,35 +1644,35 @@ class VSTQQQQWBPseudo<InstrItinClass itin>
                 "$addr.addr = $wb">;
 
 //   VST1     : Vector Store (multiple single elements)
-class VST1D<bits<4> op7_4, string Dt>
-  : NLdSt<0,0b00,0b0111,op7_4, (outs), (ins addrmode6:$Rn, VecListOneD:$Vd),
+class VST1D<bits<4> op7_4, string Dt, Operand AddrMode>
+  : NLdSt<0,0b00,0b0111,op7_4, (outs), (ins AddrMode:$Rn, VecListOneD:$Vd),
           IIC_VST1, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-class VST1Q<bits<4> op7_4, string Dt>
-  : NLdSt<0,0b00,0b1010,op7_4, (outs), (ins addrmode6:$Rn, VecListDPair:$Vd),
+class VST1Q<bits<4> op7_4, string Dt, Operand AddrMode>
+  : NLdSt<0,0b00,0b1010,op7_4, (outs), (ins AddrMode:$Rn, VecListDPair:$Vd),
           IIC_VST1x2, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
 
-def  VST1d8   : VST1D<{0,0,0,?}, "8">;
-def  VST1d16  : VST1D<{0,1,0,?}, "16">;
-def  VST1d32  : VST1D<{1,0,0,?}, "32">;
-def  VST1d64  : VST1D<{1,1,0,?}, "64">;
+def  VST1d8   : VST1D<{0,0,0,?}, "8",  addrmode6align64>;
+def  VST1d16  : VST1D<{0,1,0,?}, "16", addrmode6align64>;
+def  VST1d32  : VST1D<{1,0,0,?}, "32", addrmode6align64>;
+def  VST1d64  : VST1D<{1,1,0,?}, "64", addrmode6align64>;
 
-def  VST1q8   : VST1Q<{0,0,?,?}, "8">;
-def  VST1q16  : VST1Q<{0,1,?,?}, "16">;
-def  VST1q32  : VST1Q<{1,0,?,?}, "32">;
-def  VST1q64  : VST1Q<{1,1,?,?}, "64">;
+def  VST1q8   : VST1Q<{0,0,?,?}, "8",  addrmode6align64or128>;
+def  VST1q16  : VST1Q<{0,1,?,?}, "16", addrmode6align64or128>;
+def  VST1q32  : VST1Q<{1,0,?,?}, "32", addrmode6align64or128>;
+def  VST1q64  : VST1Q<{1,1,?,?}, "64", addrmode6align64or128>;
 
 // ...with address register writeback:
-multiclass VST1DWB<bits<4> op7_4, string Dt> {
+multiclass VST1DWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b00, 0b0111,op7_4, (outs GPR:$wb),
-                     (ins addrmode6:$Rn, VecListOneD:$Vd), IIC_VLD1u,
+                     (ins AddrMode:$Rn, VecListOneD:$Vd), IIC_VLD1u,
                      "vst1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1596,7 +1680,7 @@ multiclass VST1DWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b00,0b0111,op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VecListOneD:$Vd),
+                        (ins AddrMode:$Rn, rGPR:$Rm, VecListOneD:$Vd),
                         IIC_VLD1u,
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
@@ -1604,9 +1688,9 @@ multiclass VST1DWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
 }
-multiclass VST1QWB<bits<4> op7_4, string Dt> {
+multiclass VST1QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b00,0b1010,op7_4, (outs GPR:$wb),
-                    (ins addrmode6:$Rn, VecListDPair:$Vd), IIC_VLD1x2u,
+                    (ins AddrMode:$Rn, VecListDPair:$Vd), IIC_VLD1x2u,
                      "vst1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1614,7 +1698,7 @@ multiclass VST1QWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b00,0b1010,op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VecListDPair:$Vd),
+                        (ins AddrMode:$Rn, rGPR:$Rm, VecListDPair:$Vd),
                         IIC_VLD1x2u,
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
@@ -1623,28 +1707,28 @@ multiclass VST1QWB<bits<4> op7_4, string Dt> {
   }
 }
 
-defm VST1d8wb  : VST1DWB<{0,0,0,?}, "8">;
-defm VST1d16wb : VST1DWB<{0,1,0,?}, "16">;
-defm VST1d32wb : VST1DWB<{1,0,0,?}, "32">;
-defm VST1d64wb : VST1DWB<{1,1,0,?}, "64">;
+defm VST1d8wb  : VST1DWB<{0,0,0,?}, "8",  addrmode6align64>;
+defm VST1d16wb : VST1DWB<{0,1,0,?}, "16", addrmode6align64>;
+defm VST1d32wb : VST1DWB<{1,0,0,?}, "32", addrmode6align64>;
+defm VST1d64wb : VST1DWB<{1,1,0,?}, "64", addrmode6align64>;
 
-defm VST1q8wb  : VST1QWB<{0,0,?,?}, "8">;
-defm VST1q16wb : VST1QWB<{0,1,?,?}, "16">;
-defm VST1q32wb : VST1QWB<{1,0,?,?}, "32">;
-defm VST1q64wb : VST1QWB<{1,1,?,?}, "64">;
+defm VST1q8wb  : VST1QWB<{0,0,?,?}, "8",  addrmode6align64or128>;
+defm VST1q16wb : VST1QWB<{0,1,?,?}, "16", addrmode6align64or128>;
+defm VST1q32wb : VST1QWB<{1,0,?,?}, "32", addrmode6align64or128>;
+defm VST1q64wb : VST1QWB<{1,1,?,?}, "64", addrmode6align64or128>;
 
 // ...with 3 registers
-class VST1D3<bits<4> op7_4, string Dt>
+class VST1D3<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0, 0b00, 0b0110, op7_4, (outs),
-          (ins addrmode6:$Rn, VecListThreeD:$Vd),
+          (ins AddrMode:$Rn, VecListThreeD:$Vd),
           IIC_VST1x3, "vst1", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{4} = Rn{4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-multiclass VST1D3WB<bits<4> op7_4, string Dt> {
+multiclass VST1D3WB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
-                    (ins addrmode6:$Rn, VecListThreeD:$Vd), IIC_VLD1x3u,
+                    (ins AddrMode:$Rn, VecListThreeD:$Vd), IIC_VLD1x3u,
                      "vst1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1652,7 +1736,7 @@ multiclass VST1D3WB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VecListThreeD:$Vd),
+                        (ins AddrMode:$Rn, rGPR:$Rm, VecListThreeD:$Vd),
                         IIC_VLD1x3u,
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
@@ -1661,33 +1745,33 @@ multiclass VST1D3WB<bits<4> op7_4, string Dt> {
   }
 }
 
-def VST1d8T     : VST1D3<{0,0,0,?}, "8">;
-def VST1d16T    : VST1D3<{0,1,0,?}, "16">;
-def VST1d32T    : VST1D3<{1,0,0,?}, "32">;
-def VST1d64T    : VST1D3<{1,1,0,?}, "64">;
+def VST1d8T     : VST1D3<{0,0,0,?}, "8",  addrmode6align64>;
+def VST1d16T    : VST1D3<{0,1,0,?}, "16", addrmode6align64>;
+def VST1d32T    : VST1D3<{1,0,0,?}, "32", addrmode6align64>;
+def VST1d64T    : VST1D3<{1,1,0,?}, "64", addrmode6align64>;
 
-defm VST1d8Twb  : VST1D3WB<{0,0,0,?}, "8">;
-defm VST1d16Twb : VST1D3WB<{0,1,0,?}, "16">;
-defm VST1d32Twb : VST1D3WB<{1,0,0,?}, "32">;
-defm VST1d64Twb : VST1D3WB<{1,1,0,?}, "64">;
+defm VST1d8Twb  : VST1D3WB<{0,0,0,?}, "8",  addrmode6align64>;
+defm VST1d16Twb : VST1D3WB<{0,1,0,?}, "16", addrmode6align64>;
+defm VST1d32Twb : VST1D3WB<{1,0,0,?}, "32", addrmode6align64>;
+defm VST1d64Twb : VST1D3WB<{1,1,0,?}, "64", addrmode6align64>;
 
 def VST1d64TPseudo            : VSTQQPseudo<IIC_VST1x3>;
 def VST1d64TPseudoWB_fixed    : VSTQQWBfixedPseudo<IIC_VST1x3u>;
 def VST1d64TPseudoWB_register : VSTQQWBPseudo<IIC_VST1x3u>;
 
 // ...with 4 registers
-class VST1D4<bits<4> op7_4, string Dt>
+class VST1D4<bits<4> op7_4, string Dt, Operand AddrMode>
   : NLdSt<0, 0b00, 0b0010, op7_4, (outs),
-          (ins addrmode6:$Rn, VecListFourD:$Vd),
+          (ins AddrMode:$Rn, VecListFourD:$Vd),
           IIC_VST1x4, "vst1", Dt, "$Vd, $Rn", "",
           []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST1Instruction";
 }
-multiclass VST1D4WB<bits<4> op7_4, string Dt> {
+multiclass VST1D4WB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
-                    (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1x4u,
+                    (ins AddrMode:$Rn, VecListFourD:$Vd), IIC_VLD1x4u,
                      "vst1", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1695,7 +1779,7 @@ multiclass VST1D4WB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST1Instruction";
   }
   def _register : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VecListFourD:$Vd),
+                        (ins AddrMode:$Rn, rGPR:$Rm, VecListFourD:$Vd),
                         IIC_VLD1x4u,
                         "vst1", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
@@ -1704,15 +1788,15 @@ multiclass VST1D4WB<bits<4> op7_4, string Dt> {
   }
 }
 
-def VST1d8Q     : VST1D4<{0,0,?,?}, "8">;
-def VST1d16Q    : VST1D4<{0,1,?,?}, "16">;
-def VST1d32Q    : VST1D4<{1,0,?,?}, "32">;
-def VST1d64Q    : VST1D4<{1,1,?,?}, "64">;
+def VST1d8Q     : VST1D4<{0,0,?,?}, "8",  addrmode6align64or128or256>;
+def VST1d16Q    : VST1D4<{0,1,?,?}, "16", addrmode6align64or128or256>;
+def VST1d32Q    : VST1D4<{1,0,?,?}, "32", addrmode6align64or128or256>;
+def VST1d64Q    : VST1D4<{1,1,?,?}, "64", addrmode6align64or128or256>;
 
-defm VST1d8Qwb  : VST1D4WB<{0,0,?,?}, "8">;
-defm VST1d16Qwb : VST1D4WB<{0,1,?,?}, "16">;
-defm VST1d32Qwb : VST1D4WB<{1,0,?,?}, "32">;
-defm VST1d64Qwb : VST1D4WB<{1,1,?,?}, "64">;
+defm VST1d8Qwb  : VST1D4WB<{0,0,?,?}, "8",  addrmode6align64or128or256>;
+defm VST1d16Qwb : VST1D4WB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VST1d32Qwb : VST1D4WB<{1,0,?,?}, "32", addrmode6align64or128or256>;
+defm VST1d64Qwb : VST1D4WB<{1,1,?,?}, "64", addrmode6align64or128or256>;
 
 def VST1d64QPseudo            : VSTQQPseudo<IIC_VST1x4>;
 def VST1d64QPseudoWB_fixed    : VSTQQWBfixedPseudo<IIC_VST1x4u>;
@@ -1720,21 +1804,27 @@ def VST1d64QPseudoWB_register : VSTQQWBPseudo<IIC_VST1x4u>;
 
 //   VST2     : Vector Store (multiple 2-element structures)
 class VST2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
-            InstrItinClass itin>
-  : NLdSt<0, 0b00, op11_8, op7_4, (outs), (ins addrmode6:$Rn, VdTy:$Vd),
+            InstrItinClass itin, Operand AddrMode>
+  : NLdSt<0, 0b00, op11_8, op7_4, (outs), (ins AddrMode:$Rn, VdTy:$Vd),
           itin, "vst2", Dt, "$Vd, $Rn", "", []> {
   let Rm = 0b1111;
   let Inst{5-4} = Rn{5-4};
   let DecoderMethod = "DecodeVLDST2Instruction";
 }
 
-def  VST2d8   : VST2<0b1000, {0,0,?,?}, "8",  VecListDPair, IIC_VST2>;
-def  VST2d16  : VST2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VST2>;
-def  VST2d32  : VST2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VST2>;
+def  VST2d8   : VST2<0b1000, {0,0,?,?}, "8",  VecListDPair, IIC_VST2,
+                     addrmode6align64or128>;
+def  VST2d16  : VST2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VST2,
+                     addrmode6align64or128>;
+def  VST2d32  : VST2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VST2,
+                     addrmode6align64or128>;
 
-def  VST2q8   : VST2<0b0011, {0,0,?,?}, "8",  VecListFourD, IIC_VST2x2>;
-def  VST2q16  : VST2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VST2x2>;
-def  VST2q32  : VST2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VST2x2>;
+def  VST2q8   : VST2<0b0011, {0,0,?,?}, "8",  VecListFourD, IIC_VST2x2,
+                     addrmode6align64or128or256>;
+def  VST2q16  : VST2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VST2x2,
+                     addrmode6align64or128or256>;
+def  VST2q32  : VST2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VST2x2,
+                     addrmode6align64or128or256>;
 
 def  VST2q8Pseudo  : VSTQQPseudo<IIC_VST2x2>;
 def  VST2q16Pseudo : VSTQQPseudo<IIC_VST2x2>;
@@ -1742,9 +1832,9 @@ def  VST2q32Pseudo : VSTQQPseudo<IIC_VST2x2>;
 
 // ...with address register writeback:
 multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
-                   RegisterOperand VdTy> {
+                   RegisterOperand VdTy, Operand AddrMode> {
   def _fixed : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
-                     (ins addrmode6:$Rn, VdTy:$Vd), IIC_VLD1u,
+                     (ins AddrMode:$Rn, VdTy:$Vd), IIC_VLD1u,
                      "vst2", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1752,16 +1842,16 @@ multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
     let DecoderMethod = "DecodeVLDST2Instruction";
   }
   def _register : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VdTy:$Vd), IIC_VLD1u,
+                        (ins AddrMode:$Rn, rGPR:$Rm, VdTy:$Vd), IIC_VLD1u,
                         "vst2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
     let Inst{5-4} = Rn{5-4};
     let DecoderMethod = "DecodeVLDST2Instruction";
   }
 }
-multiclass VST2QWB<bits<4> op7_4, string Dt> {
+multiclass VST2QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
   def _fixed : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
-                     (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1u,
+                     (ins AddrMode:$Rn, VecListFourD:$Vd), IIC_VLD1u,
                      "vst2", Dt, "$Vd, $Rn!",
                      "$Rn.addr = $wb", []> {
     let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
@@ -1769,7 +1859,7 @@ multiclass VST2QWB<bits<4> op7_4, string Dt> {
     let DecoderMethod = "DecodeVLDST2Instruction";
   }
   def _register : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
-                        (ins addrmode6:$Rn, rGPR:$Rm, VecListFourD:$Vd),
+                        (ins AddrMode:$Rn, rGPR:$Rm, VecListFourD:$Vd),
                         IIC_VLD1u,
                         "vst2", Dt, "$Vd, $Rn, $Rm",
                         "$Rn.addr = $wb", []> {
@@ -1778,13 +1868,16 @@ multiclass VST2QWB<bits<4> op7_4, string Dt> {
   }
 }
 
-defm VST2d8wb    : VST2DWB<0b1000, {0,0,?,?}, "8",  VecListDPair>;
-defm VST2d16wb   : VST2DWB<0b1000, {0,1,?,?}, "16", VecListDPair>;
-defm VST2d32wb   : VST2DWB<0b1000, {1,0,?,?}, "32", VecListDPair>;
+defm VST2d8wb    : VST2DWB<0b1000, {0,0,?,?}, "8",  VecListDPair,
+                           addrmode6align64or128>;
+defm VST2d16wb   : VST2DWB<0b1000, {0,1,?,?}, "16", VecListDPair,
+                           addrmode6align64or128>;
+defm VST2d32wb   : VST2DWB<0b1000, {1,0,?,?}, "32", VecListDPair,
+                           addrmode6align64or128>;
 
-defm VST2q8wb    : VST2QWB<{0,0,?,?}, "8">;
-defm VST2q16wb   : VST2QWB<{0,1,?,?}, "16">;
-defm VST2q32wb   : VST2QWB<{1,0,?,?}, "32">;
+defm VST2q8wb    : VST2QWB<{0,0,?,?}, "8", addrmode6align64or128or256>;
+defm VST2q16wb   : VST2QWB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VST2q32wb   : VST2QWB<{1,0,?,?}, "32", addrmode6align64or128or256>;
 
 def VST2q8PseudoWB_fixed     : VSTQQWBfixedPseudo<IIC_VST2x2u>;
 def VST2q16PseudoWB_fixed    : VSTQQWBfixedPseudo<IIC_VST2x2u>;
@@ -1794,12 +1887,18 @@ def VST2q16PseudoWB_register : VSTQQWBregisterPseudo<IIC_VST2x2u>;
 def VST2q32PseudoWB_register : VSTQQWBregisterPseudo<IIC_VST2x2u>;
 
 // ...with double-spaced registers
-def VST2b8      : VST2<0b1001, {0,0,?,?}, "8",  VecListDPairSpaced, IIC_VST2>;
-def VST2b16     : VST2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VST2>;
-def VST2b32     : VST2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VST2>;
-defm VST2b8wb   : VST2DWB<0b1001, {0,0,?,?}, "8",  VecListDPairSpaced>;
-defm VST2b16wb  : VST2DWB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced>;
-defm VST2b32wb  : VST2DWB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced>;
+def VST2b8      : VST2<0b1001, {0,0,?,?}, "8",  VecListDPairSpaced, IIC_VST2,
+                      addrmode6align64or128>;
+def VST2b16     : VST2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VST2,
+                      addrmode6align64or128>;
+def VST2b32     : VST2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VST2,
+                      addrmode6align64or128>;
+defm VST2b8wb   : VST2DWB<0b1001, {0,0,?,?}, "8",  VecListDPairSpaced,
+                          addrmode6align64or128>;
+defm VST2b16wb  : VST2DWB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced,
+                          addrmode6align64or128>;
+defm VST2b32wb  : VST2DWB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced,
+                          addrmode6align64or128>;
 
 //   VST3     : Vector Store (multiple 3-element structures)
 class VST3D<bits<4> op11_8, bits<4> op7_4, string Dt>
@@ -2267,9 +2366,9 @@ def : Pat<(v2f64 (dword_alignedload addrmode6:$addr)),
 def : Pat<(dword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
           (VST1q64 addrmode6:$addr, QPR:$value)>;
 def : Pat<(v2f64 (word_alignedload addrmode6:$addr)),
-          (VLD1q32 addrmode6:$addr)>;
+          (VLD1q32 addrmode6:$addr)>, Requires<[IsLE]>;
 def : Pat<(word_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
-          (VST1q32 addrmode6:$addr, QPR:$value)>;
+          (VST1q32 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
 def : Pat<(v2f64 (hword_alignedload addrmode6:$addr)),
           (VLD1q16 addrmode6:$addr)>, Requires<[IsLE]>;
 def : Pat<(hword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
@@ -2357,14 +2456,14 @@ class N2VDIntnp<bits<2> op17_16, bits<3> op10_8, bit op7,
               InstrItinClass itin, string OpcodeStr, string Dt,
               ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
   : N2Vnp<0b10, op17_16, op10_8, op7, 0,  (outs DPR:$Vd), (ins DPR:$Vm),
-          itin, OpcodeStr, Dt, ResTy, OpTy,
+          itin, OpcodeStr, Dt,
           [(set DPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vm))))]>;
 
 class N2VQIntnp<bits<2> op17_16, bits<3> op10_8, bit op7,
               InstrItinClass itin, string OpcodeStr, string Dt,
               ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
   : N2Vnp<0b10, op17_16, op10_8, op7, 1,  (outs QPR:$Vd), (ins QPR:$Vm),
-          itin, OpcodeStr, Dt, ResTy, OpTy,
+          itin, OpcodeStr, Dt,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vm))))]>;
 
 // Similar to NV2VQIntnp with some more encoding bits exposed (crypto).
@@ -2372,7 +2471,7 @@ class N2VQIntXnp<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op6,
               bit op7, InstrItinClass itin, string OpcodeStr, string Dt,
               ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
   : N2Vnp<op19_18, op17_16, op10_8, op7, op6,  (outs QPR:$Vd), (ins QPR:$Vm),
-          itin, OpcodeStr, Dt, ResTy, OpTy,
+          itin, OpcodeStr, Dt,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vm))))]>;
 
 // Same as N2VQIntXnp but with Vd as a src register.
@@ -2381,7 +2480,7 @@ class N2VQIntX2np<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op6,
               ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
   : N2Vnp<op19_18, op17_16, op10_8, op7, op6,
           (outs QPR:$Vd), (ins QPR:$src, QPR:$Vm),
-          itin, OpcodeStr, Dt, ResTy, OpTy,
+          itin, OpcodeStr, Dt,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$src), (OpTy QPR:$Vm))))]> {
   let Constraints = "$src = $Vd";
 }
@@ -2555,7 +2654,6 @@ class N3VDIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 SDPatternOperator IntOp, bit Commutable>
   : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
           (outs DPR:$Vd), (ins DPR:$Vn, DPR:$Vm), N3RegFrm, itin, OpcodeStr, Dt,
-          ResTy, OpTy, IntOp, Commutable,
           [(set DPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vn), (OpTy DPR:$Vm))))]>;
 
 class N3VDIntSL<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
@@ -2609,7 +2707,6 @@ class N3VQIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 SDPatternOperator IntOp, bit Commutable>
   : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
           (outs QPR:$Vd), (ins QPR:$Vn, QPR:$Vm), f, itin, OpcodeStr, Dt,
-          ResTy, OpTy, IntOp, Commutable,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vn), (OpTy QPR:$Vm))))]>;
 
 // Same as N3VQIntnp but with Vd as a src register.
@@ -2618,8 +2715,8 @@ class N3VQInt3np<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 string Dt, ValueType ResTy, ValueType OpTy,
                 SDPatternOperator IntOp, bit Commutable>
   : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
-          (outs QPR:$Vd), (ins QPR:$src, QPR:$Vn, QPR:$Vm), f, itin, OpcodeStr,
-          Dt, ResTy, OpTy, IntOp, Commutable,
+          (outs QPR:$Vd), (ins QPR:$src, QPR:$Vn, QPR:$Vm),
+          f, itin, OpcodeStr, Dt,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$src), (OpTy QPR:$Vn),
                                        (OpTy QPR:$Vm))))]> {
   let Constraints = "$src = $Vd";
@@ -2939,7 +3036,6 @@ class N3VLIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
                 SDPatternOperator IntOp, bit Commutable>
   : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
           (outs QPR:$Vd), (ins DPR:$Vn, DPR:$Vm), N3RegFrm, itin, OpcodeStr, Dt,
-          ResTy, OpTy, IntOp, Commutable,
           [(set QPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vn), (OpTy DPR:$Vm))))]>;
 
 class N3VLIntSL<bit op24, bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
@@ -5245,6 +5341,35 @@ def VMOVv4f32 : N1ModImm<1, 0b000, 0b1111, 0, 1, 0, 1, (outs QPR:$Vd),
                          [(set QPR:$Vd, (v4f32 (NEONvmovFPImm timm:$SIMM)))]>;
 } // isReMaterializable
 
