Update LLVM for rebase to r212749.

Includes a cherry-pick of: r212948 - fixes a small issue with atomic calls Change-Id: Ib97bd980b59f18142a69506400911a6009d9df18
author: Stephen Hines <srhines@google.com> 2014-07-21 00:45:20 -0700
committer: Stephen Hines <srhines@google.com> 2014-07-25 00:48:57 -0700
commit: cd81d94322a39503e4a3e87b6ee03d4fcb3465fb (patch)
tree: 81b7dd2bb4370a392f31d332a566c903b5744764 /lib/Target/AArch64
parent: 0c5f13c0c4499eaf42ab5e9e2ceabd4e20e36861 (diff)
download: external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.zip
external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.tar.gz
external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.tar.bz2
42 files changed, 1780 insertions, 537 deletions
diff --git a/lib/Target/AArch64/AArch64.td b/lib/Target/AArch64/AArch64.td
index 1ad5ac8..e6a27c3 100644
--- a/lib/Target/AArch64/AArch64.td
+++ b/lib/Target/AArch64/AArch64.td
@@ -60,6 +60,7 @@ def AArch64InstrInfo : InstrInfo;
 // AArch64 Processors supported.
 //
 include "AArch64SchedA53.td"
+include "AArch64SchedA57.td"
 include "AArch64SchedCyclone.td"
 
 def ProcA53     : SubtargetFeature<"a53", "ARMProcFamily", "CortexA53",
@@ -89,7 +90,7 @@ def : ProcessorModel<"generic", NoSchedModel, [FeatureFPARMv8,
                                               FeatureCRC]>;
 
 def : ProcessorModel<"cortex-a53", CortexA53Model, [ProcA53]>;
-def : ProcessorModel<"cortex-a57", NoSchedModel, [ProcA57]>;
+def : ProcessorModel<"cortex-a57", CortexA57Model, [ProcA57]>;
 def : ProcessorModel<"cyclone", CycloneModel, [ProcCyclone]>;
 
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/AArch64/AArch64AddressTypePromotion.cpp b/lib/Target/AArch64/AArch64AddressTypePromotion.cpp
index 04906f6..ab2c4b7 100644
--- a/lib/Target/AArch64/AArch64AddressTypePromotion.cpp
+++ b/lib/Target/AArch64/AArch64AddressTypePromotion.cpp
@@ -214,8 +214,8 @@ AArch64AddressTypePromotion::shouldConsiderSExt(const Instruction *SExt) const {
   if (SExt->getType() != ConsideredSExtType)
     return false;
 
-  for (const Use &U : SExt->uses()) {
-    if (isa<GetElementPtrInst>(*U))
+  for (const User *U : SExt->users()) {
+    if (isa<GetElementPtrInst>(U))
       return true;
   }
 
@@ -267,8 +267,7 @@ AArch64AddressTypePromotion::propagateSignExtension(Instructions &SExtInsts) {
     }
 
     // Now try to get through the chain of definitions.
-    while (isa<Instruction>(SExt->getOperand(0))) {
-      Instruction *Inst = dyn_cast<Instruction>(SExt->getOperand(0));
+    while (auto *Inst = dyn_cast<Instruction>(SExt->getOperand(0))) {
       DEBUG(dbgs() << "Try to get through:\n" << *Inst << '\n');
       if (!canGetThrough(Inst) || !shouldGetThrough(Inst)) {
         // We cannot get through something that is not an Instruction
@@ -285,10 +284,10 @@ AArch64AddressTypePromotion::propagateSignExtension(Instructions &SExtInsts) {
         // assertion on the type as all involved sext operation may have not
         // been moved yet.
         while (!Inst->