+// Add support for bytes replication feature, so it could be GAS compatible.
+// E.g. instructions below:
+// "vmov.i32 d0, 0xffffffff"
+// "vmov.i32 d0, 0xabababab"
+// "vmov.i16 d0, 0xabab"
+// are incorrect, but we could deal with such cases.
+// For last two instructions, for example, it should emit:
+// "vmov.i8 d0, 0xab"
+def : NEONInstAlias<"vmov${p}.i16 $Vd, $Vm",
+                    (VMOVv8i8 DPR:$Vd, nImmVMOVI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $Vm",
+                    (VMOVv8i8 DPR:$Vd, nImmVMOVI32ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i16 $Vd, $Vm",
+                    (VMOVv16i8 QPR:$Vd, nImmVMOVI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $Vm",
+                    (VMOVv16i8 QPR:$Vd, nImmVMOVI32ByteReplicate:$Vm, pred:$p)>;
+
+// Also add same support for VMVN instructions. So instruction:
+// "vmvn.i32 d0, 0xabababab"
+// actually means:
+// "vmov.i8 d0, 0x54"
+def : NEONInstAlias<"vmvn${p}.i16 $Vd, $Vm",
+                    (VMOVv8i8 DPR:$Vd, nImmVMVNI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $Vm",
+                    (VMOVv8i8 DPR:$Vd, nImmVMVNI32ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i16 $Vd, $Vm",
+                    (VMOVv16i8 QPR:$Vd, nImmVMVNI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $Vm",
+                    (VMOVv16i8 QPR:$Vd, nImmVMVNI32ByteReplicate:$Vm, pred:$p)>;
 
 // On some CPUs the two instructions "vmov.i32 dD, #0" and "vmov.i32 qD, #0"
 // require zero cycles to execute so they should be used wherever possible for
@@ -5617,22 +5742,22 @@ def VCVTxu2fq : N2VCvtQ<1, 1, 0b1110, 0, 1, "vcvt", "f32.u32",
                         v4f32, v4i32, int_arm_neon_vcvtfxu2fp>;
 }
 
-def : NEONInstAlias<"vcvt${p}.s32.f32 $Dd, $Dm, #0", 
+def : NEONInstAlias<"vcvt${p}.s32.f32 $Dd, $Dm, #0",
                     (VCVTf2sd DPR:$Dd, DPR:$Dm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.u32.f32 $Dd, $Dm, #0", 
+def : NEONInstAlias<"vcvt${p}.u32.f32 $Dd, $Dm, #0",
                     (VCVTf2ud DPR:$Dd, DPR:$Dm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.f32.s32 $Dd, $Dm, #0", 
+def : NEONInstAlias<"vcvt${p}.f32.s32 $Dd, $Dm, #0",
                     (VCVTs2fd DPR:$Dd, DPR:$Dm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.f32.u32 $Dd, $Dm, #0", 
+def : NEONInstAlias<"vcvt${p}.f32.u32 $Dd, $Dm, #0",
                     (VCVTu2fd DPR:$Dd, DPR:$Dm, pred:$p)>;
 
-def : NEONInstAlias<"vcvt${p}.s32.f32 $Qd, $Qm, #0", 
+def : NEONInstAlias<"vcvt${p}.s32.f32 $Qd, $Qm, #0",
                     (VCVTf2sq QPR:$Qd, QPR:$Qm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.u32.f32 $Qd, $Qm, #0", 
+def : NEONInstAlias<"vcvt${p}.u32.f32 $Qd, $Qm, #0",
                     (VCVTf2uq QPR:$Qd, QPR:$Qm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.f32.s32 $Qd, $Qm, #0", 
+def : NEONInstAlias<"vcvt${p}.f32.s32 $Qd, $Qm, #0",
                     (VCVTs2fq QPR:$Qd, QPR:$Qm, pred:$p)>;
-def : NEONInstAlias<"vcvt${p}.f32.u32 $Qd, $Qm, #0", 
+def : NEONInstAlias<"vcvt${p}.f32.u32 $Qd, $Qm, #0",
                     (VCVTu2fq QPR:$Qd, QPR:$Qm, pred:$p)>;
 
 
@@ -6051,67 +6176,145 @@ def : Pat<(f32 (bitconvert GPR:$a)),
 //===----------------------------------------------------------------------===//
 
 // bit_convert
-def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v8i8  DPR:$src))), (v1i64 DPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (v1i64 DPR:$src)>;
+  def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (v1i64 DPR:$src)>;
+  def : Pat<(v1i64 (bitconvert (v8i8  DPR:$src))), (v1i64 DPR:$src)>;
+}
 def : Pat<(v1i64 (bitconvert (f64   DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v8i8  DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (f64   DPR:$src))), (v2i32 DPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (v1i64 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (v2i32 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (v2i32 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v8i8  DPR:$src))), (v2i32 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (f64   DPR:$src))), (v2i32 DPR:$src)>;
+}
 def : Pat<(v2i32 (bitconvert (v2f32 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v8i8  DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (f64   DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v8i8  (bitconvert (v1i64 DPR:$src))), (v8i8  DPR:$src)>;
-def : Pat<(v8i8  (bitconvert (v2i32 DPR:$src))), (v8i8  DPR:$src)>;
-def : Pat<(v8i8  (bitconvert (v4i16 DPR:$src))), (v8i8  DPR:$src)>;
-def : Pat<(v8i8  (bitconvert (f64   DPR:$src))), (v8i8  DPR:$src)>;
-def : Pat<(v8i8  (bitconvert (v2f32 DPR:$src))), (v8i8  DPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (v4i16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (v4i16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v8i8  DPR:$src))), (v4i16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (f64   DPR:$src))), (v4i16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (v4i16 DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v1i64 DPR:$src))), (v8i8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v2i32 DPR:$src))), (v8i8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v4i16 DPR:$src))), (v8i8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (f64   DPR:$src))), (v8i8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v2f32 DPR:$src))), (v8i8  DPR:$src)>;
+}
 def : Pat<(f64   (bitconvert (v1i64 DPR:$src))), (f64   DPR:$src)>;
-def : Pat<(f64   (bitconvert (v2i32 DPR:$src))), (f64   DPR:$src)>;
-def : Pat<(f64   (bitconvert (v4i16 DPR:$src))), (f64   DPR:$src)>;
-def : Pat<(f64   (bitconvert (v8i8  DPR:$src))), (f64   DPR:$src)>;
-def : Pat<(f64   (bitconvert (v2f32 DPR:$src))), (f64   DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (f64   DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (v2f32 DPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(f64   (bitconvert (v2i32 DPR:$src))), (f64   DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v4i16 DPR:$src))), (f64   DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v8i8  DPR:$src))), (f64   DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v2f32 DPR:$src))), (f64   DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (f64   DPR:$src))), (v2f32 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (v2f32 DPR:$src)>;
+}
 def : Pat<(v2f32 (bitconvert (v2i32 DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v8i8  DPR:$src))), (v2f32 DPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (v2f32 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (v8i8  DPR:$src))), (v2f32 DPR:$src)>;
+}
 
-def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (v2i64 QPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (v2i64 QPR:$src)>;
+  def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (v2i64 QPR:$src)>;
+  def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (v2i64 QPR:$src)>;
+}
 def : Pat<(v2i64 (bitconvert (v2f64 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (v4i32 QPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (v2i64 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (v4i32 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (v4i32 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (v4i32 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (v4i32 QPR:$src)>;
+}
 def : Pat<(v4i32 (bitconvert (v4f32 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (v4f32 QPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (v8i16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (v8i16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (v8i16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (v8i16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (v8i16 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (v16i8 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (v16i8 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (v16i8 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (v16i8 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (v16i8 QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (v4f32 QPR:$src)>;
+}
 def : Pat<(v4f32 (bitconvert (v4i32 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (v4f32 QPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (v4f32 QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (v4f32 QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (v4f32 QPR:$src)>;
+}
 def : Pat<(v2f64 (bitconvert (v2i64 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
+let Predicates = [IsLE] in {
+  def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (v2f64 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
+}
+
+let Predicates = [IsBE] in {
+  // 64 bit conversions
+  def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (VREV64d16 DPR:$src)>;
+  def : Pat<(v1i64 (bitconvert (v8i8  DPR:$src))), (VREV64d8  DPR:$src)>;
+  def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (VREV32d16 DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (v8i8  DPR:$src))), (VREV32d8  DPR:$src)>;
+  def : Pat<(v2i32 (bitconvert (f64   DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (VREV64d16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (VREV32d16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v8i8  DPR:$src))), (VREV16d8  DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (f64   DPR:$src))), (VREV64d16 DPR:$src)>;
+  def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (VREV32d16 DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v1i64 DPR:$src))), (VREV64d8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v2i32 DPR:$src))), (VREV32d8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v4i16 DPR:$src))), (VREV16d8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (f64   DPR:$src))), (VREV64d8  DPR:$src)>;
+  def : Pat<(v8i8  (bitconvert (v2f32 DPR:$src))), (VREV32d8  DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v2i32 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v4i16 DPR:$src))), (VREV64d16 DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v8i8  DPR:$src))), (VREV64d8  DPR:$src)>;
+  def : Pat<(f64   (bitconvert (v2f32 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (f64   DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (VREV64d32 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (VREV32d16 DPR:$src)>;
+  def : Pat<(v2f32 (bitconvert (v8i8  DPR:$src))), (VREV32d8  DPR:$src)>;
+
+  // 128 bit conversions
+  def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (VREV64q16 QPR:$src)>;
+  def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (VREV64q8  QPR:$src)>;
+  def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (VREV32q16 QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (VREV32q8  QPR:$src)>;
+  def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (VREV64q16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (VREV32q16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (VREV16q8  QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (VREV64q16 QPR:$src)>;
+  def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (VREV32q16 QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (VREV64q8  QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (VREV32q8  QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (VREV16q8  QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (VREV64q8  QPR:$src)>;
+  def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (VREV32q8  QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (VREV32q16 QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (VREV32q8  QPR:$src)>;
+  def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (VREV64q32 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (VREV64q16 QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (VREV64q8  QPR:$src)>;
+  def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (VREV64q32 QPR:$src)>;
+}
 
 // Fold extracting an element out of a v2i32 into a vfp register.
 def : Pat<(f32 (bitconvert (i32 (extractelt (v2i32 DPR:$src), imm:$lane)))),
@@ -6120,7 +6323,7 @@ def : Pat<(f32 (bitconvert (i32 (extractelt (v2i32 DPR:$src), imm:$lane)))),
 // Vector lengthening move with load, matching extending loads.
 
 // extload, zextload and sextload for a standard lengthening load. Example:
-// Lengthen_Single<"8", "i16", "8"> = 
+// Lengthen_Single<"8", "i16", "8"> =
 //     Pat<(v8i16 (extloadvi8 addrmode6:$addr))
 //         (VMOVLuv8i16 (VLD1d8 addrmode6:$addr,
 //                              (f64 (IMPLICIT_DEF)), (i32 0)))>;
@@ -6147,7 +6350,7 @@ multiclass Lengthen_Single<string DestLanes, string DestTy, string SrcTy> {
 // half the lanes available. Example:
 // Lengthen_HalfSingle<"4", "i16", "8", "i16", "i8"> =
 //     Pat<(v4i16 (extloadvi8 addrmode6oneL32:$addr)),
-//         (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr, 
+//         (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
 //                                      (f64 (IMPLICIT_DEF)), (i32 0))),
 //                         dsub_0)>;
 multiclass Lengthen_HalfSingle<string DestLanes, string DestTy, string SrcTy,
@@ -6257,7 +6460,7 @@ defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64">;
 // Triple lengthening - v2i8 -> v2i16 -> v2i32 -> v2i64
 def : Pat<(v2i64 (extloadvi8 addrmode6:$addr)),
       (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
-         (VLD1LNd16 addrmode6:$addr, 
+         (VLD1LNd16 addrmode6:$addr,
                     (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
 def : Pat<(v2i64 (zextloadvi8 addrmode6:$addr)),
       (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
@@ -6311,379 +6514,442 @@ defm : NEONDTAnyInstAlias<"vorr${p}", "$Vdn, $Vm",
 // VLD1 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD1LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr",
-                 (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+                      pred:$p)>;
 def VLD1LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr",
-                 (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VLD1LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr",
-                 (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 
 def VLD1LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr!",
-                 (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+                      pred:$p)>;
 def VLD1LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr!",
-                 (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VLD1LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr!",
-                 (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VLD1LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD1LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD1LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 
 // VST1 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VST1LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr",
-                 (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+                      pred:$p)>;
 def VST1LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr",
-                 (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VST1LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr",
-                 (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 
 def VST1LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr!",
-                 (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+                      pred:$p)>;
 def VST1LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr!",
-                 (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VST1LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr!",
-                 (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VST1LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST1LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST1LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VLD2 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD2LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr",
-                 (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+                  pred:$p)>;
 def VLD2LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr",
-                 (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VLD2LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr",
-                 (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD2LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr",
-                 (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VLD2LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr",
-                 (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 
 def VLD2LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr!",
-                 (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VLD2LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr!",
-                 (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VLD2LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr!",
-                 (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 def VLD2LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr!",
-                 (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VLD2LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr!",
-                 (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 def VLD2LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD2LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD2LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD2LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD2LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 
 // VST2 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VST2LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr",
-                 (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VST2LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr",
-                 (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VST2LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr",
-                 (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 def VST2LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr",
-                 (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VST2LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr",
-                 (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 
 def VST2LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr!",
-                 (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+                      pred:$p)>;
 def VST2LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr!",
-                 (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VST2LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr!",
-                 (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 def VST2LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr!",
-                 (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+                      pred:$p)>;
 def VST2LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr!",
-                 (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+                 (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+                      pred:$p)>;
 def VST2LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST2LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".16","$list, $addr, $Rm",
-                  (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST2LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST2LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".16","$list, $addr, $Rm",
-                  (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST2LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VLD3 all-lanes pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD3DUPdAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPdAsm_16: NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPdAsm_32: NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqAsm_16: NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqAsm_32: NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 
 def VLD3DUPdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+                    pred:$p)>;
 def VLD3DUPdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3DUPdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3DUPdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3DUPqWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3DUPqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3DUPqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+                  (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 
 // VLD3 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD3LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
-               (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 
 def VLD3LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VLD3LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListThreeDHWordIndexed:$list,
+                       addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
 def VLD3LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListThreeQHWordIndexed:$list,
+                       addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
 def VLD3LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeQWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VLD3 multiple structure pseudo-instructions. These need special handling for
 // the vector operands that the normal instructions don't yet model.
 // FIXME: Remove these when the register classes and instructions are updated.
 def VLD3dAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3dAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3dAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 
 def VLD3dWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3dWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3dWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3qWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VLD3dWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3dWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3dWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3qWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3qWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD3qWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VST3 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VST3LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
-               (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
-               (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
-               (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
-               (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
-               (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 
 def VST3LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+                    pred:$p)>;
 def VST3LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListThreeDHWordIndexed:$list,
+                       addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
 def VST3LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeDWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListThreeQHWordIndexed:$list,
+                       addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
 def VST3LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeQWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 
@@ -6691,168 +6957,190 @@ def VST3LNqWB_register_Asm_32 :
 // the vector operands that the normal instructions don't yet model.
 // FIXME: Remove these when the register classes and instructions are updated.
 def VST3dAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3dAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3dAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 
 def VST3dWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3dWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3dWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3qWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
-               (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
 def VST3dWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3dWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3dWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeD:$list, addrmode6:$addr,
+                  (ins VecListThreeD:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3qWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3qWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST3qWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListThreeQ:$list, addrmode6:$addr,
+                  (ins VecListThreeQ:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VLD4 all-lanes pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD4DUPdAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
+                    pred:$p)>;
 def VLD4DUPdAsm_16: NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
+                    pred:$p)>;
 def VLD4DUPdAsm_32: NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign64or128:$addr,
+                    pred:$p)>;
 def VLD4DUPqAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
+                    pred:$p)>;
 def VLD4DUPqAsm_16: NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
+                    pred:$p)>;
 def VLD4DUPqAsm_32: NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign64or128:$addr,
+                    pred:$p)>;
 
 def VLD4DUPdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
+                    pred:$p)>;
 def VLD4DUPdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
+                    pred:$p)>;
 def VLD4DUPdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDAllLanes:$list, addrmode6dupalign64or128:$addr,
+                    pred:$p)>;
 def VLD4DUPqWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
+                    pred:$p)>;
 def VLD4DUPqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
+                    pred:$p)>;
 def VLD4DUPqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQAllLanes:$list, addrmode6dupalign64or128:$addr,
+                    pred:$p)>;
 def VLD4DUPdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourDAllLanes:$list, addrmode6:$addr,
+                  (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4DUPdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourDAllLanes:$list, addrmode6:$addr,
+                  (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4DUPdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourDAllLanes:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourDAllLanes:$list,
+                       addrmode6dupalign64or128:$addr, rGPR:$Rm, pred:$p)>;
 def VLD4DUPqWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourQAllLanes:$list, addrmode6:$addr,
+                  (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4DUPqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourQAllLanes:$list, addrmode6:$addr,
+                  (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4DUPqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourQAllLanes:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourQAllLanes:$list,
+                       addrmode6dupalign64or128:$addr, rGPR:$Rm, pred:$p)>;
 
 
 // VLD4 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD4LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
-               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+                    pred:$p)>;
 def VLD4LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VLD4LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VLD4LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VLD4LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 
 def VLD4LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
-               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+                    pred:$p)>;
 def VLD4LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VLD4LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VLD4LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VLD4LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VLD4LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourDWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourDWordIndexed:$list,
+                       addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
 def VLD4LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourQHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourQWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourQWordIndexed:$list,
+                       addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
 
 
 
@@ -6860,168 +7148,202 @@ def VLD4LNqWB_register_Asm_32 :
 // the vector operands that the normal instructions don't yet model.
 // FIXME: Remove these when the register classes and instructions are updated.
 def VLD4dAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4dAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4dAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 
 def VLD4dWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4dWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4dWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4qWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                pred:$p)>;
 def VLD4dWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4dWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4dWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4qWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4qWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VLD4qWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VST4 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VST4LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
-               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+                    pred:$p)>;
 def VST4LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
-               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VST4LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
-               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VST4LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
-               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VST4LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
-               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 
 def VST4LNdWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
-               (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+                    pred:$p)>;
 def VST4LNdWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
-               (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VST4LNdWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
-               (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VST4LNqWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
-               (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+                    pred:$p)>;
 def VST4LNqWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
-               (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+                    pred:$p)>;
 def VST4LNdWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourDByteIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4LNdWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourDHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4LNdWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourDWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourDWordIndexed:$list,
+                       addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
 def VST4LNqWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourQHWordIndexed:$list, addrmode6:$addr,
+                  (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4LNqWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourQWordIndexed:$list, addrmode6:$addr,
-                       rGPR:$Rm, pred:$p)>;
+                  (ins VecListFourQWordIndexed:$list,
+                       addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
 
 
 // VST4 multiple structure pseudo-instructions. These need special handling for
 // the vector operands that the normal instructions don't yet model.
 // FIXME: Remove these when the register classes and instructions are updated.
 def VST4dAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4dAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4dAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 
 def VST4dWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4dWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4dWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
-               (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qWB_fixed_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qWB_fixed_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4qWB_fixed_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
-               (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+               (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+                    pred:$p)>;
 def VST4dWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4dWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4dWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourD:$list, addrmode6:$addr,
+                  (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4qWB_register_Asm_8 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4qWB_register_Asm_16 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 def VST4qWB_register_Asm_32 :
         NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
-                  (ins VecListFourQ:$list, addrmode6:$addr,
+                  (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
                        rGPR:$Rm, pred:$p)>;
 
 // VMOV/VMVN takes an optional datatype suffix
diff --git a/lib/Target/ARM/ARMInstrThumb.td b/lib/Target/ARM/ARMInstrThumb.td
index 754295f..e17f73a 100644
--- a/lib/Target/ARM/ARMInstrThumb.td
+++ b/lib/Target/ARM/ARMInstrThumb.td
@@ -269,7 +269,8 @@ class T1SystemEncoding<bits<8> opc>
   let Inst{7-0} = opc;
 }
 
-def tHINT : T1pI<(outs), (ins imm0_15:$imm), NoItinerary, "hint", "\t$imm", []>,
+def tHINT : T1pI<(outs), (ins imm0_15:$imm), NoItinerary, "hint", "\t$imm",
+                 [(int_arm_hint imm0_15:$imm)]>,
             T1SystemEncoding<0x00>,
             Requires<[IsThumb, HasV6M]> {
   bits<4> imm;
@@ -288,7 +289,6 @@ def : tHintAlias<"sev$p", (tHINT 4, pred:$p)>; // A8.6.157
 def : tInstAlias<"sevl$p", (tHINT 5, pred:$p)> {
   let Predicates = [IsThumb2, HasV8];
 }
-def : T2Pat<(int_arm_sevl), (tHINT 5)>;
 
 // The imm operand $val can be used by a debugger to store more information
 // about the breakpoint.
@@ -1193,6 +1193,15 @@ def tTST :                      // A8.6.230
                [(ARMcmpZ (and_su tGPR:$Rn, tGPR:$Rm), 0)]>,
                Sched<[WriteALU]>;
 
+// A8.8.247  UDF - Undefined (Encoding T1)
+def tUDF : TI<(outs), (ins imm0_255:$imm8), IIC_Br, "udf\t$imm8",
+              [(int_arm_undefined imm0_255:$imm8)]>, Encoding16 {
+  bits<8> imm8;
+  let Inst{15-12} = 0b1101;
+  let Inst{11-8} = 0b1110;
+  let Inst{7-0} = imm8;
+}
+
 // Zero-extend byte
 def tUXTB :                     // A8.6.262
   T1pIMiscEncode<{0,0,1,0,1,1,?}, (outs tGPR:$Rd), (ins tGPR:$Rm),
@@ -1308,6 +1317,18 @@ def : T1Pat<(addc   tGPR:$lhs, imm8_255_neg:$rhs),
 def : T1Pat<(subc   tGPR:$lhs, tGPR:$rhs),
             (tSUBrr tGPR:$lhs, tGPR:$rhs)>;
 
+// Bswap 16 with load/store
+def : T1Pat<(srl (bswap (extloadi16 t_addrmode_rrs2:$addr)), (i32 16)),
+            (tREV16 (tLDRHr t_addrmode_rrs2:$addr))>;
+def : T1Pat<(srl (bswap (extloadi16 t_addrmode_is2:$addr)), (i32 16)),
+            (tREV16 (tLDRHi t_addrmode_is2:$addr))>;
+def : T1Pat<(truncstorei16 (srl (bswap tGPR:$Rn), (i32 16)),
+                           t_addrmode_rrs2:$addr),
+            (tSTRHr (tREV16 tGPR:$Rn), t_addrmode_rrs2:$addr)>;
+def : T1Pat<(truncstorei16 (srl (bswap tGPR:$Rn), (i32 16)),
+                           t_addrmode_is2:$addr),
+            (tSTRHi(tREV16 tGPR:$Rn), t_addrmode_is2:$addr)>;
+
 // ConstantPool
 def : T1Pat<(ARMWrapper  tconstpool  :$dst), (tLEApcrel tconstpool  :$dst)>;
 
diff --git a/lib/Target/ARM/ARMInstrThumb2.td b/lib/Target/ARM/ARMInstrThumb2.td
index 387bd60..c30d6ab 100644
--- a/lib/Target/ARM/ARMInstrThumb2.td
+++ b/lib/Target/ARM/ARMInstrThumb2.td
@@ -1445,7 +1445,7 @@ defm t2STRH:T2I_st<0b01,"strh", IIC_iStore_bh_i, IIC_iStore_bh_si,
 // Store doubleword
 let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in
 def t2STRDi8 : T2Ii8s4<1, 0, 0, (outs),
-                       (ins GPR:$Rt, GPR:$Rt2, t2addrmode_imm8s4:$addr),
+                       (ins rGPR:$Rt, rGPR:$Rt2, t2addrmode_imm8s4:$addr),
                IIC_iStore_d_r, "strd", "\t$Rt, $Rt2, $addr", "", []>;
 
 // Indexed stores
@@ -1676,7 +1676,7 @@ defm t2PLI    : T2Ipl<0, 1, "pli">,  Requires<[IsThumb2,HasV7]>;
 // pci variant is very similar to i12, but supports negative offsets
 // from the PC. Only PLD and PLI have pci variants (not PLDW)
 class T2Iplpci<bits<1> inst, string opc> : T2Iso<(outs), (ins t2ldrlabel:$addr),
-               IIC_Preload, opc, "\t$addr", 
+               IIC_Preload, opc, "\t$addr",
                [(ARMPreload (ARMWrapper tconstpool:$addr),
                 (i32 0), (i32 inst))]>, Sched<[WritePreLd]> {
   let Inst{31-25} = 0b1111100;
@@ -1918,7 +1918,7 @@ def t2MOVi16 : T2I<(outs rGPR:$Rd), (ins imm0_65535_expr:$imm), IIC_iMOVi,
   let DecoderMethod = "DecodeT2MOVTWInstruction";
 }
 
-def : t2InstAlias<"mov${p} $Rd, $imm", 
+def : t2InstAlias<"mov${p} $Rd, $imm",
                   (t2MOVi16 rGPR:$Rd, imm256_65535_expr:$imm, pred:$p)>;
 
 def t2MOVi16_ga_pcrel : PseudoInst<(outs rGPR:$Rd),
@@ -2407,6 +2407,19 @@ def t2UBFX: T2TwoRegBitFI<
   let Inst{15} = 0;
 }
 
+// A8.8.247  UDF - Undefined (Encoding T2)
+def t2UDF : T2XI<(outs), (ins imm0_65535:$imm16), IIC_Br, "udf.w\t$imm16",
+                 [(int_arm_undefined imm0_65535:$imm16)]> {
+  bits<16> imm16;
+  let Inst{31-29} = 0b111;
+  let Inst{28-27} = 0b10;
+  let Inst{26-20} = 0b1111111;
+  let Inst{19-16} = imm16{15-12};
+  let Inst{15} = 0b1;
+  let Inst{14-12} = 0b010;
+  let Inst{11-0} = imm16{11-0};
+}
+
 // A8.6.18  BFI - Bitfield insert (Encoding T1)
 let Constraints = "$src = $Rd" in {
   def t2BFI : T2TwoRegBitFI<(outs rGPR:$Rd),
@@ -3495,8 +3508,8 @@ def t2B   : T2I<(outs), (ins uncondbrtarget:$target), IIC_Br,
   let Inst{25-16} = target{20-11};
   let Inst{10-0} = target{10-0};
   let DecoderMethod = "DecodeT2BInstruction";
-  let AsmMatchConverter = "cvtThumbBranches"; 
-} 
+  let AsmMatchConverter = "cvtThumbBranches";
+}
 
 let isNotDuplicable = 1, isIndirectBranch = 1 in {
 def t2BR_JT : t2PseudoInst<(outs),
@@ -3671,7 +3684,8 @@ def : t2InstAlias<"cps.w $mode", (t2CPS1p imm0_31:$mode), 0>;
 
 // A6.3.4 Branches and miscellaneous control
 // Table A6-14 Change Processor State, and hint instructions
-def t2HINT : T2I<(outs), (ins imm0_239:$imm), NoItinerary, "hint", ".w\t$imm",[]> {
+def t2HINT : T2I<(outs), (ins imm0_239:$imm), NoItinerary, "hint", ".w\t$imm",
+                  [(int_arm_hint imm0_239:$imm)]> {
   bits<8> imm;
   let Inst{31-3} = 0b11110011101011111000000000000;
   let Inst{7-0} = imm;
@@ -3698,7 +3712,7 @@ def t2DBG : T2I<(outs), (ins imm0_15:$opt), NoItinerary, "dbg", "\t$opt", []> {
 
 // Secure Monitor Call is a system instruction.
 // Option = Inst{19-16}
-def t2SMC : T2I<(outs), (ins imm0_15:$opt), NoItinerary, "smc", "\t$opt", 
+def t2SMC : T2I<(outs), (ins imm0_15:$opt), NoItinerary, "smc", "\t$opt",
                 []>, Requires<[IsThumb2, HasTrustZone]> {
   let Inst{31-27} = 0b11110;
   let Inst{26-20} = 0b1111111;
@@ -4278,7 +4292,7 @@ def : t2InstAlias<"sbc${s}${p} $Rd, $Rn, $ShiftedRm",
 
 // Aliases for ADD without the ".w" optional width specifier.
 def : t2InstAlias<"add${s}${p} $Rd, $Rn, $imm",
-        (t2ADDri GPRnopc:$Rd, GPRnopc:$Rn, t2_so_imm:$imm, pred:$p, 
+        (t2ADDri GPRnopc:$Rd, GPRnopc:$Rn, t2_so_imm:$imm, pred:$p,
          cc_out:$s)>;
 def : t2InstAlias<"add${p} $Rd, $Rn, $imm",
            (t2ADDri12 GPRnopc:$Rd, GPR:$Rn, imm0_4095:$imm, pred:$p)>;
diff --git a/lib/Target/ARM/ARMJITInfo.cpp b/lib/Target/ARM/ARMJITInfo.cpp
index 73c6eb7..8821c2d 100644
--- a/lib/Target/ARM/ARMJITInfo.cpp
+++ b/lib/Target/ARM/ARMJITInfo.cpp
@@ -11,7 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "jit"
 #include "ARMJITInfo.h"
 #include "ARMConstantPoolValue.h"
 #include "ARMRelocations.h"
@@ -25,6 +24,8 @@
 #include <cstdlib>
 using namespace llvm;
 
+#define DEBUG_TYPE "jit"
+
 void ARMJITInfo::replaceMachineCodeForFunction(void *Old, void *New) {
   report_fatal_error("ARMJITInfo::replaceMachineCodeForFunction");
 }
@@ -319,13 +320,13 @@ void ARMJITInfo::relocate(void *Function, MachineRelocation *MR,
       break;
     }
     case ARM::reloc_arm_movw: {
-      ResultPtr = ResultPtr & 0xFFFF; 
+      ResultPtr = ResultPtr & 0xFFFF;
       *((intptr_t*)RelocPos) |= ResultPtr & 0xFFF;
       *((intptr_t*)RelocPos) |= ((ResultPtr >> 12) & 0xF) << 16;
       break;
     }
     case ARM::reloc_arm_movt: {
-      ResultPtr = (ResultPtr >> 16) & 0xFFFF; 
+      ResultPtr = (ResultPtr >> 16) & 0xFFFF;
       *((intptr_t*)RelocPos) |= ResultPtr & 0xFFF;
       *((intptr_t*)RelocPos) |= ((ResultPtr >> 12) & 0xF) << 16;
       break;
diff --git a/lib/Target/ARM/ARMLoadStoreOptimizer.cpp b/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
index 48e0bd7..ee7df54 100644
--- a/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
+++ b/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
@@ -12,13 +12,14 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-ldst-opt"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMBaseRegisterInfo.h"
+#include "ARMISelLowering.h"
 #include "ARMMachineFunctionInfo.h"
 #include "ARMSubtarget.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
+#include "Thumb1RegisterInfo.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
@@ -42,6 +43,8 @@
 #include "llvm/Target/TargetRegisterInfo.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-ldst-opt"
+
 STATISTIC(NumLDMGened , "Number of ldm instructions generated");
 STATISTIC(NumSTMGened , "Number of stm instructions generated");
 STATISTIC(NumVLDMGened, "Number of vldm instructions generated");
@@ -65,9 +68,10 @@ namespace {
     const TargetInstrInfo *TII;
     const TargetRegisterInfo *TRI;
     const ARMSubtarget *STI;
+    const TargetLowering *TL;
     ARMFunctionInfo *AFI;
     RegScavenger *RS;
-    bool isThumb2;
+    bool isThumb1, isThumb2;
 
     bool runOnMachineFunction(MachineFunction &Fn) override;
 
@@ -93,7 +97,10 @@ namespace {
     void findUsesOfImpDef(SmallVectorImpl<MachineOperand *> &UsesOfImpDefs,
                           const MemOpQueue &MemOps, unsigned DefReg,
                           unsigned RangeBegin, unsigned RangeEnd);
-
+    void UpdateBaseRegUses(MachineBasicBlock &MBB,
+                           MachineBasicBlock::iterator MBBI,
+                           DebugLoc dl, unsigned Base, unsigned WordOffset,
+                           ARMCC::CondCodes Pred, unsigned PredReg);
     bool MergeOps(MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
                   int Offset, unsigned Base, bool BaseKill, int Opcode,
                   ARMCC::CondCodes Pred, unsigned PredReg, unsigned Scratch,
@@ -119,7 +126,6 @@ namespace {
                       ARMCC::CondCodes Pred, unsigned PredReg,
                       unsigned Scratch, MemOpQueue &MemOps,
                       SmallVectorImpl<MachineBasicBlock::iterator> &Merges);
-
     void AdvanceRS(MachineBasicBlock &MBB, MemOpQueue &MemOps);
     bool FixInvalidRegPairOp(MachineBasicBlock &MBB,
                              MachineBasicBlock::iterator &MBBI);
@@ -159,6 +165,21 @@ static int getLoadStoreMultipleOpcode(int Opcode, ARM_AM::AMSubMode Mode) {
     case ARM_AM::db: return ARM::STMDB;
     case ARM_AM::ib: return ARM::STMIB;
     }
+  case ARM::tLDRi:
+    // tLDMIA is writeback-only - unless the base register is in the input
+    // reglist.
+    ++NumLDMGened;
+    switch (Mode) {
+    default: llvm_unreachable("Unhandled submode!");
+    case ARM_AM::ia: return ARM::tLDMIA;
+    }
+  case ARM::tSTRi:
+    // There is no non-writeback tSTMIA either.
+    ++NumSTMGened;
+    switch (Mode) {
+    default: llvm_unreachable("Unhandled submode!");
+    case ARM_AM::ia: return ARM::tSTMIA_UPD;
+    }
   case ARM::t2LDRi8:
   case ARM::t2LDRi12:
     ++NumLDMGened;
@@ -217,6 +238,9 @@ AMSubMode getLoadStoreMultipleSubMode(int Opcode) {
   case ARM::LDMIA_UPD:
   case ARM::STMIA:
   case ARM::STMIA_UPD:
+  case ARM::tLDMIA:
+  case ARM::tLDMIA_UPD:
+  case ARM::tSTMIA_UPD:
   case ARM::t2LDMIA_RET:
   case ARM::t2LDMIA:
   case ARM::t2LDMIA_UPD:
@@ -263,12 +287,20 @@ AMSubMode getLoadStoreMultipleSubMode(int Opcode) {
   } // end namespace ARM_AM
 } // end namespace llvm
 
+static bool isT1i32Load(unsigned Opc) {
+  return Opc == ARM::tLDRi;
+}
+
 static bool isT2i32Load(unsigned Opc) {
   return Opc == ARM::t2LDRi12 || Opc == ARM::t2LDRi8;
 }
 
 static bool isi32Load(unsigned Opc) {
-  return Opc == ARM::LDRi12 || isT2i32Load(Opc);
+  return Opc == ARM::LDRi12 || isT1i32Load(Opc) || isT2i32Load(Opc) ;
+}
+
+static bool isT1i32Store(unsigned Opc) {
+  return Opc == ARM::tSTRi;
 }
 
 static bool isT2i32Store(unsigned Opc) {
@@ -276,7 +308,102 @@ static bool isT2i32Store(unsigned Opc) {
 }
 
 static bool isi32Store(unsigned Opc) {
-  return Opc == ARM::STRi12 || isT2i32Store(Opc);
+  return Opc == ARM::STRi12 || isT1i32Store(Opc) || isT2i32Store(Opc);
+}
+
+static unsigned getImmScale(unsigned Opc) {
+  switch (Opc) {
+  default: llvm_unreachable("Unhandled opcode!");
+  case ARM::tLDRi:
+  case ARM::tSTRi:
+    return 1;
+  case ARM::tLDRHi:
+  case ARM::tSTRHi:
+    return 2;
+  case ARM::tLDRBi:
+  case ARM::tSTRBi:
+    return 4;
+  }
+}
+
+/// Update future uses of the base register with the offset introduced
+/// due to writeback. This function only works on Thumb1.
+void
+ARMLoadStoreOpt::UpdateBaseRegUses(MachineBasicBlock &MBB,
+                                   MachineBasicBlock::iterator MBBI,
+                                   DebugLoc dl, unsigned Base,
+                                   unsigned WordOffset,
+                                   ARMCC::CondCodes Pred, unsigned PredReg) {
+  assert(isThumb1 && "Can only update base register uses for Thumb1!");
+
+  // Start updating any instructions with immediate offsets. Insert a sub before
+  // the first non-updateable instruction (if any).
+  for (; MBBI != MBB.end(); ++MBBI) {
+    if (MBBI->readsRegister(Base)) {
+      unsigned Opc = MBBI->getOpcode();
+      int Offset;
+      bool InsertSub = false;
+
+      if (Opc == ARM::tLDRi  || Opc == ARM::tSTRi  ||
+          Opc == ARM::tLDRHi || Opc == ARM::tSTRHi ||
+          Opc == ARM::tLDRBi || Opc == ARM::tSTRBi) {
+        // Loads and stores with immediate offsets can be updated, but only if
+        // the new offset isn't negative.
+        // The MachineOperand containing the offset immediate is the last one
+        // before predicates.
+        MachineOperand &MO =
+          MBBI->getOperand(MBBI->getDesc().getNumOperands() - 3);
+        // The offsets are scaled by 1, 2 or 4 depending on the Opcode
+        Offset = MO.getImm() - WordOffset * getImmScale(Opc);
+        if (Offset >= 0)
+          MO.setImm(Offset);
+        else
+          InsertSub = true;
+
+      } else if (Opc == ARM::tSUBi8 || Opc == ARM::tADDi8) {
+        // SUB/ADD using this register. Merge it with the update.
+        // If the merged offset is too large, insert a new sub instead.
+        MachineOperand &MO =
+          MBBI->getOperand(MBBI->getDesc().getNumOperands() - 3);
+        Offset = (Opc == ARM::tSUBi8) ?
+          MO.getImm() + WordOffset * 4 :
+          MO.getImm() - WordOffset * 4 ;
+        if (TL->isLegalAddImmediate(Offset)) {
+          MO.setImm(Offset);
+          // The base register has now been reset, so exit early.
+          return;
+        } else {
+          InsertSub = true;
+        }
+
+      } else {
+        // Can't update the instruction.
+        InsertSub = true;
+      }
+
+      if (InsertSub) {
+        // An instruction above couldn't be updated, so insert a sub.
+        AddDefaultT1CC(BuildMI(MBB, MBBI, dl, TII->get(ARM::tSUBi8), Base))
+          .addReg(Base, getKillRegState(true)).addImm(WordOffset * 4)
+          .addImm(Pred).addReg(PredReg);
+        return;
+      }
+    }
+
+    if (MBBI->killsRegister(Base))
+      // Register got killed. Stop updating.
+      return;
+  }
+
+  // The end of the block was reached. This means register liveness escapes the
+  // block, and it's necessary to insert a sub before the last instruction.
+  if (MBB.succ_size() > 0)
+    // But only insert the SUB if there is actually a successor block.
+    // FIXME: Check more carefully if register is live at this point, e.g. by
+    // also examining the successor block's register liveness information.
+    AddDefaultT1CC(BuildMI(MBB, --MBBI, dl, TII->get(ARM::tSUBi8), Base))
+      .addReg(Base, getKillRegState(true)).addImm(WordOffset * 4)
+      .addImm(Pred).addReg(PredReg);
 }
 
 /// MergeOps - Create and insert a LDM or STM with Base as base register and
@@ -296,18 +423,19 @@ ARMLoadStoreOpt::MergeOps(MachineBasicBlock &MBB,
     return false;
 
   ARM_AM::AMSubMode Mode = ARM_AM::ia;
-  // VFP and Thumb2 do not support IB or DA modes.
+  // VFP and Thumb2 do not support IB or DA modes. Thumb1 only supports IA.
   bool isNotVFP = isi32Load(Opcode) || isi32Store(Opcode);
-  bool haveIBAndDA = isNotVFP && !isThumb2;
-  if (Offset == 4 && haveIBAndDA)
+  bool haveIBAndDA = isNotVFP && !isThumb2 && !isThumb1;
+
+  if (Offset == 4 && haveIBAndDA) {
     Mode = ARM_AM::ib;
-  else if (Offset == -4 * (int)NumRegs + 4 && haveIBAndDA)
+  } else if (Offset == -4 * (int)NumRegs + 4 && haveIBAndDA) {
     Mode = ARM_AM::da;
-  else if (Offset == -4 * (int)NumRegs && isNotVFP)
+  } else if (Offset == -4 * (int)NumRegs && isNotVFP && !isThumb1) {
     // VLDM/VSTM do not support DB mode without also updating the base reg.
     Mode = ARM_AM::db;
-  else if (Offset != 0) {
-    // Check if this is a supported opcode before we insert instructions to
+  } else if (Offset != 0) {
+    // Check if this is a supported opcode before inserting instructions to
     // calculate a new base register.
     if (!getLoadStoreMultipleOpcode(Opcode, Mode)) return false;
 
@@ -318,41 +446,98 @@ ARMLoadStoreOpt::MergeOps(MachineBasicBlock &MBB,
       return false;
 
     unsigned NewBase;
-    if (isi32Load(Opcode))
+    if (isi32Load(Opcode)) {
       // If it is a load, then just use one of the destination register to
       // use as the new base.
       NewBase = Regs[NumRegs-1].first;
-    else {
+    } else {
       // Use the scratch register to use as a new base.
       NewBase = Scratch;
       if (NewBase == 0)
         return false;
     }
-    int BaseOpc = !isThumb2 ? ARM::ADDri : ARM::t2ADDri;
+
+    int BaseOpc =
+      isThumb2 ? ARM::t2ADDri :
+      isThumb1 ? ARM::tADDi8  : ARM::ADDri;
+
     if (Offset < 0) {
-      BaseOpc = !isThumb2 ? ARM::SUBri : ARM::t2SUBri;
+      BaseOpc =
+        isThumb2 ? ARM::t2SUBri :
+        isThumb1 ? ARM::tSUBi8  : ARM::SUBri;
       Offset = - Offset;
     }
-    int ImmedOffset = isThumb2
-      ? ARM_AM::getT2SOImmVal(Offset) : ARM_AM::getSOImmVal(Offset);
-    if (ImmedOffset == -1)
-      // FIXME: Try t2ADDri12 or t2SUBri12?
-      return false;  // Probably not worth it then.
-
-    BuildMI(MBB, MBBI, dl, TII->get(BaseOpc), NewBase)
-      .addReg(Base, getKillRegState(BaseKill)).addImm(Offset)
-      .addImm(Pred).addReg(PredReg).addReg(0);
+
+    if (!TL->isLegalAddImmediate(Offset))
+      // FIXME: Try add with register operand?
+      return false; // Probably not worth it then.
+
+    if (isThumb1) {
+      if (Base != NewBase) {
+        // Need to insert a MOV to the new base first.
+        // FIXME: If the immediate fits in 3 bits, use ADD instead.
+        BuildMI(MBB, MBBI, dl, TII->get(ARM::tMOVr), NewBase)
+          .addReg(Base, getKillRegState(BaseKill))
+          .addImm(Pred).addReg(PredReg);
+      }
+      AddDefaultT1CC(BuildMI(MBB, MBBI, dl, TII->get(BaseOpc), NewBase))
+        .addReg(NewBase, getKillRegState(true)).addImm(Offset)
+        .addImm(Pred).addReg(PredReg);
+    } else {
+      BuildMI(MBB, MBBI, dl, TII->get(BaseOpc), NewBase)
+        .addReg(Base, getKillRegState(BaseKill)).addImm(Offset)
+        .addImm(Pred).addReg(PredReg).addReg(0);
+    }
+
     Base = NewBase;
-    BaseKill = true;  // New base is always killed right its use.
+    BaseKill = true; // New base is always killed straight away.
   }
 
   bool isDef = (isi32Load(Opcode) || Opcode == ARM::VLDRS ||
                 Opcode == ARM::VLDRD);
+
+  // Get LS multiple opcode. Note that for Thumb1 this might be an opcode with
+  // base register writeback.
   Opcode = getLoadStoreMultipleOpcode(Opcode, Mode);
   if (!Opcode) return false;
-  MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII->get(Opcode))
-    .addReg(Base, getKillRegState(BaseKill))
-    .addImm(Pred).addReg(PredReg);
+
+  bool Writeback = isThumb1; // Thumb1 LDM/STM have base reg writeback.
+
+  // Exception: If the base register is in the input reglist, Thumb1 LDM is
+  // non-writeback. Check for this.
+  if (Opcode == ARM::tLDRi && isThumb1)
+    for (unsigned I = 0; I < NumRegs; ++I)
+      if (Base == Regs[I].first) {
+        Writeback = false;
+        break;
+      }
+
+  MachineInstrBuilder MIB;
+
+  if (Writeback) {
+    if (Opcode == ARM::tLDMIA)
+      // Update tLDMIA with writeback if necessary.
+      Opcode = ARM::tLDMIA_UPD;
+
+    // The base isn't dead after a merged instruction with writeback. Update
+    // future uses of the base with the added offset (if possible), or reset
+    // the base register as necessary.
+    if (!BaseKill)
+      UpdateBaseRegUses(MBB, MBBI, dl, Base, NumRegs, Pred, PredReg);
+
+    MIB = BuildMI(MBB, MBBI, dl, TII->get(Opcode));
+
+    // Thumb1: we might need to set base writeback when building the MI.
+    MIB.addReg(Base, getDefRegState(true))
+       .addReg(Base, getKillRegState(BaseKill));
+  } else {
+    // No writeback, simply build the MachineInstr.
+    MIB = BuildMI(MBB, MBBI, dl, TII->get(Opcode));
+    MIB.addReg(Base, getKillRegState(BaseKill));
+  }
+
+  MIB.addImm(Pred).addReg(PredReg);
+
   for (unsigned i = 0; i != NumRegs; ++i)
     MIB = MIB.addReg(Regs[i].first, getDefRegState(isDef)
                      | getKillRegState(Regs[i].second));
@@ -492,7 +677,7 @@ void ARMLoadStoreOpt::MergeOpsUpdate(MachineBasicBlock &MBB,
   // affected uses.
   for (SmallVectorImpl<MachineOperand *>::iterator I = UsesOfImpDefs.begin(),
                                                    E = UsesOfImpDefs.end();
-       I != E; ++I)
+                                                   I != E; ++I)
     (*I)->setIsUndef();
 
   for (unsigned i = memOpsBegin; i < memOpsEnd; ++i) {
@@ -589,7 +774,6 @@ ARMLoadStoreOpt::MergeLDR_STR(MachineBasicBlock &MBB, unsigned SIndex,
   bool BaseKill = Loc->findRegisterUseOperandIdx(Base, true) != -1;
   MergeOpsUpdate(MBB, MemOps, SIndex, MemOps.size(), insertAfter, SOffset,
                  Base, BaseKill, Opcode, Pred, PredReg, Scratch, dl, Merges);
-  return;
 }
 
 static bool definesCPSR(MachineInstr *MI) {
@@ -616,6 +800,7 @@ static bool isMatchingDecrement(MachineInstr *MI, unsigned Base,
   bool CheckCPSRDef = false;
   switch (MI->getOpcode()) {
   default: return false;
+  case ARM::tSUBi8:
   case ARM::t2SUBri:
   case ARM::SUBri:
     CheckCPSRDef = true;
@@ -628,10 +813,11 @@ static bool isMatchingDecrement(MachineInstr *MI, unsigned Base,
   if (Bytes == 0 || (Limit && Bytes >= Limit))
     return false;
 
-  unsigned Scale = (MI->getOpcode() == ARM::tSUBspi) ? 4 : 1; // FIXME
+  unsigned Scale = (MI->getOpcode() == ARM::tSUBspi ||
+                    MI->getOpcode() == ARM::tSUBi8) ? 4 : 1; // FIXME
   if (!(MI->getOperand(0).getReg() == Base &&
         MI->getOperand(1).getReg() == Base &&
-        (MI->getOperand(2).getImm()*Scale) == Bytes &&
+        (MI->getOperand(2).getImm() * Scale) == Bytes &&
         getInstrPredicate(MI, MyPredReg) == Pred &&
         MyPredReg == PredReg))
     return false;
@@ -649,6 +835,7 @@ static bool isMatchingIncrement(MachineInstr *MI, unsigned Base,
   bool CheckCPSRDef = false;
   switch (MI->getOpcode()) {
   default: return false;
+  case ARM::tADDi8:
   case ARM::t2ADDri:
   case ARM::ADDri:
     CheckCPSRDef = true;
@@ -661,10 +848,11 @@ static bool isMatchingIncrement(MachineInstr *MI, unsigned Base,
     // Make sure the offset fits in 8 bits.
     return false;
 
-  unsigned Scale = (MI->getOpcode() == ARM::tADDspi) ? 4 : 1; // FIXME
+  unsigned Scale = (MI->getOpcode() == ARM::tADDspi ||
+                    MI->getOpcode() == ARM::tADDi8) ? 4 : 1; // FIXME
   if (!(MI->getOperand(0).getReg() == Base &&
         MI->getOperand(1).getReg() == Base &&
-        (MI->getOperand(2).getImm()*Scale) == Bytes &&
+        (MI->getOperand(2).getImm() * Scale) == Bytes &&
         getInstrPredicate(MI, MyPredReg) == Pred &&
         MyPredReg == PredReg))
     return false;
@@ -677,6 +865,8 @@ static inline unsigned getLSMultipleTransferSize(MachineInstr *MI) {
   default: return 0;
   case ARM::LDRi12:
   case ARM::STRi12:
+  case ARM::tLDRi:
+  case ARM::tSTRi:
   case ARM::t2LDRi8:
   case ARM::t2LDRi12:
   case ARM::t2STRi8:
@@ -695,6 +885,9 @@ static inline unsigned getLSMultipleTransferSize(MachineInstr *MI) {
   case ARM::STMDA:
   case ARM::STMDB:
   case ARM::STMIB:
+  case ARM::tLDMIA:
+  case ARM::tLDMIA_UPD:
+  case ARM::tSTMIA_UPD:
   case ARM::t2LDMIA:
   case ARM::t2LDMDB:
   case ARM::t2STMIA:
@@ -791,6 +984,9 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLSMultiple(MachineBasicBlock &MBB,
                                                MachineBasicBlock::iterator MBBI,
                                                bool &Advance,
                                                MachineBasicBlock::iterator &I) {
+  // Thumb1 is already using updating loads/stores.
+  if (isThumb1) return false;
+
   MachineInstr *MI = MBBI;
   unsigned Base = MI->getOperand(0).getReg();
   bool BaseKill = MI->getOperand(0).isKill();
@@ -927,6 +1123,10 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineBasicBlock &MBB,
                                                const TargetInstrInfo *TII,
                                                bool &Advance,
                                                MachineBasicBlock::iterator &I) {
+  // Thumb1 doesn't have updating LDR/STR.
+  // FIXME: Use LDM/STM with single register instead.
+  if (isThumb1) return false;
+
   MachineInstr *MI = MBBI;
   unsigned Base = MI->getOperand(1).getReg();
   bool BaseKill = MI->getOperand(1).isKill();
@@ -1002,7 +1202,7 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineBasicBlock &MBB,
     return false;
 
   if (isAM5) {
-    // VLDM[SD}_UPD, VSTM[SD]_UPD
+    // VLDM[SD]_UPD, VSTM[SD]_UPD
     // (There are no base-updating versions of VLDR/VSTR instructions, but the
     // updating load/store-multiple instructions can be used with only one
     // register.)
@@ -1100,6 +1300,8 @@ static bool isMemoryOp(const MachineInstr *MI) {
     return MI->getOperand(1).isReg();
   case ARM::LDRi12:
   case ARM::STRi12:
+  case ARM::tLDRi:
+  case ARM::tSTRi:
   case ARM::t2LDRi8:
   case ARM::t2LDRi12:
   case ARM::t2STRi8:
@@ -1137,6 +1339,10 @@ static int getMemoryOpOffset(const MachineInstr *MI) {
       Opcode == ARM::LDRi12   || Opcode == ARM::STRi12)
     return OffField;
 
+  // Thumb1 immediate offsets are scaled by 4
+  if (Opcode == ARM::tLDRi || Opcode == ARM::tSTRi)
+    return OffField * 4;
+
   int Offset = isAM3 ? ARM_AM::getAM3Offset(OffField)
     : ARM_AM::getAM5Offset(OffField) * 4;
   if (isAM3) {
@@ -1408,16 +1614,20 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
       if (MBBI == E)
         // Reach the end of the block, try merging the memory instructions.
         TryMerge = true;
-    } else
+    } else {
       TryMerge = true;
+    }
 
     if (TryMerge) {
       if (NumMemOps > 1) {
         // Try to find a free register to use as a new base in case it's needed.
         // First advance to the instruction just before the start of the chain.
         AdvanceRS(MBB, MemOps);
+
         // Find a scratch register.
-        unsigned Scratch = RS->FindUnusedReg(&ARM::GPRRegClass);
+        unsigned Scratch =
+          RS->FindUnusedReg(isThumb1 ? &ARM::tGPRRegClass : &ARM::GPRRegClass);
+
         // Process the load / store instructions.
         RS->forward(std::prev(MBBI));
 
@@ -1483,6 +1693,8 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
 /// =>
 ///   ldmfd sp!, {..., pc}
 bool ARMLoadStoreOpt::MergeReturnIntoLDM(MachineBasicBlock &MBB) {
+  // Thumb1 LDM doesn't allow high registers.
+  if (isThumb1) return false;
   if (MBB.empty()) return false;
 
   MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
@@ -1513,12 +1725,14 @@ bool ARMLoadStoreOpt::MergeReturnIntoLDM(MachineBasicBlock &MBB) {
 
 bool ARMLoadStoreOpt::runOnMachineFunction(MachineFunction &Fn) {
   const TargetMachine &TM = Fn.getTarget();
+  TL = TM.getTargetLowering();
   AFI = Fn.getInfo<ARMFunctionInfo>();
   TII = TM.getInstrInfo();
   TRI = TM.getRegisterInfo();
   STI = &TM.getSubtarget<ARMSubtarget>();
   RS = new RegScavenger();
   isThumb2 = AFI->isThumb2Function();
+  isThumb1 = AFI->isThumbFunction() && !isThumb2;
 
   bool Modified = false;
   for (MachineFunction::iterator MFI = Fn.begin(), E = Fn.end(); MFI != E;
@@ -1666,11 +1880,11 @@ ARMPreAllocLoadStoreOpt::CanFormLdStDWord(MachineInstr *Op0, MachineInstr *Op1,
   // FIXME: VLDRS / VSTRS -> VLDRD / VSTRD
   unsigned Scale = 1;
   unsigned Opcode = Op0->getOpcode();
-  if (Opcode == ARM::LDRi12)
+  if (Opcode == ARM::LDRi12) {
     NewOpc = ARM::LDRD;
-  else if (Opcode == ARM::STRi12)
+  } else if (Opcode == ARM::STRi12) {
     NewOpc = ARM::STRD;
-  else if (Opcode == ARM::t2LDRi8 || Opcode == ARM::t2LDRi12) {
+  } else if (Opcode == ARM::t2LDRi8 || Opcode == ARM::t2LDRi12) {
     NewOpc = ARM::t2LDRDi8;
     Scale = 4;
     isT2 = true;
@@ -1678,8 +1892,9 @@ ARMPreAllocLoadStoreOpt::CanFormLdStDWord(MachineInstr *Op0, MachineInstr *Op1,
     NewOpc = ARM::t2STRDi8;
     Scale = 4;
     isT2 = true;
-  } else
+  } else {
     return false;
+  }
 
   // Make sure the base address satisfies i64 ld / st alignment requirement.
   // At the moment, we ignore the memoryoperand's value.
@@ -1746,8 +1961,8 @@ bool ARMPreAllocLoadStoreOpt::RescheduleOps(MachineBasicBlock *MBB,
   while (Ops.size() > 1) {
     unsigned FirstLoc = ~0U;
     unsigned LastLoc = 0;
-    MachineInstr *FirstOp = 0;
-    MachineInstr *LastOp = 0;
+    MachineInstr *FirstOp = nullptr;
+    MachineInstr *LastOp = nullptr;
     int LastOffset = 0;
     unsigned LastOpcode = 0;
     unsigned LastBytes = 0;
diff --git a/lib/Target/ARM/ARMOptimizeBarriersPass.cpp b/lib/Target/ARM/ARMOptimizeBarriersPass.cpp
index 20619fa..2a49255 100644
--- a/lib/Target/ARM/ARMOptimizeBarriersPass.cpp
+++ b/lib/Target/ARM/ARMOptimizeBarriersPass.cpp
@@ -8,8 +8,6 @@
 //
 //===------------------------------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "double barriers"
-
 #include "ARM.h"
 #include "ARMMachineFunctionInfo.h"
 #include "ARMInstrInfo.h"
@@ -17,6 +15,8 @@
 #include "llvm/CodeGen/MachineFunctionPass.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "double barriers"
+
 STATISTIC(NumDMBsRemoved, "Number of DMBs removed");
 
 namespace {
@@ -25,9 +25,9 @@ public:
   static char ID;
   ARMOptimizeBarriersPass() : MachineFunctionPass(ID) {}
 
-  virtual bool runOnMachineFunction(MachineFunction &Fn);
+  bool runOnMachineFunction(MachineFunction &Fn) override;
 
-  virtual const char *getPassName() const {
+  const char *getPassName() const override {
     return "optimise barriers pass";
   }
 
diff --git a/lib/Target/ARM/ARMRegisterInfo.td b/lib/Target/ARM/ARMRegisterInfo.td
index 7f0fe05..b290e7f 100644
--- a/lib/Target/ARM/ARMRegisterInfo.td
+++ b/lib/Target/ARM/ARMRegisterInfo.td
@@ -116,13 +116,13 @@ def D15 : ARMReg<15, "d15", [S30, S31]>, DwarfRegNum<[271]>;
 }
 
 // VFP3 defines 16 additional double registers
-def D16 : ARMFReg<16, "d16">, DwarfRegNum<[272]>; 
+def D16 : ARMFReg<16, "d16">, DwarfRegNum<[272]>;
 def D17 : ARMFReg<17, "d17">, DwarfRegNum<[273]>;
 def D18 : ARMFReg<18, "d18">, DwarfRegNum<[274]>;
 def D19 : ARMFReg<19, "d19">, DwarfRegNum<[275]>;
 def D20 : ARMFReg<20, "d20">, DwarfRegNum<[276]>;
 def D21 : ARMFReg<21, "d21">, DwarfRegNum<[277]>;
-def D22 : ARMFReg<22, "d22">, DwarfRegNum<[278]>; 
+def D22 : ARMFReg<22, "d22">, DwarfRegNum<[278]>;
 def D23 : ARMFReg<23, "d23">, DwarfRegNum<[279]>;
 def D24 : ARMFReg<24, "d24">, DwarfRegNum<[280]>;
 def D25 : ARMFReg<25, "d25">, DwarfRegNum<[281]>;
@@ -158,11 +158,11 @@ def Q15 : ARMReg<15, "q15", [D30, D31]>;
 // Current Program Status Register.
 // We model fpscr with two registers: FPSCR models the control bits and will be
 // reserved. FPSCR_NZCV models the flag bits and will be unreserved. APSR_NZCV
-// models the APSR when it's accessed by some special instructions. In such cases 
+// models the APSR when it's accessed by some special instructions. In such cases
 // it has the same encoding as PC.
 def CPSR       : ARMReg<0,  "cpsr">;
 def APSR       : ARMReg<1,  "apsr">;
-def APSR_NZCV  : ARMReg<15, "apsr_nzcv">; 
+def APSR_NZCV  : ARMReg<15, "apsr_nzcv">;
 def SPSR       : ARMReg<2,  "spsr">;
 def FPSCR      : ARMReg<3,  "fpscr">;
 def FPSCR_NZCV : ARMReg<3,  "fpscr_nzcv"> {
diff --git a/lib/Target/ARM/ARMScheduleV6.td b/lib/Target/ARM/ARMScheduleV6.td
index 0ace9bc..57d0bfb 100644
--- a/lib/Target/ARM/ARMScheduleV6.td
+++ b/lib/Target/ARM/ARMScheduleV6.td
@@ -93,7 +93,7 @@ def ARMV6Itineraries : ProcessorItineraries<
   InstrItinData<IIC_iMAC32   , [InstrStage<2, [V6_Pipe]>], [5, 1, 1, 2]>,
   InstrItinData<IIC_iMUL64   , [InstrStage<3, [V6_Pipe]>], [6, 1, 1]>,
   InstrItinData<IIC_iMAC64   , [InstrStage<3, [V6_Pipe]>], [6, 1, 1, 2]>,
-  
+
   // Integer load pipeline
   //
   // Immediate offset
@@ -181,7 +181,7 @@ def ARMV6Itineraries : ProcessorItineraries<
   //
   // Store multiple + update
   InstrItinData<IIC_iStore_mu , [InstrStage<3, [V6_Pipe]>], [2]>,
-  
+
   // Branch
   //
   // no delay slots, so the latency of a branch is unimportant
diff --git a/lib/Target/ARM/ARMSelectionDAGInfo.cpp b/lib/Target/ARM/ARMSelectionDAGInfo.cpp
index ba3cf4d..008ad64 100644
--- a/lib/Target/ARM/ARMSelectionDAGInfo.cpp
+++ b/lib/Target/ARM/ARMSelectionDAGInfo.cpp
@@ -11,12 +11,13 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-selectiondag-info"
 #include "ARMTargetMachine.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/IR/DerivedTypes.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-selectiondag-info"
+
 ARMSelectionDAGInfo::ARMSelectionDAGInfo(const TargetMachine &TM)
   : TargetSelectionDAGInfo(TM),
     Subtarget(&TM.getSubtarget<ARMSubtarget>()) {
@@ -52,9 +53,10 @@ ARMSelectionDAGInfo::EmitTargetCodeForMemcpy(SelectionDAG &DAG, SDLoc dl,
   EVT VT = MVT::i32;
   unsigned VTSize = 4;
   unsigned i = 0;
-  const unsigned MAX_LOADS_IN_LDM = 6;
-  SDValue TFOps[MAX_LOADS_IN_LDM];
-  SDValue Loads[MAX_LOADS_IN_LDM];
+  // Emit a maximum of 4 loads in Thumb1 since we have fewer registers
+  const unsigned MAX_LOADS_IN_LDM = Subtarget->isThumb1Only() ? 4 : 6;
+  SDValue TFOps[6];
+  SDValue Loads[6];
   uint64_t SrcOff = 0, DstOff = 0;
 
   // Emit up to MAX_LOADS_IN_LDM loads, then a TokenFactor barrier, then the
@@ -71,7 +73,8 @@ ARMSelectionDAGInfo::EmitTargetCodeForMemcpy(SelectionDAG &DAG, SDLoc dl,
       TFOps[i] = Loads[i].getValue(1);
       SrcOff += VTSize;
     }
-    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, &TFOps[0], i);
+    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                        makeArrayRef(TFOps, i));
 
     for (i = 0;
          i < MAX_LOADS_IN_LDM && EmittedNumMemOps + i < NumMemOps; ++i) {
@@ -82,7 +85,8 @@ ARMSelectionDAGInfo::EmitTargetCodeForMemcpy(SelectionDAG &DAG, SDLoc dl,
                               isVolatile, false, 0);
       DstOff += VTSize;
     }
-    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, &TFOps[0], i);
+    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                        makeArrayRef(TFOps, i));
 
     EmittedNumMemOps += i;
   }
@@ -112,7 +116,8 @@ ARMSelectionDAGInfo::EmitTargetCodeForMemcpy(SelectionDAG &DAG, SDLoc dl,
     SrcOff += VTSize;
     BytesLeft -= VTSize;
   }
-  Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, &TFOps[0], i);
+  Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                      makeArrayRef(TFOps, i));
 
   i = 0;
   BytesLeft = BytesLeftSave;
@@ -133,7 +138,8 @@ ARMSelectionDAGInfo::EmitTargetCodeForMemcpy(SelectionDAG &DAG, SDLoc dl,
     DstOff += VTSize;
     BytesLeft -= VTSize;
   }
-  return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, &TFOps[0], i);
+  return DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                     makeArrayRef(TFOps, i));
 }
 
 // Adjust parameters for memset, EABI uses format (ptr, size, value),
@@ -146,7 +152,8 @@ EmitTargetCodeForMemset(SelectionDAG &DAG, SDLoc dl,
                         unsigned Align, bool isVolatile,
                         MachinePointerInfo DstPtrInfo) const {
   // Use default for non-AAPCS (or MachO) subtargets
-  if (!Subtarget->isAAPCS_ABI() || Subtarget->isTargetMachO())
+  if (!Subtarget->isAAPCS_ABI() || Subtarget->isTargetMachO() ||
+      Subtarget->isTargetWindows())
     return SDValue();
 
   const ARMTargetLowering &TLI =
@@ -179,22 +186,14 @@ EmitTargetCodeForMemset(SelectionDAG &DAG, SDLoc dl,
   Args.push_back(Entry);
 
   // Emit __eabi_memset call
-  TargetLowering::CallLoweringInfo CLI(Chain,
-                    Type::getVoidTy(*DAG.getContext()), // return type
-                    false, // return sign ext
-                    false, // return zero ext
-                    false, // is var arg
-                    false, // is in regs
-                    0,     // number of fixed arguments
-                    TLI.getLibcallCallingConv(RTLIB::MEMSET), // call conv
-                    false, // is tail call
-                    false, // does not return
-                    false, // is return val used
-                    DAG.getExternalSymbol(TLI.getLibcallName(RTLIB::MEMSET),
-                                          TLI.getPointerTy()), // callee
-                    Args, DAG, dl);
-  std::pair<SDValue,SDValue> CallResult =
-    TLI.LowerCallTo(CLI);
-
+  TargetLowering::CallLoweringInfo CLI(DAG);
+  CLI.setDebugLoc(dl).setChain(Chain)
+    .setCallee(TLI.getLibcallCallingConv(RTLIB::MEMSET),
+               Type::getVoidTy(*DAG.getContext()),
+               DAG.getExternalSymbol(TLI.getLibcallName(RTLIB::MEMSET),
+                                     TLI.getPointerTy()), &Args, 0)
+    .setDiscardResult();
+
+  std::pair<SDValue,SDValue> CallResult = TLI.LowerCallTo(CLI);
   return CallResult.second;
 }
diff --git a/lib/Target/ARM/ARMSubtarget.cpp b/lib/Target/ARM/ARMSubtarget.cpp
index 73e2018..5b204f6 100644
--- a/lib/Target/ARM/ARMSubtarget.cpp
+++ b/lib/Target/ARM/ARMSubtarget.cpp
@@ -21,12 +21,14 @@
 #include "llvm/Target/TargetInstrInfo.h"
 #include "llvm/Target/TargetOptions.h"
 
+using namespace llvm;
+
+#define DEBUG_TYPE "arm-subtarget"
+
 #define GET_SUBTARGETINFO_TARGET_DESC
 #define GET_SUBTARGETINFO_CTOR
 #include "ARMGenSubtargetInfo.inc"
 
-using namespace llvm;
-
 static cl::opt<bool>
 ReserveR9("arm-reserve-r9", cl::Hidden,
           cl::desc("Reserve R9, making it unavailable as GPR"));
diff --git a/lib/Target/ARM/ARMSubtarget.h b/lib/Target/ARM/ARMSubtarget.h
index 3855419..38536b2 100644
--- a/lib/Target/ARM/ARMSubtarget.h
+++ b/lib/Target/ARM/ARMSubtarget.h
@@ -31,7 +31,7 @@ class TargetOptions;
 class ARMSubtarget : public ARMGenSubtargetInfo {
 protected:
   enum ARMProcFamilyEnum {
-    Others, CortexA5, CortexA7, CortexA8, CortexA9, CortexA12, CortexA15, 
+    Others, CortexA5, CortexA7, CortexA8, CortexA9, CortexA12, CortexA15,
     CortexR5, Swift, CortexA53, CortexA57, Krait
   };
   enum ARMProcClassEnum {
@@ -242,9 +242,7 @@ protected:
   /// getMaxInlineSizeThreshold - Returns the maximum memset / memcpy size
   /// that still makes it profitable to inline the call.
   unsigned getMaxInlineSizeThreshold() const {
-    // FIXME: For now, we don't lower memcpy's to loads / stores for Thumb1.
-    // Change this once Thumb1 ldmia / stmia support is added.
-    return isThumb1Only() ? 0 : 64;
+    return 64;
   }
   /// ParseSubtargetFeatures - Parses features string setting specified
   /// subtarget options.  Definition of function is auto generated by tblgen.
@@ -396,7 +394,7 @@ public:
   bool isLittle() const { return IsLittle; }
 
   unsigned getMispredictionPenalty() const;
-  
+
   /// This function returns true if the target has sincos() routine in its
   /// compiler runtime or math libraries.
   bool hasSinCos() const;
diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 4ae539a..8876227 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -228,7 +228,7 @@ TargetPassConfig *ARMBaseTargetMachine::createPassConfig(PassManagerBase &PM) {
 bool ARMPassConfig::addPreISel() {
   const ARMSubtarget *Subtarget = &getARMSubtarget();
   if (Subtarget->hasAnyDataBarrier() && !Subtarget->isThumb1Only())
-    addPass(createARMAtomicExpandPass(TM));
+    addPass(createAtomicExpandLoadLinkedPass(TM));
 
   if (TM->getOptLevel() != CodeGenOpt::None)
     addPass(createGlobalMergePass(TM));
@@ -247,8 +247,7 @@ bool ARMPassConfig::addInstSelector() {
 }
 
 bool ARMPassConfig::addPreRegAlloc() {
-  // FIXME: temporarily disabling load / store optimization pass for Thumb1.
-  if (getOptLevel() != CodeGenOpt::None && !getARMSubtarget().isThumb1Only())
+  if (getOptLevel() != CodeGenOpt::None)
     addPass(createARMLoadStoreOptimizationPass(true));
   if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isCortexA9())
     addPass(createMLxExpansionPass());
@@ -262,12 +261,10 @@ bool ARMPassConfig::addPreRegAlloc() {
 }
 
 bool ARMPassConfig::addPreSched2() {
-  // FIXME: temporarily disabling load / store optimization pass for Thumb1.
   if (getOptLevel() != CodeGenOpt::None) {
-    if (!getARMSubtarget().isThumb1Only()) {
-      addPass(createARMLoadStoreOptimizationPass());
-      printAndVerify("After ARM load / store optimizer");
-    }
+    addPass(createARMLoadStoreOptimizationPass());
+    printAndVerify("After ARM load / store optimizer");
+
     if (getARMSubtarget().hasNEON())
       addPass(createExecutionDependencyFixPass(&ARM::DPRRegClass));
   }
diff --git a/lib/Target/ARM/ARMTargetMachine.h b/lib/Target/ARM/ARMTargetMachine.h
index 0c80a95..664c992 100644
--- a/lib/Target/ARM/ARMTargetMachine.h
+++ b/lib/Target/ARM/ARMTargetMachine.h
@@ -23,7 +23,6 @@
 #include "Thumb1FrameLowering.h"
 #include "Thumb1InstrInfo.h"
 #include "Thumb2InstrInfo.h"
-#include "llvm/ADT/OwningPtr.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/Target/TargetMachine.h"
@@ -102,7 +101,7 @@ class ARMTargetMachine : public ARMBaseTargetMachine {
 /// ARMLETargetMachine - ARM little endian target machine.
 ///
 class ARMLETargetMachine : public ARMTargetMachine {
-  virtual void anchor();
+  void anchor() override;
 public:
   ARMLETargetMachine(const Target &T, StringRef TT,
                      StringRef CPU, StringRef FS, const TargetOptions &Options,
@@ -113,7 +112,7 @@ public:
 /// ARMBETargetMachine - ARM big endian target machine.
 ///
 class ARMBETargetMachine : public ARMTargetMachine {
-  virtual void anchor();
+  void anchor() override;
 public:
   ARMBETargetMachine(const Target &T, StringRef TT,
                      StringRef CPU, StringRef FS, const TargetOptions &Options,
@@ -128,12 +127,12 @@ public:
 class ThumbTargetMachine : public ARMBaseTargetMachine {
   virtual void anchor();
   // Either Thumb1InstrInfo or Thumb2InstrInfo.
-  OwningPtr<ARMBaseInstrInfo> InstrInfo;
+  std::unique_ptr<ARMBaseInstrInfo> InstrInfo;
   const DataLayout    DL;   // Calculates type size & alignment
   ARMTargetLowering   TLInfo;
   ARMSelectionDAGInfo TSInfo;
   // Either Thumb1FrameLowering or ARMFrameLowering.
-  OwningPtr<ARMFrameLowering> FrameLowering;
+  std::unique_ptr<ARMFrameLowering> FrameLowering;
 public:
   ThumbTargetMachine(const Target &T, StringRef TT,
                      StringRef CPU, StringRef FS,
@@ -169,7 +168,7 @@ public:
 /// ThumbLETargetMachine - Thumb little endian target machine.
 ///
 class ThumbLETargetMachine : public ThumbTargetMachine {
-  virtual void anchor();
+  void anchor() override;
 public:
   ThumbLETargetMachine(const Target &T, StringRef TT,
                      StringRef CPU, StringRef FS, const TargetOptions &Options,
@@ -180,10 +179,10 @@ public:
 /// ThumbBETargetMachine - Thumb big endian target machine.
 ///
 class ThumbBETargetMachine : public ThumbTargetMachine {
-  virtual void anchor();
+  void anchor() override;
 public:
-  ThumbBETargetMachine(const Target &T, StringRef TT,
-                       StringRef CPU, StringRef FS, const TargetOptions &Options,
+  ThumbBETargetMachine(const Target &T, StringRef TT, StringRef CPU,
+                       StringRef FS, const TargetOptions &Options,
                        Reloc::Model RM, CodeModel::Model CM,
                        CodeGenOpt::Level OL);
 };
diff --git a/lib/Target/ARM/ARMTargetObjectFile.cpp b/lib/Target/ARM/ARMTargetObjectFile.cpp
index 3379f85..48238bf 100644
--- a/lib/Target/ARM/ARMTargetObjectFile.cpp
+++ b/lib/Target/ARM/ARMTargetObjectFile.cpp
@@ -11,6 +11,7 @@
 #include "ARMSubtarget.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/IR/Mangler.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSectionELF.h"
@@ -31,7 +32,7 @@ void ARMElfTargetObjectFile::Initialize(MCContext &Ctx,
   InitializeELF(isAAPCS_ABI);
 
   if (isAAPCS_ABI) {
-    LSDASection = NULL;
+    LSDASection = nullptr;
   }
 
   AttributesSection =
@@ -45,6 +46,10 @@ const MCExpr *ARMElfTargetObjectFile::getTTypeGlobalReference(
     const GlobalValue *GV, unsigned Encoding, Mangler &Mang,
     const TargetMachine &TM, MachineModuleInfo *MMI,
     MCStreamer &Streamer) const {
+  if (TM.getMCAsmInfo()->getExceptionHandlingType() != ExceptionHandling::ARM)
+    return TargetLoweringObjectFileELF::getTTypeGlobalReference(
+        GV, Encoding, Mang, TM, MMI, Streamer);
+
   assert(Encoding == DW_EH_PE_absptr && "Can handle absptr encoding only");
 
   return MCSymbolRefExpr::Create(TM.getSymbol(GV, Mang),
diff --git a/lib/Target/ARM/ARMTargetObjectFile.h b/lib/Target/ARM/ARMTargetObjectFile.h
index 5f8d612..c926421 100644
--- a/lib/Target/ARM/ARMTargetObjectFile.h
+++ b/lib/Target/ARM/ARMTargetObjectFile.h
@@ -23,7 +23,7 @@ protected:
 public:
   ARMElfTargetObjectFile() :
     TargetLoweringObjectFileELF(),
-    AttributesSection(NULL)
+    AttributesSection(nullptr)
   {}
 
   void Initialize(MCContext &Ctx, const TargetMachine &TM) override;
diff --git a/lib/Target/ARM/ARMTargetTransformInfo.cpp b/lib/Target/ARM/ARMTargetTransformInfo.cpp
index d3b43cd..57df7da 100644
--- a/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -14,7 +14,6 @@
 ///
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "armtti"
 #include "ARM.h"
 #include "ARMTargetMachine.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
@@ -23,8 +22,10 @@
 #include "llvm/Target/TargetLowering.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "armtti"
+
 // Declare the pass initialization routine locally as target-specific passes
-// don't havve a target-wide initialization entry point, and so we rely on the
+// don't have a target-wide initialization entry point, and so we rely on the
 // pass constructor initialization.
 namespace llvm {
 void initializeARMTTIPass(PassRegistry &);
@@ -42,7 +43,7 @@ class ARMTTI final : public ImmutablePass, public TargetTransformInfo {
   unsigned getScalarizationOverhead(Type *Ty, bool Insert, bool Extract) const;
 
 public:
-  ARMTTI() : ImmutablePass(ID), TM(0), ST(0), TLI(0) {
+  ARMTTI() : ImmutablePass(ID), TM(nullptr), ST(nullptr), TLI(nullptr) {
     llvm_unreachable("This pass cannot be directly constructed");
   }
 
diff --git a/lib/Target/ARM/Android.mk b/lib/Target/ARM/Android.mk
index 4be95aa..095955b 100644
--- a/lib/Target/ARM/Android.mk
+++ b/lib/Target/ARM/Android.mk
@@ -17,7 +17,6 @@ arm_codegen_TBLGEN_TABLES := \
 arm_codegen_SRC_FILES := \
   A15SDOptimizer.cpp \
   ARMAsmPrinter.cpp \
-  ARMAtomicExpandPass.cpp \
   ARMBaseInstrInfo.cpp \
   ARMBaseRegisterInfo.cpp \
   ARMCodeEmitter.cpp \
diff --git a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index 9c57a24..5cdf394 100644
--- a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -13,7 +13,6 @@
 #include "MCTargetDesc/ARMArchName.h"
 #include "MCTargetDesc/ARMBaseInfo.h"
 #include "MCTargetDesc/ARMMCExpr.h"
-#include "llvm/ADT/OwningPtr.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
@@ -23,9 +22,7 @@
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCDisassembler.h"
-#include "llvm/MC/MCELF.h"
 #include "llvm/MC/MCELFStreamer.h"
-#include "llvm/MC/MCELFSymbolFlags.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCInstrDesc.h"
@@ -345,7 +342,8 @@ public:
   };
 
   ARMAsmParser(MCSubtargetInfo &_STI, MCAsmParser &_Parser,
-               const MCInstrInfo &MII)
+               const MCInstrInfo &MII,
+               const MCTargetOptions &Options)
       : MCTargetAsmParser(), STI(_STI), Parser(_Parser), MII(MII), UC(_Parser) {
     MCAsmParserExtension::Initialize(_Parser);
 
@@ -416,7 +414,7 @@ class ARMOperand : public MCParsedAsmOperand {
     k_Token
   } Kind;
 
-  SMLoc StartLoc, EndLoc;
+  SMLoc StartLoc, EndLoc, AlignmentLoc;
   SmallVector<unsigned, 8> Registers;
 
   struct CCOp {
@@ -633,6 +631,12 @@ public:
   /// operand.
   SMRange getLocRange() const { return SMRange(StartLoc, EndLoc); }
 
+  /// getAlignmentLoc - Get the location of the Alignment token of this operand.
+  SMLoc getAlignmentLoc() const {
+    assert(Kind == k_Memory && "Invalid access!");
+    return AlignmentLoc;
+  }
+
   ARMCC::CondCodes getCondCode() const {
     assert(Kind == k_CondCode && "Invalid access!");
     return CC.Val;
@@ -1089,12 +1093,12 @@ public:
   bool isPostIdxReg() const {
     return Kind == k_PostIndexRegister && PostIdxReg.ShiftTy ==ARM_AM::no_shift;
   }
-  bool isMemNoOffset(bool alignOK = false) const {
+  bool isMemNoOffset(bool alignOK = false, unsigned Alignment = 0) const {
     if (!isMem())
       return false;
     // No offset of any kind.
-    return Memory.OffsetRegNum == 0 && Memory.OffsetImm == 0 &&
-     (alignOK || Memory.Alignment == 0);
+    return Memory.OffsetRegNum == 0 && Memory.OffsetImm == nullptr &&
+     (alignOK || Memory.Alignment == Alignment);
   }
   bool isMemPCRelImm12() const {
     if (!isMem() || Memory.OffsetRegNum != 0 || Memory.Alignment != 0)
@@ -1110,6 +1114,65 @@ public:
   bool isAlignedMemory() const {
     return isMemNoOffset(true);
   }
+  bool isAlignedMemoryNone() const {
+    return isMemNoOffset(false, 0);
+  }
+  bool isDupAlignedMemoryNone() const {
+    return isMemNoOffset(false, 0);
+  }
+  bool isAlignedMemory16() const {
+    if (isMemNoOffset(false, 2)) // alignment in bytes for 16-bits is 2.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isDupAlignedMemory16() const {
+    if (isMemNoOffset(false, 2)) // alignment in bytes for 16-bits is 2.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isAlignedMemory32() const {
+    if (isMemNoOffset(false, 4)) // alignment in bytes for 32-bits is 4.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isDupAlignedMemory32() const {
+    if (isMemNoOffset(false, 4)) // alignment in bytes for 32-bits is 4.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isAlignedMemory64() const {
+    if (isMemNoOffset(false, 8)) // alignment in bytes for 64-bits is 8.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isDupAlignedMemory64() const {
+    if (isMemNoOffset(false, 8)) // alignment in bytes for 64-bits is 8.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isAlignedMemory64or128() const {
+    if (isMemNoOffset(false, 8)) // alignment in bytes for 64-bits is 8.
+      return true;
+    if (isMemNoOffset(false, 16)) // alignment in bytes for 128-bits is 16.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isDupAlignedMemory64or128() const {
+    if (isMemNoOffset(false, 8)) // alignment in bytes for 64-bits is 8.
+      return true;
+    if (isMemNoOffset(false, 16)) // alignment in bytes for 128-bits is 16.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
+  bool isAlignedMemory64or128or256() const {
+    if (isMemNoOffset(false, 8)) // alignment in bytes for 64-bits is 8.
+      return true;
+    if (isMemNoOffset(false, 16)) // alignment in bytes for 128-bits is 16.
+      return true;
+    if (isMemNoOffset(false, 32)) // alignment in bytes for 256-bits is 32.
+      return true;
+    return isMemNoOffset(false, 0);
+  }
   bool isAddrMode2() const {
     if (!isMem() || Memory.Alignment != 0) return false;
     // Check for register offset.
@@ -1545,7 +1608,10 @@ public:
   }
 
   bool isNEONi16splat() const {
-    if (!isImm()) return false;
+    if (isNEONByteReplicate(2))
+      return false; // Leave that for bytes replication and forbid by default.
+    if (!isImm())
+      return false;
     const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
     // Must be a constant.
     if (!CE) return false;
@@ -1555,7 +1621,10 @@ public:
   }
 
   bool isNEONi32splat() const {
-    if (!isImm()) return false;
+    if (isNEONByteReplicate(4))
+      return false; // Leave that for bytes replication and forbid by default.
+    if (!isImm())
+      return false;
     const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
     // Must be a constant.
     if (!CE) return false;
@@ -1567,11 +1636,36 @@ public:
       (Value >= 0x01000000 && Value <= 0xff000000);
   }
 
+  bool isNEONByteReplicate(unsigned NumBytes) const {
+    if (!isImm())
+      return false;
+    const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
+    // Must be a constant.
+    if (!CE)
+      return false;
+    int64_t Value = CE->getValue();
+    if (!Value)
+      return false; // Don't bother with zero.
+
+    unsigned char B = Value & 0xff;
+    for (unsigned i = 1; i < NumBytes; ++i) {
+      Value >>= 8;
+      if ((Value & 0xff) != B)
+        return false;
+    }
+    return true;
+  }
+  bool isNEONi16ByteReplicate() const { return isNEONByteReplicate(2); }
+  bool isNEONi32ByteReplicate() const { return isNEONByteReplicate(4); }
   bool isNEONi32vmov() const {
-    if (!isImm()) return false;
+    if (isNEONByteReplicate(4))
+      return false; // Let it to be classified as byte-replicate case.
+    if (!isImm())
+      return false;
     const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
     // Must be a constant.
-    if (!CE) return false;
+    if (!CE)
+      return false;
     int64_t Value = CE->getValue();
     // i32 value with set bits only in one byte X000, 0X00, 00X0, or 000X,
     // for VMOV/VMVN only, 00Xf or 0Xff are also accepted.
@@ -1612,7 +1706,7 @@ public:
 
   void addExpr(MCInst &Inst, const MCExpr *Expr) const {
     // Add as immediates when possible.  Null MCExpr = 0.
-    if (Expr == 0)
+    if (!Expr)
       Inst.addOperand(MCOperand::CreateImm(0));
     else if (const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(Expr))
       Inst.addOperand(MCOperand::CreateImm(CE->getValue()));
@@ -1926,6 +2020,50 @@ public:
     Inst.addOperand(MCOperand::CreateImm(Memory.Alignment));
   }
 
+  void addDupAlignedMemoryNoneOperands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemoryNoneOperands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemory16Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addDupAlignedMemory16Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemory32Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addDupAlignedMemory32Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemory64Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addDupAlignedMemory64Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemory64or128Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addDupAlignedMemory64or128Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
+  void addAlignedMemory64or128or256Operands(MCInst &Inst, unsigned N) const {
+    addAlignedMemoryOperands(Inst, N);
+  }
+
   void addAddrMode2Operands(MCInst &Inst, unsigned N) const {
     assert(N == 3 && "Invalid number of operands!");
     int32_t Val = Memory.OffsetImm ? Memory.OffsetImm->getValue() : 0;
@@ -2275,6 +2413,19 @@ public:
     Inst.addOperand(MCOperand::CreateImm(Value));
   }
 
+  void addNEONinvByteReplicateOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    // The immediate encodes the type of constant as well as the value.
+    const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
+    unsigned Value = CE->getValue();
+    assert((Inst.getOpcode() == ARM::VMOVv8i8 ||
+            Inst.getOpcode() == ARM::VMOVv16i8) &&
+           "All vmvn instructions that wants to replicate non-zero byte "
+           "always must be replaced with VMOVv8i8 or VMOVv16i8.");
+    unsigned B = ((~Value) & 0xff);
+    B |= 0xe00; // cmode = 0b1110
+    Inst.addOperand(MCOperand::CreateImm(B));
+  }
   void addNEONi32vmovOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     // The immediate encodes the type of constant as well as the value.
@@ -2289,6 +2440,19 @@ public:
     Inst.addOperand(MCOperand::CreateImm(Value));
   }
 
+  void addNEONvmovByteReplicateOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    // The immediate encodes the type of constant as well as the value.
+    const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
+    unsigned Value = CE->getValue();
+    assert((Inst.getOpcode() == ARM::VMOVv8i8 ||
+            Inst.getOpcode() == ARM::VMOVv16i8) &&
+           "All instructions that wants to replicate non-zero byte "
+           "always must be replaced with VMOVv8i8 or VMOVv16i8.");
+    unsigned B = Value & 0xff;
+    B |= 0xe00; // cmode = 0b1110
+    Inst.addOperand(MCOperand::CreateImm(B));
+  }
   void addNEONi32vmovNegOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     // The immediate encodes the type of constant as well as the value.
@@ -2523,7 +2687,8 @@ public:
                                unsigned ShiftImm,
                                unsigned Alignment,
                                bool isNegative,
-                               SMLoc S, SMLoc E) {
+                               SMLoc S, SMLoc E,
+                               SMLoc AlignmentLoc = SMLoc()) {
     ARMOperand *Op = new ARMOperand(k_Memory);
     Op->Memory.BaseRegNum = BaseRegNum;
     Op->Memory.OffsetImm = OffsetImm;
@@ -2534,6 +2699,7 @@ public:
     Op->Memory.isNegative = isNegative;
     Op->StartLoc = S;
     Op->EndLoc = E;
+    Op->AlignmentLoc = AlignmentLoc;
     return Op;
   }
 
@@ -2806,7 +2972,7 @@ int ARMAsmParser::tryParseShiftRegister(
   // The source register for the shift has already been added to the
   // operand list, so we need to pop it off and combine it into the shifted
   // register operand instead.
-  OwningPtr<ARMOperand> PrevOp((ARMOperand*)Operands.pop_back_val());
+  std::unique_ptr<ARMOperand> PrevOp((ARMOperand*)Operands.pop_back_val());
   if (!PrevOp->isReg())
     return Error(PrevOp->getStartLoc(), "shift must be of a register");
   int SrcReg = PrevOp->getReg();
@@ -2825,7 +2991,7 @@ int ARMAsmParser::tryParseShiftRegister(
         Parser.getTok().is(AsmToken::Dollar)) {
       Parser.Lex(); // Eat hash.
       SMLoc ImmLoc = Parser.getTok().getLoc();
-      const MCExpr *ShiftExpr = 0;
+      const MCExpr *ShiftExpr = nullptr;
       if (getParser().parseExpression(ShiftExpr, EndLoc)) {
         Error(ImmLoc, "invalid immediate shift value");
         return -1;
@@ -2855,12 +3021,12 @@ int ARMAsmParser::tryParseShiftRegister(
       EndLoc = Parser.getTok().getEndLoc();
       ShiftReg = tryParseRegister();
       if (ShiftReg == -1) {
-        Error (L, "expected immediate or register in shift operand");
+        Error(L, "expected immediate or register in shift operand");
         return -1;
       }
     } else {
-      Error (Parser.getTok().getLoc(),
-                    "expected immediate or register in shift operand");
+      Error(Parser.getTok().getLoc(),
+            "expected immediate or register in shift operand");
       return -1;
     }
   }
@@ -4323,8 +4489,9 @@ parseMemory(SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
     E = Tok.getEndLoc();
     Parser.Lex(); // Eat right bracket token.
 
-    Operands.push_back(ARMOperand::CreateMem(BaseRegNum, 0, 0, ARM_AM::no_shift,
-                                             0, 0, false, S, E));
+    Operands.push_back(ARMOperand::CreateMem(BaseRegNum, nullptr, 0,
+                                             ARM_AM::no_shift, 0, 0, false,
+                                             S, E));
 
     // If there's a pre-indexing writeback marker, '!', just add it as a token
     // operand. It's rather odd, but syntactically valid.
@@ -4346,6 +4513,7 @@ parseMemory(SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
   if (Parser.getTok().is(AsmToken::Colon)) {
     Parser.Lex(); // Eat the ':'.
     E = Parser.getTok().getLoc();
+    SMLoc AlignmentLoc = Tok.getLoc();
 
     const MCExpr *Expr;
     if (getParser().parseExpression(Expr))
@@ -4378,9 +4546,9 @@ parseMemory(SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
 
     // Don't worry about range checking the value here. That's handled by
     // the is*() predicates.
-    Operands.push_back(ARMOperand::CreateMem(BaseRegNum, 0, 0,
+    Operands.push_back(ARMOperand::CreateMem(BaseRegNum, nullptr, 0,
                                              ARM_AM::no_shift, 0, Align,
-                                             false, S, E));
+                                             false, S, E, AlignmentLoc));
 
     // If there's a pre-indexing writeback marker, '!', just add it as a token
     // operand.
@@ -4471,7 +4639,7 @@ parseMemory(SmallVectorImpl<MCParsedAsmOperand*> &Operands) {
   E = Parser.getTok().getEndLoc();
   Parser.Lex(); // Eat right bracket token.
 
-  Operands.push_back(ARMOperand::CreateMem(BaseRegNum, 0, OffsetRegNum,
+  Operands.push_back(ARMOperand::CreateMem(BaseRegNum, nullptr, OffsetRegNum,
                                            ShiftType, ShiftImm, 0, isNegative,
                                            S, E));
 
@@ -4926,8 +5094,9 @@ getMnemonicAcceptInfo(StringRef Mnemonic, StringRef FullInst,
 
   if (Mnemonic == "bkpt" || Mnemonic == "cbnz" || Mnemonic == "setend" ||
       Mnemonic == "cps" ||  Mnemonic == "it" ||  Mnemonic == "cbz" ||
-      Mnemonic == "trap" || Mnemonic == "hlt" || Mnemonic.startswith("crc32") ||
-      Mnemonic.startswith("cps") || Mnemonic.startswith("vsel") ||
+      Mnemonic == "trap" || Mnemonic == "hlt" || Mnemonic == "udf" ||
+      Mnemonic.startswith("crc32") || Mnemonic.startswith("cps") ||
+      Mnemonic.startswith("vsel") ||
       Mnemonic == "vmaxnm" || Mnemonic == "vminnm" || Mnemonic == "vcvta" ||
       Mnemonic == "vcvtn" || Mnemonic == "vcvtp" || Mnemonic == "vcvtm" ||
       Mnemonic == "vrinta" || Mnemonic == "vrintn" || Mnemonic == "vrintp" ||
@@ -5404,21 +5573,24 @@ bool ARMAsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
   }
 
   // GNU Assembler extension (compatibility)
-  if ((Mnemonic == "ldrd" || Mnemonic == "strd") && !isThumb() &&
-      Operands.size() == 4) {
-    ARMOperand *Op = static_cast<ARMOperand *>(Operands[2]);
-    assert(Op->isReg() && "expected register argument");
+  if ((Mnemonic == "ldrd" || Mnemonic == "strd")) {
+    ARMOperand *Op2 = static_cast<ARMOperand *>(Operands[2]);
+    ARMOperand *Op3 = static_cast<ARMOperand *>(Operands[3]);
+    if (Op3->isMem()) {
+      assert(Op2->isReg() && "expected register argument");
 
-    unsigned SuperReg = MRI->getMatchingSuperReg(
-        Op->getReg(), ARM::gsub_0, &MRI->getRegClass(ARM::GPRPairRegClassID));
+      unsigned SuperReg = MRI->getMatchingSuperReg(
+          Op2->getReg(), ARM::gsub_0, &MRI->getRegClass(ARM::GPRPairRegClassID));
 
-    assert(SuperReg && "expected register pair");
+      assert(SuperReg && "expected register pair");
 
-    unsigned PairedReg = MRI->getSubReg(SuperReg, ARM::gsub_1);
+      unsigned PairedReg = MRI->getSubReg(SuperReg, ARM::gsub_1);
 
-    Operands.insert(Operands.begin() + 3,
-                    ARMOperand::CreateReg(PairedReg, Op->getStartLoc(),
-                                          Op->getEndLoc()));
+      Operands.insert(Operands.begin() + 3,
+                      ARMOperand::CreateReg(PairedReg,
+                                            Op2->getStartLoc(),
+                                            Op2->getEndLoc()));
+    }
   }
 
   // FIXME: As said above, this is all a pretty gross hack.  This instruction
@@ -5748,6 +5920,30 @@ validateInstruction(MCInst &Inst,
       return Error(Operands[Op]->getStartLoc(), "branch target out of range");
     break;
   }
+  case ARM::MOVi16:
+  case ARM::t2MOVi16:
+  case ARM::t2MOVTi16:
+    {
+    // We want to avoid misleadingly allowing something like "mov r0, <symbol>"
+    // especially when we turn it into a movw and the expression <symbol> does
+    // not have a :lower16: or :upper16 as part of the expression.  We don't
+    // want the behavior of silently truncating, which can be unexpected and
+    // lead to bugs that are difficult to find since this is an easy mistake
+    // to make.
+    int i = (Operands[3]->isImm()) ? 3 : 4;
+    ARMOperand *Op = static_cast<ARMOperand*>(Operands[i]);
+    const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(Op->getImm());
+    if (CE) break;
+    const MCExpr *E = dyn_cast<MCExpr>(Op->getImm());
+    if (!E) break;
+    const ARMMCExpr *ARM16Expr = dyn_cast<ARMMCExpr>(E);
+    if (!ARM16Expr || (ARM16Expr->getKind() != ARMMCExpr::VK_ARM_HI16 &&
+                       ARM16Expr->getKind() != ARMMCExpr::VK_ARM_LO16)) {
+      return Error(Op->getStartLoc(),
+	     "immediate expression for mov requires :lower16: or :upper16");
+      break;
+    }
+    }
   }
 
   return false;
@@ -5898,7 +6094,7 @@ static unsigned getRealVLDOpcode(unsigned Opc, unsigned &Spacing) {
   case ARM::VLD3DUPdWB_fixed_Asm_16: Spacing = 1; return ARM::VLD3DUPd16_UPD;
   case ARM::VLD3DUPdWB_fixed_Asm_32: Spacing = 1; return ARM::VLD3DUPd32_UPD;
   case ARM::VLD3DUPqWB_fixed_Asm_8: Spacing = 1; return ARM::VLD3DUPq8_UPD;
-  case ARM::VLD3DUPqWB_fixed_Asm_16: Spacing = 1; return ARM::VLD3DUPq16_UPD;
+  case ARM::VLD3DUPqWB_fixed_Asm_16: Spacing = 2; return ARM::VLD3DUPq16_UPD;
   case ARM::VLD3DUPqWB_fixed_Asm_32: Spacing = 2; return ARM::VLD3DUPq32_UPD;
   case ARM::VLD3DUPdWB_register_Asm_8:  Spacing = 1; return ARM::VLD3DUPd8_UPD;
   case ARM::VLD3DUPdWB_register_Asm_16: Spacing = 1; return ARM::VLD3DUPd16_UPD;
@@ -7860,9 +8056,11 @@ unsigned ARMAsmParser::checkTargetMatchPredicate(MCInst &Inst) {
   return Match_Success;
 }
 
-template<> inline bool IsCPSRDead<MCInst>(MCInst* Instr) {
+namespace llvm {
+template <> inline bool IsCPSRDead<MCInst>(MCInst *Instr) {
   return true; // In an assembly source, no need to second-guess
 }
+}
 
 static const char *getSubtargetFeatureName(unsigned Val);
 bool ARMAsmParser::
@@ -7965,6 +8163,42 @@ MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
     if (ErrorLoc == SMLoc()) ErrorLoc = IDLoc;
     return Error(ErrorLoc, "immediate operand must be in the range [0,239]");
   }
+  case Match_AlignedMemoryRequiresNone:
+  case Match_DupAlignedMemoryRequiresNone:
+  case Match_AlignedMemoryRequires16:
+  case Match_DupAlignedMemoryRequires16:
+  case Match_AlignedMemoryRequires32:
+  case Match_DupAlignedMemoryRequires32:
+  case Match_AlignedMemoryRequires64:
+  case Match_DupAlignedMemoryRequires64:
+  case Match_AlignedMemoryRequires64or128:
+  case Match_DupAlignedMemoryRequires64or128:
+  case Match_AlignedMemoryRequires64or128or256:
+  {
+    SMLoc ErrorLoc = ((ARMOperand*)Operands[ErrorInfo])->getAlignmentLoc();
+    if (ErrorLoc == SMLoc()) ErrorLoc = IDLoc;
+    switch (MatchResult) {
+      default:
+        llvm_unreachable("Missing Match_Aligned type");
+      case Match_AlignedMemoryRequiresNone:
+      case Match_DupAlignedMemoryRequiresNone:
+        return Error(ErrorLoc, "alignment must be omitted");
+      case Match_AlignedMemoryRequires16:
+      case Match_DupAlignedMemoryRequires16:
+        return Error(ErrorLoc, "alignment must be 16 or omitted");
+      case Match_AlignedMemoryRequires32:
+      case Match_DupAlignedMemoryRequires32:
+        return Error(ErrorLoc, "alignment must be 32 or omitted");
+      case Match_AlignedMemoryRequires64:
+      case Match_DupAlignedMemoryRequires64:
+        return Error(ErrorLoc, "alignment must be 64 or omitted");
+      case Match_AlignedMemoryRequires64or128:
+      case Match_DupAlignedMemoryRequires64or128:
+        return Error(ErrorLoc, "alignment must be 64, 128 or omitted");
+      case Match_AlignedMemoryRequires64or128or256:
+        return Error(ErrorLoc, "alignment must be 64, 128, 256 or omitted");
+    }
+  }
   }
 
   llvm_unreachable("Implement any new match types added!");
@@ -7972,6 +8206,10 @@ MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
 
 /// parseDirective parses the arm specific directives
 bool ARMAsmParser::ParseDirective(AsmToken DirectiveID) {
+  const MCObjectFileInfo::Environment Format =
+    getContext().getObjectFileInfo()->getObjectFileType();
+  bool IsMachO = Format == MCObjectFileInfo::IsMachO;
+
   StringRef IDVal = DirectiveID.getIdentifier();
   if (IDVal == ".word")
     return parseLiteralValues(4, DirectiveID.getLoc());
@@ -7989,16 +8227,6 @@ bool ARMAsmParser::ParseDirective(AsmToken DirectiveID) {
     return parseDirectiveSyntax(DirectiveID.getLoc());
   else if (IDVal == ".unreq")
     return parseDirectiveUnreq(DirectiveID.getLoc());
-  else if (IDVal == ".arch")
-    return parseDirectiveArch(DirectiveID.getLoc());
-  else if (IDVal == ".eabi_attribute")
-    return parseDirectiveEabiAttr(DirectiveID.getLoc());
-  else if (IDVal == ".cpu")
-    return parseDirectiveCPU(DirectiveID.getLoc());
-  else if (IDVal == ".fpu")
-    return parseDirectiveFPU(DirectiveID.getLoc());
-  else if (IDVal == ".fnstart")
-    return parseDirectiveFnStart(DirectiveID.getLoc());
   else if (IDVal == ".fnend")
     return parseDirectiveFnEnd(DirectiveID.getLoc());
   else if (IDVal == ".cantunwind")
@@ -8015,12 +8243,6 @@ bool ARMAsmParser::ParseDirective(AsmToken DirectiveID) {
     return parseDirectiveRegSave(DirectiveID.getLoc(), false);
   else if (IDVal == ".vsave")
     return parseDirectiveRegSave(DirectiveID.getLoc(), true);
-  else if (IDVal == ".inst")
-    return parseDirectiveInst(DirectiveID.getLoc());
-  else if (IDVal == ".inst.n")
-    return parseDirectiveInst(DirectiveID.getLoc(), 'n');
-  else if (IDVal == ".inst.w")
-    return parseDirectiveInst(DirectiveID.getLoc(), 'w');
   else if (IDVal == ".ltorg" || IDVal == ".pool")
     return parseDirectiveLtorg(DirectiveID.getLoc());
   else if (IDVal == ".even")
@@ -8029,18 +8251,38 @@ bool ARMAsmParser::ParseDirective(AsmToken DirectiveID) {
     return parseDirectivePersonalityIndex(DirectiveID.getLoc());
   else if (IDVal == ".unwind_raw")
     return parseDirectiveUnwindRaw(DirectiveID.getLoc());
-  else if (IDVal == ".tlsdescseq")
-    return parseDirectiveTLSDescSeq(DirectiveID.getLoc());
   else if (IDVal == ".movsp")
     return parseDirectiveMovSP(DirectiveID.getLoc());
-  else if (IDVal == ".object_arch")
-    return parseDirectiveObjectArch(DirectiveID.getLoc());
   else if (IDVal == ".arch_extension")
     return parseDirectiveArchExtension(DirectiveID.getLoc());
   else if (IDVal == ".align")
     return parseDirectiveAlign(DirectiveID.getLoc());
   else if (IDVal == ".thumb_set")
     return parseDirectiveThumbSet(DirectiveID.getLoc());
+
+  if (!IsMachO) {
+    if (IDVal == ".arch")
+      return parseDirectiveArch(DirectiveID.getLoc());
+    else if (IDVal == ".cpu")
+      return parseDirectiveCPU(DirectiveID.getLoc());
+    else if (IDVal == ".eabi_attribute")
+      return parseDirectiveEabiAttr(DirectiveID.getLoc());
+    else if (IDVal == ".fpu")
+      return parseDirectiveFPU(DirectiveID.getLoc());
+    else if (IDVal == ".fnstart")
+      return parseDirectiveFnStart(DirectiveID.getLoc());
+    else if (IDVal == ".inst")
+      return parseDirectiveInst(DirectiveID.getLoc());
+    else if (IDVal == ".inst.n")
+      return parseDirectiveInst(DirectiveID.getLoc(), 'n');
+    else if (IDVal == ".inst.w")
+      return parseDirectiveInst(DirectiveID.getLoc(), 'w');
+    else if (IDVal == ".object_arch")
+      return parseDirectiveObjectArch(DirectiveID.getLoc());
+    else if (IDVal == ".tlsdescseq")
+      return parseDirectiveTLSDescSeq(DirectiveID.getLoc());
+  }
+
   return true;
 }
 
@@ -8121,32 +8363,6 @@ void ARMAsmParser::onLabelParsed(MCSymbol *Symbol) {
   if (NextSymbolIsThumb) {
     getParser().getStreamer().EmitThumbFunc(Symbol);
     NextSymbolIsThumb = false;
-    return;
-  }
-
-  if (!isThumb())
-    return;
-
-  const MCObjectFileInfo::Environment Format =
-    getContext().getObjectFileInfo()->getObjectFileType();
-  switch (Format) {
-  case MCObjectFileInfo::IsCOFF: {
-    const MCSymbolData &SD =
-      getParser().getStreamer().getOrCreateSymbolData(Symbol);
-    char Type = COFF::IMAGE_SYM_DTYPE_FUNCTION << COFF::SCT_COMPLEX_TYPE_SHIFT;
-    if (SD.getFlags() & (Type << COFF::SF_TypeShift))
-      getParser().getStreamer().EmitThumbFunc(Symbol);
-    break;
-  }
-  case MCObjectFileInfo::IsELF: {
-    const MCSymbolData &SD =
-      getParser().getStreamer().getOrCreateSymbolData(Symbol);
-    if (MCELF::GetType(SD) & (ELF::STT_FUNC << ELF_STT_Shift))
-      getParser().getStreamer().EmitThumbFunc(Symbol);
-    break;
-  }
-  case MCObjectFileInfo::IsMachO:
-    break;
   }
 }
 
@@ -8303,14 +8519,6 @@ bool ARMAsmParser::parseDirectiveUnreq(SMLoc L) {
 /// parseDirectiveArch
 ///  ::= .arch token
 bool ARMAsmParser::parseDirectiveArch(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".arch directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   StringRef Arch = getParser().parseStringToEndOfStatement().trim();
 
   unsigned ID = StringSwitch<unsigned>(Arch)
@@ -8334,14 +8542,6 @@ bool ARMAsmParser::parseDirectiveArch(SMLoc L) {
 ///  ::= .eabi_attribute int, int [, "str"]
 ///  ::= .eabi_attribute Tag_name, int [, "str"]
 bool ARMAsmParser::parseDirectiveEabiAttr(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".eabi_attribute directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   int64_t Tag;
   SMLoc TagLoc;
   TagLoc = Parser.getTok().getLoc();
@@ -8447,14 +8647,6 @@ bool ARMAsmParser::parseDirectiveEabiAttr(SMLoc L) {
 /// parseDirectiveCPU
 ///  ::= .cpu str
 bool ARMAsmParser::parseDirectiveCPU(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".cpu directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   StringRef CPU = getParser().parseStringToEndOfStatement().trim();
   getTargetStreamer().emitTextAttribute(ARMBuildAttrs::CPU_name, CPU);
   return false;
@@ -8463,14 +8655,6 @@ bool ARMAsmParser::parseDirectiveCPU(SMLoc L) {
 /// parseDirectiveFPU
 ///  ::= .fpu str
 bool ARMAsmParser::parseDirectiveFPU(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".fpu directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   StringRef FPU = getParser().parseStringToEndOfStatement().trim();
 
   unsigned ID = StringSwitch<unsigned>(FPU)
@@ -8490,14 +8674,6 @@ bool ARMAsmParser::parseDirectiveFPU(SMLoc L) {
 /// parseDirectiveFnStart
 ///  ::= .fnstart
 bool ARMAsmParser::parseDirectiveFnStart(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".fnstart directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   if (UC.hasFnStart()) {
     Error(L, ".fnstart starts before the end of previous one");
     UC.emitFnStartLocNotes();
@@ -8777,14 +8953,6 @@ bool ARMAsmParser::parseDirectiveRegSave(SMLoc L, bool IsVector) {
 ///  ::= .inst.n opcode [, ...]
 ///  ::= .inst.w opcode [, ...]
 bool ARMAsmParser::parseDirectiveInst(SMLoc Loc, char Suffix) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(Loc, ".inst directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   int Width;
 
   if (isThumb()) {
@@ -9033,14 +9201,6 @@ bool ARMAsmParser::parseDirectiveUnwindRaw(SMLoc L) {
 /// parseDirectiveTLSDescSeq
 ///   ::= .tlsdescseq tls-variable
 bool ARMAsmParser::parseDirectiveTLSDescSeq(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".tlsdescseq directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   if (getLexer().isNot(AsmToken::Identifier)) {
     TokError("expected variable after '.tlsdescseq' directive");
     Parser.eatToEndOfStatement();
@@ -9128,14 +9288,6 @@ bool ARMAsmParser::parseDirectiveMovSP(SMLoc L) {
 /// parseDirectiveObjectArch
 ///   ::= .object_arch name
 bool ARMAsmParser::parseDirectiveObjectArch(SMLoc L) {
-  const MCAsmInfo *MAI = getParser().getStreamer().getContext().getAsmInfo();
-  bool isMachO = MAI->hasSubsectionsViaSymbols();
-  if (isMachO) {
-    Error(L, ".object_arch directive not valid for Mach-O");
-    Parser.eatToEndOfStatement();
-    return false;
-  }
-
   if (getLexer().isNot(AsmToken::Identifier)) {
     Error(getLexer().getLoc(), "unexpected token");
     Parser.eatToEndOfStatement();
@@ -9221,36 +9373,7 @@ bool ARMAsmParser::parseDirectiveThumbSet(SMLoc L) {
   Lex();
 
   MCSymbol *Alias = getContext().GetOrCreateSymbol(Name);
-  if (const MCSymbolRefExpr *SRE = dyn_cast<MCSymbolRefExpr>(Value)) {
-    MCSymbol *Sym = getContext().LookupSymbol(SRE->getSymbol().getName());
-    if (!Sym->isDefined()) {
-      getStreamer().EmitSymbolAttribute(Sym, MCSA_Global);
-      getStreamer().EmitAssignment(Alias, Value);
-      return false;
-    }
-
-    const MCObjectFileInfo::Environment Format =
-      getContext().getObjectFileInfo()->getObjectFileType();
-    switch (Format) {
-    case MCObjectFileInfo::IsCOFF: {
-      char Type = COFF::IMAGE_SYM_DTYPE_FUNCTION << COFF::SCT_COMPLEX_TYPE_SHIFT;
-      getStreamer().EmitCOFFSymbolType(Type);
-      // .set values are always local in COFF
-      getStreamer().EmitSymbolAttribute(Alias, MCSA_Local);
-      break;
-    }
-    case MCObjectFileInfo::IsELF:
-      getStreamer().EmitSymbolAttribute(Alias, MCSA_ELF_TypeFunction);
-      break;
-    case MCObjectFileInfo::IsMachO:
-      break;
-    }
-  }
-
-  // FIXME: set the function as being a thumb function via the assembler
-  getStreamer().EmitThumbFunc(Alias);
-  getStreamer().EmitAssignment(Alias, Value);
-
+  getTargetStreamer().emitThumbSet(Alias, Value);
   return false;
 }
 
@@ -9365,8 +9488,8 @@ unsigned ARMAsmParser::validateTargetOperandClass(MCParsedAsmOperand *AsmOp,
       int64_t Value;
       if (!SOExpr->EvaluateAsAbsolute(Value))
         return Match_Success;
-      assert((Value >= INT32_MIN && Value <= INT32_MAX) &&
-             "expression value must be representiable in 32 bits");
+      assert((Value >= INT32_MIN && Value <= UINT32_MAX) &&
+             "expression value must be representable in 32 bits");
     }
     break;
   case MCK_GPRPair:
diff --git a/lib/Target/ARM/CMakeLists.txt b/lib/Target/ARM/CMakeLists.txt
index 8e14883..9b5fa75 100644
--- a/lib/Target/ARM/CMakeLists.txt
+++ b/lib/Target/ARM/CMakeLists.txt
@@ -17,7 +17,6 @@ add_public_tablegen_target(ARMCommonTableGen)
 add_llvm_target(ARMCodeGen
   A15SDOptimizer.cpp
   ARMAsmPrinter.cpp
-  ARMAtomicExpandPass.cpp
   ARMBaseInstrInfo.cpp
   ARMBaseRegisterInfo.cpp
   ARMCodeEmitter.cpp
diff --git a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
index 9e40381..4d4038d 100644
--- a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
+++ b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
@@ -7,8 +7,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "arm-disassembler"
-
 #include "llvm/MC/MCDisassembler.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
 #include "MCTargetDesc/ARMBaseInfo.h"
@@ -29,6 +27,8 @@
 
 using namespace llvm;
 
+#define DEBUG_TYPE "arm-disassembler"
+
 typedef MCDisassembler::DecodeStatus DecodeStatus;
 
 namespace {
@@ -90,8 +90,8 @@ class ARMDisassembler : public MCDisassembler {
 public:
   /// Constructor     - Initializes the disassembler.
   ///
-  ARMDisassembler(const MCSubtargetInfo &STI) :
-    MCDisassembler(STI) {
+  ARMDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx) :
+    MCDisassembler(STI, Ctx) {
   }
 
   ~ARMDisassembler() {
@@ -109,8 +109,8 @@ class ThumbDisassembler : public MCDisassembler {
 public:
   /// Constructor     - Initializes the disassembler.
   ///
-  ThumbDisassembler(const MCSubtargetInfo &STI) :
-    MCDisassembler(STI) {
+  ThumbDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx) :
+    MCDisassembler(STI, Ctx) {
   }
 
   ~ThumbDisassembler() {
@@ -400,12 +400,16 @@ static DecodeStatus DecodeMRRC2(llvm::MCInst &Inst, unsigned Val,
                                 uint64_t Address, const void *Decoder);
 #include "ARMGenDisassemblerTables.inc"
 
-static MCDisassembler *createARMDisassembler(const Target &T, const MCSubtargetInfo &STI) {
-  return new ARMDisassembler(STI);
+static MCDisassembler *createARMDisassembler(const Target &T,
+                                             const MCSubtargetInfo &STI,
+                                             MCContext &Ctx) {
+  return new ARMDisassembler(STI, Ctx);
 }
 
-static MCDisassembler *createThumbDisassembler(const Target &T, const MCSubtargetInfo &STI) {
-  return new ThumbDisassembler(STI);
+static MCDisassembler *createThumbDisassembler(const Target &T,
+                                               const MCSubtargetInfo &STI,
+                                               MCContext &Ctx) {
+  return new ThumbDisassembler(STI, Ctx);
 }
 
 DecodeStatus ARMDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
diff --git a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
index da3fe01..e4b785d 100644
--- a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
+++ b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
@@ -11,7 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "asm-printer"
 #include "ARMInstPrinter.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
 #include "MCTargetDesc/ARMBaseInfo.h"
@@ -23,6 +22,8 @@
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "asm-printer"
+
 #include "ARMGenAsmWriter.inc"
 
 /// translateShiftImm - Convert shift immediate from 0-31 to 1-32 for printing.
diff --git a/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp b/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
index 1db517f..7acd9cc 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
@@ -306,8 +306,36 @@ bool ARMAsmBackend::writeNopData(uint64_t Count, MCObjectWriter *OW) const {
   return true;
 }
 
+static uint32_t swapHalfWords(uint32_t Value, bool IsLittleEndian) {
+  if (IsLittleEndian) {
+    // Note that the halfwords are stored high first and low second in thumb;
+    // so we need to swap the fixup value here to map properly.
+    uint32_t Swapped = (Value & 0xFFFF0000) >> 16;
+    Swapped |= (Value & 0x0000FFFF) << 16;
+    return Swapped;
+  }
+  else
+    return Value;
+}
+
+static uint32_t joinHalfWords(uint32_t FirstHalf, uint32_t SecondHalf,
+                              bool IsLittleEndian) {
+  uint32_t Value;
+
+  if (IsLittleEndian) {
+    Value = (SecondHalf & 0xFFFF) << 16;
+    Value |= (FirstHalf & 0xFFFF);
+  } else {
+    Value = (SecondHalf & 0xFFFF);
+    Value |= (FirstHalf & 0xFFFF) << 16;
+  }
+
+  return Value;
+}
+
 static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
-                                 bool IsPCRel, MCContext *Ctx) {
+                                 bool IsPCRel, MCContext *Ctx,
+                                 bool IsLittleEndian) {
   unsigned Kind = Fixup.getKind();
   switch (Kind) {
   default:
@@ -316,6 +344,10 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
   case FK_Data_2:
   case FK_Data_4:
     return Value;
+  case FK_SecRel_2:
+    return Value;
+  case FK_SecRel_4:
+    return Value;
   case ARM::fixup_arm_movt_hi16:
     if (!IsPCRel)
       Value >>= 16;
@@ -342,9 +374,7 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     // inst{14-12} = Mid3;
     // inst{7-0} = Lo8;
     Value = (Hi4 << 16) | (i << 26) | (Mid3 << 12) | (Lo8);
-    uint64_t swapped = (Value & 0xFFFF0000) >> 16;
-    swapped |= (Value & 0x0000FFFF) << 16;
-    return swapped;
+    return swapHalfWords(Value, IsLittleEndian);
   }
   case ARM::fixup_arm_ldst_pcrel_12:
     // ARM PC-relative values are offset by 8.
@@ -364,11 +394,8 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
 
     // Same addressing mode as fixup_arm_pcrel_10,
     // but with 16-bit halfwords swapped.
-    if (Kind == ARM::fixup_t2_ldst_pcrel_12) {
-      uint64_t swapped = (Value & 0xFFFF0000) >> 16;
-      swapped |= (Value & 0x0000FFFF) << 16;
-      return swapped;
-    }
+    if (Kind == ARM::fixup_t2_ldst_pcrel_12)
+      return swapHalfWords(Value, IsLittleEndian);
 
     return Value;
   }
@@ -401,9 +428,7 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     out |= (Value & 0x700) << 4;
     out |= (Value & 0x0FF);
 
-    uint64_t swapped = (out & 0xFFFF0000) >> 16;
-    swapped |= (out & 0x0000FFFF) << 16;
-    return swapped;
+    return swapHalfWords(out, IsLittleEndian);
   }
 
   case ARM::fixup_arm_condbranch:
@@ -434,9 +459,7 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     out |= (Value & 0x1FF800)  << 5; // imm6 field
     out |= (Value & 0x0007FF);        // imm11 field
 
-    uint64_t swapped = (out & 0xFFFF0000) >> 16;
-    swapped |= (out & 0x0000FFFF) << 16;
-    return swapped;
+    return swapHalfWords(out, IsLittleEndian);
   }
   case ARM::fixup_t2_condbranch: {
     Value = Value - 4;
@@ -449,9 +472,7 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     out |= (Value & 0x1F800) << 5; // imm6 field
     out |= (Value & 0x007FF);      // imm11 field
 
-    uint32_t swapped = (out & 0xFFFF0000) >> 16;
-    swapped |= (out & 0x0000FFFF) << 16;
-    return swapped;
+    return swapHalfWords(out, IsLittleEndian);
   }
   case ARM::fixup_arm_thumb_bl: {
     // The value doesn't encode the low bit (always zero) and is offset by
@@ -475,13 +496,10 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     uint32_t imm10Bits = (offset & 0x1FF800) >> 11;
     uint32_t imm11Bits = (offset & 0x000007FF);
 
-    uint32_t Binary = 0;
-    uint32_t firstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10Bits);
-    uint32_t secondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
+    uint32_t FirstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10Bits);
+    uint32_t SecondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
                           (uint16_t)imm11Bits);
-    Binary |= secondHalf << 16;
-    Binary |= firstHalf;
-    return Binary;
+    return joinHalfWords(FirstHalf, SecondHalf, IsLittleEndian);
   }
   case ARM::fixup_arm_thumb_blx: {
     // The value doesn't encode the low two bits (always zero) and is offset by
@@ -508,13 +526,10 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
     uint32_t imm10HBits = (offset & 0xFFC00) >> 10;
     uint32_t imm10LBits = (offset & 0x3FF);
 
-    uint32_t Binary = 0;
-    uint32_t firstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10HBits);
-    uint32_t secondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
+    uint32_t FirstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10HBits);
+    uint32_t SecondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
                           ((uint16_t)imm10LBits) << 1);
-    Binary |= secondHalf << 16;
-    Binary |= firstHalf;
-    return Binary;
+    return joinHalfWords(FirstHalf, SecondHalf, IsLittleEndian);
   }
   case ARM::fixup_arm_thumb_cp:
     // Offset by 4, and don't encode the low two bits. Two bytes of that
@@ -566,11 +581,8 @@ static unsigned adjustFixupValue(const MCFixup &Fixup, uint64_t Value,
 
     // Same addressing mode as fixup_arm_pcrel_10, but with 16-bit halfwords
     // swapped.
-    if (Kind == ARM::fixup_t2_pcrel_10) {
-      uint32_t swapped = (Value & 0xFFFF0000) >> 16;
-      swapped |= (Value & 0x0000FFFF) << 16;
-      return swapped;
-    }
+    if (Kind == ARM::fixup_t2_pcrel_10)
+      return swapHalfWords(Value, IsLittleEndian);
 
     return Value;
   }
@@ -603,7 +615,7 @@ void ARMAsmBackend::processFixupValue(const MCAssembler &Asm,
   // the offset when the destination has the same MCFragment.
   if (A && (unsigned)Fixup.getKind() == ARM::fixup_arm_thumb_bl) {
     const MCSymbol &Sym = A->getSymbol().AliasedSymbol();
-    MCSymbolData &SymData = Asm.getSymbolData(Sym);
+    const MCSymbolData &SymData = Asm.getSymbolData(Sym);
     IsResolved = (SymData.getFragment() == DF);
   }
   // We must always generate a relocation for BL/BLX instructions if we have
@@ -618,7 +630,8 @@ void ARMAsmBackend::processFixupValue(const MCAssembler &Asm,
   // Try to get the encoded value for the fixup as-if we're mapping it into
   // the instruction. This allows adjustFixupValue() to issue a diagnostic
   // if the value aren't invalid.
-  (void)adjustFixupValue(Fixup, Value, false, &Asm.getContext());
+  (void)adjustFixupValue(Fixup, Value, false, &Asm.getContext(),
+                         IsLittleEndian);
 }
 
 /// getFixupKindNumBytes - The number of bytes the fixup may change.
@@ -662,6 +675,11 @@ static unsigned getFixupKindNumBytes(unsigned Kind) {
   case ARM::fixup_t2_movt_hi16:
   case ARM::fixup_t2_movw_lo16:
     return 4;
+
+  case FK_SecRel_2:
+    return 2;
+  case FK_SecRel_4:
+    return 4;
   }
 }
 
@@ -716,7 +734,7 @@ void ARMAsmBackend::applyFixup(const MCFixup &Fixup, char *Data,
                                unsigned DataSize, uint64_t Value,
                                bool IsPCRel) const {
   unsigned NumBytes = getFixupKindNumBytes(Fixup.getKind());
-  Value = adjustFixupValue(Fixup, Value, IsPCRel, nullptr);
+  Value = adjustFixupValue(Fixup, Value, IsPCRel, nullptr, IsLittleEndian);
   if (!Value) return;           // Doesn't change encoding.
 
   unsigned Offset = Fixup.getOffset();
@@ -724,8 +742,11 @@ void ARMAsmBackend::applyFixup(const MCFixup &Fixup, char *Data,
 
   // Used to point to big endian bytes.
   unsigned FullSizeBytes;
-  if (!IsLittleEndian)
+  if (!IsLittleEndian) {
     FullSizeBytes = getFixupKindContainerSizeBytes(Fixup.getKind());
+    assert((Offset + FullSizeBytes) <= DataSize && "Invalid fixup size!");
+    assert(NumBytes <= FullSizeBytes && "Invalid fixup size!");
+  }
 
   // For each byte of the fragment that the fixup touches, mask in the bits from
   // the fixup value. The Value has been "split up" into the appropriate
@@ -737,6 +758,15 @@ void ARMAsmBackend::applyFixup(const MCFixup &Fixup, char *Data,
 }
 
 namespace {
+// FIXME: This should be in a separate file.
+class ARMWinCOFFAsmBackend : public ARMAsmBackend {
+public:
+  ARMWinCOFFAsmBackend(const Target &T, const StringRef &Triple)
+    : ARMAsmBackend(T, Triple, true) { }
+  MCObjectWriter *createObjectWriter(raw_ostream &OS) const override {
+    return createARMWinCOFFObjectWriter(OS, /*Is64Bit=*/false);
+  }
+};
 
 // FIXME: This should be in a separate file.
 // ELF is an ELF of course...
@@ -777,7 +807,9 @@ MCAsmBackend *llvm::createARMAsmBackend(const Target &T,
                                         bool isLittle) {
   Triple TheTriple(TT);
 
-  if (TheTriple.isOSBinFormatMachO()) {
+  switch (TheTriple.getObjectFormat()) {
+  default: llvm_unreachable("unsupported object format");
+  case Triple::MachO: {
     MachO::CPUSubTypeARM CS =
       StringSwitch<MachO::CPUSubTypeARM>(TheTriple.getArchName())
       .Cases("armv4t", "thumbv4t", MachO::CPU_SUBTYPE_ARM_V4T)
@@ -792,15 +824,14 @@ MCAsmBackend *llvm::createARMAsmBackend(const Target &T,
 
     return new DarwinARMAsmBackend(T, TT, CS);
   }
-
-#if 0
-  // FIXME: Introduce yet another checker but assert(0).
-  if (TheTriple.isOSBinFormatCOFF())
-    assert(0 && "Windows not supported on ARM");
-#endif
-
-  uint8_t OSABI = MCELFObjectTargetWriter::getOSABI(Triple(TT).getOS());
-  return new ELFARMAsmBackend(T, TT, OSABI, isLittle);
+  case Triple::COFF:
+    assert(TheTriple.isOSWindows() && "non-Windows ARM COFF is not supported");
+    return new ARMWinCOFFAsmBackend(T, TT);
+  case Triple::ELF:
+    assert(TheTriple.isOSBinFormatELF() && "using ELF for non-ELF target");
+    uint8_t OSABI = MCELFObjectTargetWriter::getOSABI(Triple(TT).getOS());
+    return new ELFARMAsmBackend(T, TT, OSABI, isLittle);
+  }
 }
 
 MCAsmBackend *llvm::createARMLEAsmBackend(const Target &T,
diff --git a/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp b/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
index a4661b1..1c84263 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
@@ -74,7 +74,7 @@ unsigned ARMELFObjectWriter::GetRelocType(const MCValue &Target,
 unsigned ARMELFObjectWriter::GetRelocTypeInner(const MCValue &Target,
                                                const MCFixup &Fixup,
                                                bool IsPCRel) const  {
-  MCSymbolRefExpr::VariantKind Modifier = Fixup.getAccessVariant();
+  MCSymbolRefExpr::VariantKind Modifier = Target.getAccessVariant();
 
   unsigned Type = 0;
   if (IsPCRel) {
@@ -91,6 +91,9 @@ unsigned ARMELFObjectWriter::GetRelocTypeInner(const MCValue &Target,
       case MCSymbolRefExpr::VK_GOTTPOFF:
         Type = ELF::R_ARM_TLS_IE32;
         break;
+      case MCSymbolRefExpr::VK_GOTPCREL:
+        Type = ELF::R_ARM_GOT_PREL;
+        break;
       }
       break;
     case ARM::fixup_arm_blx:
@@ -167,6 +170,9 @@ unsigned ARMELFObjectWriter::GetRelocTypeInner(const MCValue &Target,
       case MCSymbolRefExpr::VK_GOTOFF:
         Type = ELF::R_ARM_GOTOFF32;
         break;
+      case MCSymbolRefExpr::VK_GOTPCREL:
+        Type = ELF::R_ARM_GOT_PREL;
+        break;
       case MCSymbolRefExpr::VK_ARM_TARGET1:
         Type = ELF::R_ARM_TARGET1;
         break;
diff --git a/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp b/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
index 5a01d26..a4d13ed 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
@@ -30,6 +30,7 @@
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCInstPrinter.h"
+#include "llvm/MC/MCObjectFileInfo.h"
 #include "llvm/MC/MCObjectStreamer.h"
 #include "llvm/MC/MCRegisterInfo.h"
 #include "llvm/MC/MCSection.h"
@@ -62,7 +63,7 @@ static const char *GetFPUName(unsigned ID) {
 #define ARM_FPU_NAME(NAME, ID) case ARM::ID: return NAME;
 #include "ARMFPUName.def"
   }
-  return NULL;
+  return nullptr;
 }
 
 static const char *GetArchName(unsigned ID) {
@@ -75,7 +76,7 @@ static const char *GetArchName(unsigned ID) {
 #define ARM_ARCH_ALIAS(NAME, ID) /* empty */
 #include "ARMArchName.def"
   }
-  return NULL;
+  return nullptr;
 }
 
 static const char *GetArchDefaultCPUName(unsigned ID) {
@@ -88,7 +89,7 @@ static const char *GetArchDefaultCPUName(unsigned ID) {
 #define ARM_ARCH_ALIAS(NAME, ID) /* empty */
 #include "ARMArchName.def"
   }
-  return NULL;
+  return nullptr;
 }
 
 static unsigned GetArchDefaultCPUArch(unsigned ID) {
@@ -139,6 +140,7 @@ class ARMTargetAsmStreamer : public ARMTargetStreamer {
   void finishAttributeSection() override;
 
   void AnnotateTLSDescriptorSequence(const MCSymbolRefExpr *SRE) override;
+  void emitThumbSet(MCSymbol *Symbol, const MCExpr *Value) override;
 
 public:
   ARMTargetAsmStreamer(MCStreamer &S, formatted_raw_ostream &OS,
@@ -260,6 +262,10 @@ ARMTargetAsmStreamer::AnnotateTLSDescriptorSequence(const MCSymbolRefExpr *S) {
   OS << "\t.tlsdescseq\t" << S->getSymbol().getName();
 }
 
+void ARMTargetAsmStreamer::emitThumbSet(MCSymbol *Symbol, const MCExpr *Value) {
+  OS << "\t.thumb_set\t" << *Symbol << ", " << *Value << '\n';
+}
+
 void ARMTargetAsmStreamer::emitInst(uint32_t Inst, char Suffix) {
   OS << "\t.inst";
   if (Suffix)
@@ -310,7 +316,7 @@ private:
     for (size_t i = 0; i < Contents.size(); ++i)
       if (Contents[i].Tag == Attribute)
         return &Contents[i];
-    return 0;
+    return nullptr;
   }
 
   void setAttributeItem(unsigned Attribute, unsigned Value,
@@ -406,8 +412,10 @@ private:
   void emitFPU(unsigned FPU) override;
   void emitInst(uint32_t Inst, char Suffix = '\0') override;
   void finishAttributeSection() override;
+  void emitLabel(MCSymbol *Symbol) override;
 
   void AnnotateTLSDescriptorSequence(const MCSymbolRefExpr *SRE) override;
+  void emitThumbSet(MCSymbol *Symbol, const MCExpr *Value) override;
 
   size_t calculateContentSize() const;
 
@@ -415,7 +423,7 @@ public:
   ARMTargetELFStreamer(MCStreamer &S)
     : ARMTargetStreamer(S), CurrentVendor("aeabi"), FPU(ARM::INVALID_FPU),
       Arch(ARM::INVALID_ARCH), EmittedArch(ARM::INVALID_ARCH),
-      AttributeSection(0) {}
+      AttributeSection(nullptr) {}
 };
 
 /// Extend the generic ELFStreamer class so that it can emit mapping symbols at
@@ -531,7 +539,8 @@ public:
   /// This is one of the functions used to emit data into an ELF section, so the
   /// ARM streamer overrides it to add the appropriate mapping symbol ($d) if
   /// necessary.
-  void EmitValueImpl(const MCExpr *Value, unsigned Size) override {
+  void EmitValueImpl(const MCExpr *Value, unsigned Size,
+                     const SMLoc &Loc) override {
     EmitDataMappingSymbol();
     MCELFStreamer::EmitValueImpl(Value, Size);
   }
@@ -600,12 +609,8 @@ private:
   }
 
   void EmitThumbFunc(MCSymbol *Func) override {
-    // FIXME: Anything needed here to flag the function as thumb?
-
     getAssembler().setIsThumbFunc(Func);
-
-    MCSymbolData &SD = getAssembler().getOrCreateSymbolData(*Func);
-    SD.setFlags(SD.getFlags() | ELF_Other_ThumbFunc);
+    EmitSymbolAttribute(Func, MCSA_ELF_TypeFunction);
   }
 
   // Helper functions for ARM exception handling directives
@@ -980,10 +985,35 @@ void ARMTargetELFStreamer::finishAttributeSection() {
   Contents.clear();
   FPU = ARM::INVALID_FPU;
 }
+
+void ARMTargetELFStreamer::emitLabel(MCSymbol *Symbol) {
+  ARMELFStreamer &Streamer = getStreamer();
+  if (!Streamer.IsThumb)
+    return;
+
+  const MCSymbolData &SD = Streamer.getOrCreateSymbolData(Symbol);
+  if (MCELF::GetType(SD) & (ELF::STT_FUNC << ELF_STT_Shift))
+    Streamer.EmitThumbFunc(Symbol);
+}
+
 void
 ARMTargetELFStreamer::AnnotateTLSDescriptorSequence(const MCSymbolRefExpr *S) {
   getStreamer().EmitFixup(S, FK_Data_4);
 }
+
+void ARMTargetELFStreamer::emitThumbSet(MCSymbol *Symbol, const MCExpr *Value) {
+  if (const MCSymbolRefExpr *SRE = dyn_cast<MCSymbolRefExpr>(Value)) {
+    const MCSymbol &Sym = SRE->getSymbol();
+    if (!Sym.isDefined()) {
+      getStreamer().EmitAssignment(Symbol, Value);
+      return;
+    }
+  }
+
+  getStreamer().EmitThumbFunc(Symbol);
+  getStreamer().EmitAssignment(Symbol, Value);
+}
+
 void ARMTargetELFStreamer::emitInst(uint32_t Inst, char Suffix) {
   getStreamer().emitInst(Inst, Suffix);
 }
@@ -1012,7 +1042,7 @@ inline void ARMELFStreamer::SwitchToEHSection(const char *Prefix,
   }
 
   // Get .ARM.extab or .ARM.exidx section
-  const MCSectionELF *EHSection = NULL;
+  const MCSectionELF *EHSection = nullptr;
   if (const MCSymbol *Group = FnSection.getGroup()) {
     EHSection = getContext().getELFSection(
       EHSecName, Type, Flags | ELF::SHF_GROUP, Kind,
@@ -1049,9 +1079,9 @@ void ARMELFStreamer::EmitFixup(const MCExpr *Expr, MCFixupKind Kind) {
 }
 
 void ARMELFStreamer::Reset() {
-  ExTab = NULL;
-  FnStart = NULL;
-  Personality = NULL;
+  ExTab = nullptr;
+  FnStart = nullptr;
+  Personality = nullptr;
   PersonalityIndex = ARM::EHABI::NUM_PERSONALITY_INDEX;
   FPReg = ARM::SP;
   FPOffset = 0;
@@ -1065,7 +1095,7 @@ void ARMELFStreamer::Reset() {
 }
 
 void ARMELFStreamer::emitFnStart() {
-  assert(FnStart == 0);
+  assert(FnStart == nullptr);
   FnStart = getContext().CreateTempSymbol();
   EmitLabel(FnStart);
 }
@@ -1104,11 +1134,14 @@ void ARMELFStreamer::emitFnEnd() {
     // the second word of exception index table entry.  The size of the unwind
     // opcodes should always be 4 bytes.
     assert(PersonalityIndex == ARM::EHABI::AEABI_UNWIND_CPP_PR0 &&
-           "Compact model must use __aeabi_cpp_unwind_pr0 as personality");
+           "Compact model must use __aeabi_unwind_cpp_pr0 as personality");
     assert(Opcodes.size() == 4u &&
-           "Unwind opcode size for __aeabi_cpp_unwind_pr0 must be equal to 4");
-    EmitBytes(StringRef(reinterpret_cast<const char*>(Opcodes.data()),
-                        Opcodes.size()));
+           "Unwind opcode size for __aeabi_unwind_cpp_pr0 must be equal to 4");
+    uint64_t Intval = Opcodes[0] |
+                      Opcodes[1] << 8 |
+                      Opcodes[2] << 16 |
+                      Opcodes[3] << 24;
+    EmitIntValue(Intval, Opcodes.size());
   }
 
   // Switch to the section containing FnStart
@@ -1180,8 +1213,15 @@ void ARMELFStreamer::FlushUnwindOpcodes(bool NoHandlerData) {
   }
 
   // Emit unwind opcodes
-  EmitBytes(StringRef(reinterpret_cast<const char *>(Opcodes.data()),
-                      Opcodes.size()));
+  assert((Opcodes.size() % 4) == 0 &&
+         "Unwind opcode size for __aeabi_cpp_unwind_pr0 must be multiple of 4");
+  for (unsigned I = 0; I != Opcodes.size(); I += 4) {
+    uint64_t Intval = Opcodes[I] |
+                      Opcodes[I + 1] << 8 |
+                      Opcodes[I + 2] << 16 |
+                      Opcodes[I + 3] << 24;
+    EmitIntValue(Intval, 4);
+  }
 
   // According to ARM EHABI section 9.2, if the __aeabi_unwind_cpp_pr1() or
   // __aeabi_unwind_cpp_pr2() is used, then the handler data must be emitted
@@ -1283,13 +1323,11 @@ void ARMELFStreamer::emitUnwindRaw(int64_t Offset,
 namespace llvm {
 
 MCStreamer *createMCAsmStreamer(MCContext &Ctx, formatted_raw_ostream &OS,
-                                bool isVerboseAsm, bool useCFI,
-                                bool useDwarfDirectory,
+                                bool isVerboseAsm, bool useDwarfDirectory,
                                 MCInstPrinter *InstPrint, MCCodeEmitter *CE,
                                 MCAsmBackend *TAB, bool ShowInst) {
-  MCStreamer *S =
-      llvm::createAsmStreamer(Ctx, OS, isVerboseAsm, useCFI, useDwarfDirectory,
-                              InstPrint, CE, TAB, ShowInst);
+  MCStreamer *S = llvm::createAsmStreamer(
+      Ctx, OS, isVerboseAsm, useDwarfDirectory, InstPrint, CE, TAB, ShowInst);
   new ARMTargetAsmStreamer(*S, OS, *InstPrint, isVerboseAsm);
   return S;
 }
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.cpp b/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.cpp
index b7f96e0..7a19208 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.cpp
@@ -25,7 +25,7 @@ ARMMCAsmInfoDarwin::ARMMCAsmInfoDarwin(StringRef TT) {
       (TheTriple.getArch() == Triple::thumbeb))
     IsLittleEndian = false;
 
-  Data64bitsDirective = 0;
+  Data64bitsDirective = nullptr;
   CommentString = "@";
   Code16Directive = ".code\t16";
   Code32Directive = ".code\t32";
@@ -50,7 +50,7 @@ ARMELFMCAsmInfo::ARMELFMCAsmInfo(StringRef TT) {
   // ".comm align is in bytes but .align is pow-2."
   AlignmentIsInBytes = false;
 
-  Data64bitsDirective = 0;
+  Data64bitsDirective = nullptr;
   CommentString = "@";
   Code16Directive = ".code\t16";
   Code32Directive = ".code\t32";
@@ -59,7 +59,14 @@ ARMELFMCAsmInfo::ARMELFMCAsmInfo(StringRef TT) {
   SupportsDebugInformation = true;
 
   // Exceptions handling
-  ExceptionsType = ExceptionHandling::ARM;
+  switch (TheTriple.getOS()) {
+  case Triple::NetBSD:
+    ExceptionsType = ExceptionHandling::DwarfCFI;
+    break;
+  default:
+    ExceptionsType = ExceptionHandling::ARM;
+    break;
+  }
 
   // foo(plt) instead of foo@plt
   UseParensForSymbolVariant = true;
@@ -89,6 +96,7 @@ void ARMCOFFMCAsmInfoGNU::anchor() { }
 
 ARMCOFFMCAsmInfoGNU::ARMCOFFMCAsmInfoGNU() {
   AlignmentIsInBytes = false;
+  HasSingleParameterDotFile = true;
 
   CommentString = "@";
   Code16Directive = ".code\t16";
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.h b/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.h
index beaf6a4..51cfa0a 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.h
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCAsmInfo.h
@@ -35,13 +35,13 @@ namespace llvm {
   };
 
   class ARMCOFFMCAsmInfoMicrosoft : public MCAsmInfoMicrosoft {
-    void anchor();
+    void anchor() override;
   public:
     explicit ARMCOFFMCAsmInfoMicrosoft();
   };
 
   class ARMCOFFMCAsmInfoGNU : public MCAsmInfoGNUCOFF {
-    void anchor();
+    void anchor() override;
   public:
     explicit ARMCOFFMCAsmInfoGNU();
   };
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp b/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
index 5564e0a..5b51a52 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
@@ -11,7 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "mccodeemitter"
 #include "MCTargetDesc/ARMMCTargetDesc.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
 #include "MCTargetDesc/ARMBaseInfo.h"
@@ -31,6 +30,8 @@
 
 using namespace llvm;
 
+#define DEBUG_TYPE "mccodeemitter"
+
 STATISTIC(MCNumEmitted, "Number of MC instructions emitted.");
 STATISTIC(MCNumCPRelocations, "Number of constant pool relocations created.");
 
@@ -1036,16 +1037,17 @@ ARMMCCodeEmitter::getHiLo16ImmOpValue(const MCInst &MI, unsigned OpIdx,
                                        : ARM::fixup_arm_movw_lo16);
       break;
     }
+
     Fixups.push_back(MCFixup::Create(0, E, Kind, MI.getLoc()));
     return 0;
   }
   // If the expression doesn't have :upper16: or :lower16: on it,
-  // it's just a plain immediate expression, and those evaluate to
+  // it's just a plain immediate expression, previously those evaluated to
   // the lower 16 bits of the expression regardless of whether
-  // we have a movt or a movw.
-  Kind = MCFixupKind(isThumb2(STI) ? ARM::fixup_t2_movw_lo16
-                                   : ARM::fixup_arm_movw_lo16);
-  Fixups.push_back(MCFixup::Create(0, E, Kind, MI.getLoc()));
+  // we have a movt or a movw, but that led to misleadingly results.
+  // This is now disallowed in the the AsmParser in validateInstruction()
+  // so this should never happen.
+  assert(0 && "expression without :upper16: or :lower16:");
   return 0;
 }
 
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCExpr.cpp b/lib/Target/ARM/MCTargetDesc/ARMMCExpr.cpp
index fc8505b..87ea875 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCExpr.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCExpr.cpp
@@ -7,12 +7,13 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "armmcexpr"
 #include "ARMMCExpr.h"
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "armmcexpr"
+
 const ARMMCExpr*
 ARMMCExpr::Create(VariantKind Kind, const MCExpr *Expr,
                        MCContext &Ctx) {
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
index 949a3d5..04d63a7 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
@@ -21,6 +21,7 @@
 #include "llvm/MC/MCInstrAnalysis.h"
 #include "llvm/MC/MCInstrInfo.h"
 #include "llvm/MC/MCRegisterInfo.h"
+#include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSubtargetInfo.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/TargetRegistry.h"
@@ -106,9 +107,11 @@ std::string ARM_MC::ParseARMTriple(StringRef TT, StringRef CPU) {
     unsigned SubVer = TT[Idx];
     if (SubVer == '8') {
       if (NoCPU)
-        // v8a: FeatureDB, FeatureFPARMv8, FeatureNEON, FeatureDSPThumb2, FeatureMP,
-        //      FeatureHWDiv, FeatureHWDivARM, FeatureTrustZone, FeatureT2XtPk, FeatureCrypto, FeatureCRC
-        ARMArchFeature = "+v8,+db,+fp-armv8,+neon,+t2dsp,+mp,+hwdiv,+hwdiv-arm,+trustzone,+t2xtpk,+crypto,+crc";
+        // v8a: FeatureDB, FeatureFPARMv8, FeatureNEON, FeatureDSPThumb2,
+        //      FeatureMP, FeatureHWDiv, FeatureHWDivARM, FeatureTrustZone,
+        //      FeatureT2XtPk, FeatureCrypto, FeatureCRC
+        ARMArchFeature = "+v8,+db,+fp-armv8,+neon,+t2dsp,+mp,+hwdiv,+hwdiv-arm,"
+                         "+trustzone,+t2xtpk,+crypto,+crc";
       else
         // Use CPU to figure out the exact features
         ARMArchFeature = "+v8";
@@ -245,7 +248,7 @@ static MCAsmInfo *createARMMCAsmInfo(const MCRegisterInfo &MRI, StringRef TT) {
   }
 
   unsigned Reg = MRI.getDwarfRegNum(ARM::SP, true);
-  MAI->addInitialFrameState(MCCFIInstruction::createDefCfa(0, Reg, 0));
+  MAI->addInitialFrameState(MCCFIInstruction::createDefCfa(nullptr, Reg, 0));
 
   return MAI;
 }
@@ -273,18 +276,20 @@ static MCStreamer *createMCStreamer(const Target &T, StringRef TT,
                                     bool NoExecStack) {
   Triple TheTriple(TT);
 
-  if (TheTriple.isOSBinFormatMachO()) {
+  switch (TheTriple.getObjectFormat()) {
+  default: llvm_unreachable("unsupported object format");
+  case Triple::MachO: {
     MCStreamer *S = createMachOStreamer(Ctx, MAB, OS, Emitter, false);
     new ARMTargetStreamer(*S);
     return S;
   }
-
-  if (TheTriple.isOSWindows()) {
-    llvm_unreachable("ARM does not support Windows COFF format");
+  case Triple::COFF:
+    assert(TheTriple.isOSWindows() && "non-Windows ARM COFF is not supported");
+    return createARMWinCOFFStreamer(Ctx, MAB, *Emitter, OS);
+  case Triple::ELF:
+    return createARMELFStreamer(Ctx, MAB, OS, Emitter, false, NoExecStack,
+                                TheTriple.getArch() == Triple::thumb);
   }
-
-  return createARMELFStreamer(Ctx, MAB, OS, Emitter, false, NoExecStack,
-                              TheTriple.getArch() == Triple::thumb);
 }
 
 static MCInstPrinter *createARMMCInstPrinter(const Target &T,
@@ -295,7 +300,7 @@ static MCInstPrinter *createARMMCInstPrinter(const Target &T,
                                              const MCSubtargetInfo &STI) {
   if (SyntaxVariant == 0)
     return new ARMInstPrinter(MAI, MII, MRI, STI);
-  return 0;
+  return nullptr;
 }
 
 static MCRelocationInfo *createARMMCRelocationInfo(StringRef TT,
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.h b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.h
index e81876f..8853a8c 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.h
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.h
@@ -47,8 +47,7 @@ namespace ARM_MC {
 }
 
 MCStreamer *createMCAsmStreamer(MCContext &Ctx, formatted_raw_ostream &OS,
-                                bool isVerboseAsm, bool useCFI,
-                                bool useDwarfDirectory,
+                                bool isVerboseAsm, bool useDwarfDirectory,
                                 MCInstPrinter *InstPrint, MCCodeEmitter *CE,
                                 MCAsmBackend *TAB, bool ShowInst);
 
@@ -78,6 +77,11 @@ MCAsmBackend *createThumbLEAsmBackend(const Target &T, const MCRegisterInfo &MRI
 MCAsmBackend *createThumbBEAsmBackend(const Target &T, const MCRegisterInfo &MRI,
                                       StringRef TT, StringRef CPU);
 
+/// createARMWinCOFFStreamer - Construct a PE/COFF machine code streamer which
+/// will generate a PE/COFF object file.
+MCStreamer *createARMWinCOFFStreamer(MCContext &Context, MCAsmBackend &MAB,
+                                     MCCodeEmitter &Emitter, raw_ostream &OS);
+
 /// createARMELFObjectWriter - Construct an ELF Mach-O object writer.
 MCObjectWriter *createARMELFObjectWriter(raw_ostream &OS,
                                          uint8_t OSABI,
@@ -89,6 +93,8 @@ MCObjectWriter *createARMMachObjectWriter(raw_ostream &OS,
                                           uint32_t CPUType,
                                           uint32_t CPUSubtype);
 
+/// createARMWinCOFFObjectWriter - Construct an ARM PE/COFF object writer.
+MCObjectWriter *createARMWinCOFFObjectWriter(raw_ostream &OS, bool Is64Bit);
 
 /// createARMMachORelocationInfo - Construct ARM Mach-O relocation info.
 MCRelocationInfo *createARMMachORelocationInfo(MCContext &Ctx);
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMachObjectWriter.cpp b/lib/Target/ARM/MCTargetDesc/ARMMachObjectWriter.cpp
index 3bf5cf1..ecfa4e5 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMachObjectWriter.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMachObjectWriter.cpp
@@ -156,7 +156,7 @@ RecordARMScatteredHalfRelocation(MachObjectWriter *Writer,
 
   // See <reloc.h>.
   const MCSymbol *A = &Target.getSymA()->getSymbol();
-  MCSymbolData *A_SD = &Asm.getSymbolData(*A);
+  const MCSymbolData *A_SD = &Asm.getSymbolData(*A);
 
   if (!A_SD->getFragment())
     Asm.getContext().FatalError(Fixup.getLoc(),
@@ -170,7 +170,7 @@ RecordARMScatteredHalfRelocation(MachObjectWriter *Writer,
   FixedValue += SecAddr;
 
   if (const MCSymbolRefExpr *B = Target.getSymB()) {
-    MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol());
+    const MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol());
 
     if (!B_SD->getFragment())
       Asm.getContext().FatalError(Fixup.getLoc(),
@@ -206,11 +206,11 @@ RecordARMScatteredHalfRelocation(MachObjectWriter *Writer,
     // The thumb bit shouldn't be set in the 'other-half' bit of the
     // relocation, but it will be set in FixedValue if the base symbol
     // is a thumb function. Clear it out here.
-    if (A_SD->getFlags() & SF_ThumbFunc)
+    if (Asm.isThumbFunc(A))
       FixedValue &= 0xfffffffe;
     break;
   case ARM::fixup_t2_movt_hi16:
-    if (A_SD->getFlags() & SF_ThumbFunc)
+    if (Asm.isThumbFunc(A))
       FixedValue &= 0xfffffffe;
     MovtBit = 1;
     // Fallthrough
@@ -259,7 +259,7 @@ void ARMMachObjectWriter::RecordARMScatteredRelocation(MachObjectWriter *Writer,
 
   // See <reloc.h>.
   const MCSymbol *A = &Target.getSymA()->getSymbol();
-  MCSymbolData *A_SD = &Asm.getSymbolData(*A);
+  const MCSymbolData *A_SD = &Asm.getSymbolData(*A);
 
   if (!A_SD->getFragment())
     Asm.getContext().FatalError(Fixup.getLoc(),
@@ -272,7 +272,7 @@ void ARMMachObjectWriter::RecordARMScatteredRelocation(MachObjectWriter *Writer,
   uint32_t Value2 = 0;
 
   if (const MCSymbolRefExpr *B = Target.getSymB()) {
-    MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol());
+    const MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol());
 
     if (!B_SD->getFragment())
       Asm.getContext().FatalError(Fixup.getLoc(),
@@ -378,7 +378,7 @@ void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
   }
 
   // Get the symbol data, if any.
-  MCSymbolData *SD = 0;
+  const MCSymbolData *SD = nullptr;
   if (Target.getSymA())
     SD = &Asm.getSymbolData(Target.getSymA()->getSymbol());
 
diff --git a/lib/Target/ARM/MCTargetDesc/ARMTargetStreamer.cpp b/lib/Target/ARM/MCTargetDesc/ARMTargetStreamer.cpp
index fdc0ed7..e3cfb05 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMTargetStreamer.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMTargetStreamer.cpp
@@ -109,7 +109,7 @@ ConstantPool *
 AssemblerConstantPools::getConstantPool(const MCSection *Section) {
   ConstantPoolMapTy::iterator CP = ConstantPools.find(Section);
   if (CP == ConstantPools.end())
-    return 0;
+    return nullptr;
 
   return &CP->second;
 }
@@ -246,3 +246,7 @@ void ARMTargetStreamer::AnnotateTLSDescriptorSequence(
     const MCSymbolRefExpr *SRE) {
   llvm_unreachable("unimplemented");
 }
+
+void ARMTargetStreamer::emitThumbSet(MCSymbol *Symbol, const MCExpr *Value) {
+  llvm_unreachable("unimplemented");
+}
diff --git a/lib/Target/ARM/MCTargetDesc/ARMWinCOFFObjectWriter.cpp b/lib/Target/ARM/MCTargetDesc/ARMWinCOFFObjectWriter.cpp
new file mode 100644
index 0000000..d31f1f4
--- /dev/null
+++ b/lib/Target/ARM/MCTargetDesc/ARMWinCOFFObjectWriter.cpp
@@ -0,0 +1,82 @@
+//===-- ARMWinCOFFObjectWriter.cpp - ARM Windows COFF Object Writer -- C++ -==//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "MCTargetDesc/ARMFixupKinds.h"
+#include "llvm/MC/MCFixup.h"
+#include "llvm/MC/MCValue.h"
+#include "llvm/MC/MCWinCOFFObjectWriter.h"
+#include "llvm/Support/COFF.h"
+#include "llvm/Support/Debug.h"
+
+using namespace llvm;
+
+namespace {
+class ARMWinCOFFObjectWriter : public MCWinCOFFObjectTargetWriter {
+public:
+  ARMWinCOFFObjectWriter(bool Is64Bit)
+    : MCWinCOFFObjectTargetWriter(COFF::IMAGE_FILE_MACHINE_ARMNT) {
+    assert(!Is64Bit && "AArch64 support not yet implemented");
+  }
+  virtual ~ARMWinCOFFObjectWriter() { }
+
+  unsigned getRelocType(const MCValue &Target, const MCFixup &Fixup,
+                        bool IsCrossSection) const override;
+
+  bool recordRelocation(const MCFixup &) const override;
+};
+
+unsigned ARMWinCOFFObjectWriter::getRelocType(const MCValue &Target,
+                                              const MCFixup &Fixup,
+                                              bool IsCrossSection) const {
+  assert(getMachine() == COFF::IMAGE_FILE_MACHINE_ARMNT &&
+         "AArch64 support not yet implemented");
+
+  MCSymbolRefExpr::VariantKind Modifier =
+    Target.isAbsolute() ? MCSymbolRefExpr::VK_None : Target.getSymA()->getKind();
+
+  switch (static_cast<unsigned>(Fixup.getKind())) {
+  default: llvm_unreachable("unsupported relocation type");
+  case FK_Data_4:
+    switch (Modifier) {
+    case MCSymbolRefExpr::VK_COFF_IMGREL32:
+      return COFF::IMAGE_REL_ARM_ADDR32NB;
+    case MCSymbolRefExpr::VK_SECREL:
+      return COFF::IMAGE_REL_ARM_SECREL;
+    default:
+      return COFF::IMAGE_REL_ARM_ADDR32;
+    }
+  case FK_SecRel_2:
+    return COFF::IMAGE_REL_ARM_SECTION;
+  case FK_SecRel_4:
+    return COFF::IMAGE_REL_ARM_SECREL;
+  case ARM::fixup_t2_condbranch:
+    return COFF::IMAGE_REL_ARM_BRANCH20T;
+  case ARM::fixup_t2_uncondbranch:
+    return COFF::IMAGE_REL_ARM_BRANCH24T;
+  case ARM::fixup_arm_thumb_bl:
+  case ARM::fixup_arm_thumb_blx:
+    return COFF::IMAGE_REL_ARM_BLX23T;
+  case ARM::fixup_t2_movw_lo16:
+  case ARM::fixup_t2_movt_hi16:
+    return COFF::IMAGE_REL_ARM_MOV32T;
+  }
+}
+
+bool ARMWinCOFFObjectWriter::recordRelocation(const MCFixup &Fixup) const {
+  return static_cast<unsigned>(Fixup.getKind()) != ARM::fixup_t2_movt_hi16;
+}
+}
+
+namespace llvm {
+MCObjectWriter *createARMWinCOFFObjectWriter(raw_ostream &OS, bool Is64Bit) {
+  MCWinCOFFObjectTargetWriter *MOTW = new ARMWinCOFFObjectWriter(Is64Bit);
+  return createWinCOFFObjectWriter(MOTW, OS);
+}
+}
+
diff --git a/lib/Target/ARM/MCTargetDesc/ARMWinCOFFStreamer.cpp b/lib/Target/ARM/MCTargetDesc/ARMWinCOFFStreamer.cpp
new file mode 100644
index 0000000..b344ced
--- /dev/null
+++ b/lib/Target/ARM/MCTargetDesc/ARMWinCOFFStreamer.cpp
@@ -0,0 +1,46 @@
+//===-- ARMWinCOFFStreamer.cpp - ARM Target WinCOFF Streamer ----*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "ARMMCTargetDesc.h"
+#include "llvm/MC/MCWinCOFFStreamer.h"
+
+using namespace llvm;
+
+namespace {
+class ARMWinCOFFStreamer : public MCWinCOFFStreamer {
+public:
+  ARMWinCOFFStreamer(MCContext &C, MCAsmBackend &AB, MCCodeEmitter &CE,
+                     raw_ostream &OS)
+    : MCWinCOFFStreamer(C, AB, CE, OS) { }
+
+  void EmitAssemblerFlag(MCAssemblerFlag Flag) override;
+  void EmitThumbFunc(MCSymbol *Symbol) override;
+};
+
+void ARMWinCOFFStreamer::EmitAssemblerFlag(MCAssemblerFlag Flag) {
+  switch (Flag) {
+  default: llvm_unreachable("not implemented");
+  case MCAF_SyntaxUnified:
+  case MCAF_Code16:
+    break;
+  }
+}
+
+void ARMWinCOFFStreamer::EmitThumbFunc(MCSymbol *Symbol) {
+  getAssembler().setIsThumbFunc(Symbol);
+}
+}
+
+namespace llvm {
+MCStreamer *createARMWinCOFFStreamer(MCContext &Context, MCAsmBackend &MAB,
+                                     MCCodeEmitter &Emitter, raw_ostream &OS) {
+  return new ARMWinCOFFStreamer(Context, MAB, Emitter, OS);
+}
+}
+
diff --git a/lib/Target/ARM/MCTargetDesc/Android.mk b/lib/Target/ARM/MCTargetDesc/Android.mk
index 074d29e..a5827f7 100644
--- a/lib/Target/ARM/MCTargetDesc/Android.mk
+++ b/lib/Target/ARM/MCTargetDesc/Android.mk
@@ -17,7 +17,9 @@ arm_mc_desc_SRC_FILES := \
   ARMMachObjectWriter.cpp \
   ARMMachORelocationInfo.cpp \
   ARMTargetStreamer.cpp \
-  ARMUnwindOpAsm.cpp
+  ARMUnwindOpAsm.cpp \
+  ARMWinCOFFObjectWriter.cpp \
+  ARMWinCOFFStreamer.cpp \
 
 # For the host
 # =====================================================
diff --git a/lib/Target/ARM/MCTargetDesc/CMakeLists.txt b/lib/Target/ARM/MCTargetDesc/CMakeLists.txt
index 06812d4..9582e8c 100644
--- a/lib/Target/ARM/MCTargetDesc/CMakeLists.txt
+++ b/lib/Target/ARM/MCTargetDesc/CMakeLists.txt
@@ -1,14 +1,16 @@
 add_llvm_library(LLVMARMDesc
   ARMAsmBackend.cpp
   ARMELFObjectWriter.cpp
+  ARMELFObjectWriter.cpp
   ARMELFStreamer.cpp
+  ARMMachObjectWriter.cpp
+  ARMMachORelocationInfo.cpp
   ARMMCAsmInfo.cpp
   ARMMCCodeEmitter.cpp
   ARMMCExpr.cpp
   ARMMCTargetDesc.cpp
-  ARMMachObjectWriter.cpp
-  ARMELFObjectWriter.cpp
   ARMTargetStreamer.cpp
   ARMUnwindOpAsm.cpp
-  ARMMachORelocationInfo.cpp
+  ARMWinCOFFObjectWriter.cpp
+  ARMWinCOFFStreamer.cpp
   )
diff --git a/lib/Target/ARM/MLxExpansionPass.cpp b/lib/Target/ARM/MLxExpansionPass.cpp
index 80af859..f6d24e9 100644
--- a/lib/Target/ARM/MLxExpansionPass.cpp
+++ b/lib/Target/ARM/MLxExpansionPass.cpp
@@ -12,7 +12,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "mlx-expansion"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMSubtarget.h"
@@ -28,6 +27,8 @@
 #include "llvm/Target/TargetRegisterInfo.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "mlx-expansion"
+
 static cl::opt<bool>
 ForceExapnd("expand-all-fp-mlx", cl::init(false), cl::Hidden);
 static cl::opt<unsigned>
@@ -73,7 +74,7 @@ namespace {
 }
 
 void MLxExpansion::clearStack() {
-  std::fill(LastMIs, LastMIs + 4, (MachineInstr*)0);
+  std::fill(LastMIs, LastMIs + 4, nullptr);
   MIIdx = 0;
 }
 
@@ -88,7 +89,7 @@ MachineInstr *MLxExpansion::getAccDefMI(MachineInstr *MI) const {
   // real definition MI. This is important for _sfp instructions.
   unsigned Reg = MI->getOperand(1).getReg();
   if (TargetRegisterInfo::isPhysicalRegister(Reg))
-    return 0;
+    return nullptr;
 
   MachineBasicBlock *MBB = MI->getParent();
   MachineInstr *DefMI = MRI->getVRegDef(Reg);
@@ -352,7 +353,7 @@ bool MLxExpansion::ExpandFPMLxInstructions(MachineBasicBlock &MBB) {
     if (Domain == ARMII::DomainGeneral) {
       if (++Skip == 2)
         // Assume dual issues of non-VFP / NEON instructions.
-        pushStack(0);
+        pushStack(nullptr);
     } else {
       Skip = 0;
 
diff --git a/lib/Target/ARM/README-Thumb.txt b/lib/Target/ARM/README-Thumb.txt
index a64707e..f4d9be3 100644
--- a/lib/Target/ARM/README-Thumb.txt
+++ b/lib/Target/ARM/README-Thumb.txt
@@ -215,10 +215,6 @@ etc. Almost all Thumb instructions clobber condition code.
 
 //===---------------------------------------------------------------------===//
 
-Add ldmia, stmia support.
-
-//===---------------------------------------------------------------------===//
-
 Thumb load / store address mode offsets are scaled. The values kept in the
 instruction operands are pre-scale values. This probably ought to be changed
 to avoid extra work when we convert Thumb2 instructions to Thumb1 instructions.
diff --git a/lib/Target/ARM/Thumb1FrameLowering.cpp b/lib/Target/ARM/Thumb1FrameLowering.cpp
index 2224652..be29dc5 100644
--- a/lib/Target/ARM/Thumb1FrameLowering.cpp
+++ b/lib/Target/ARM/Thumb1FrameLowering.cpp
@@ -293,7 +293,7 @@ void Thumb1FrameLowering::emitPrologue(MachineFunction &MF) const {
     AFI->setShouldRestoreSPFromFP(true);
 }
 
-static bool isCSRestore(MachineInstr *MI, const uint16_t *CSRegs) {
+static bool isCSRestore(MachineInstr *MI, const MCPhysReg *CSRegs) {
   if (MI->getOpcode() == ARM::tLDRspi &&
       MI->getOperand(1).isFI() &&
       isCalleeSavedRegister(MI->getOperand(0).getReg(), CSRegs))
@@ -328,7 +328,7 @@ void Thumb1FrameLowering::emitEpilogue(MachineFunction &MF,
   int NumBytes = (int)MFI->getStackSize();
   assert((unsigned)NumBytes >= ArgRegsSaveSize &&
          "ArgRegsSaveSize is included in NumBytes");
-  const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs();
+  const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs();
   unsigned FramePtr = RegInfo->getFrameRegister(MF);
 
   if (!AFI->hasStackFrame()) {
diff --git a/lib/Target/ARM/Thumb1RegisterInfo.h b/lib/Target/ARM/Thumb1RegisterInfo.h
index 93e2b5a..0c0abbe 100644
--- a/lib/Target/ARM/Thumb1RegisterInfo.h
+++ b/lib/Target/ARM/Thumb1RegisterInfo.h
@@ -56,7 +56,7 @@ public:
                              unsigned Reg) const override;
   void eliminateFrameIndex(MachineBasicBlock::iterator II,
                            int SPAdj, unsigned FIOperandNum,
-                           RegScavenger *RS = NULL) const override;
+                           RegScavenger *RS = nullptr) const override;
 };
 }
 
diff --git a/lib/Target/ARM/Thumb2ITBlockPass.cpp b/lib/Target/ARM/Thumb2ITBlockPass.cpp
index 406dbe0..edb9ff3 100644
--- a/lib/Target/ARM/Thumb2ITBlockPass.cpp
+++ b/lib/Target/ARM/Thumb2ITBlockPass.cpp
@@ -7,7 +7,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "thumb2-it"
 #include "ARM.h"
 #include "ARMMachineFunctionInfo.h"
 #include "Thumb2InstrInfo.h"
@@ -19,6 +18,8 @@
 #include "llvm/CodeGen/MachineInstrBundle.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "thumb2-it"
+
 STATISTIC(NumITs,        "Number of IT blocks inserted");
 STATISTIC(NumMovedInsts, "Number of predicated instructions moved");
 
diff --git a/lib/Target/ARM/Thumb2SizeReduction.cpp b/lib/Target/ARM/Thumb2SizeReduction.cpp
index 04b83fb..6267ecf 100644
--- a/lib/Target/ARM/Thumb2SizeReduction.cpp
+++ b/lib/Target/ARM/Thumb2SizeReduction.cpp
@@ -7,7 +7,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "t2-reduce-size"
 #include "ARM.h"
 #include "ARMBaseInstrInfo.h"
 #include "ARMSubtarget.h"
@@ -25,6 +24,8 @@
 #include "llvm/Target/TargetMachine.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "t2-reduce-size"
+
 STATISTIC(NumNarrows,  "Number of 32-bit instrs reduced to 16-bit ones");
 STATISTIC(Num2Addrs,   "Number of 32-bit instrs reduced to 2addr 16-bit ones");
 STATISTIC(NumLdSts,    "Number of 32-bit load / store reduced to 16-bit ones");
@@ -915,15 +916,14 @@ bool Thumb2SizeReduce::ReduceMBB(MachineBasicBlock &MBB) {
 
   // Yes, CPSR could be livein.
   bool LiveCPSR = MBB.isLiveIn(ARM::CPSR);
-  MachineInstr *BundleMI = 0;
+  MachineInstr *BundleMI = nullptr;
 
-  CPSRDef = 0;
+  CPSRDef = nullptr;
   HighLatencyCPSR = false;
 
   // Check predecessors for the latest CPSRDef.
-  for (MachineBasicBlock::pred_iterator
-       I = MBB.pred_begin(), E = MBB.pred_end(); I != E; ++I) {
-    const MBBInfo &PInfo = BlockInfo[(*I)->getNumber()];
+  for (auto *Pred : MBB.predecessors()) {
+    const MBBInfo &PInfo = BlockInfo[Pred->getNumber()];
     if (!PInfo.Visited) {
       // Since blocks are visited in RPO, this must be a back-edge.
       continue;
@@ -984,7 +984,7 @@ bool Thumb2SizeReduce::ReduceMBB(MachineBasicBlock &MBB) {
     LiveCPSR = UpdateCPSRDef(*MI, LiveCPSR, DefCPSR);
     if (MI->isCall()) {
       // Calls don't really set CPSR.
-      CPSRDef = 0;
+      CPSRDef = nullptr;
       HighLatencyCPSR = false;
       IsSelfLoop = false;
     } else if (DefCPSR) {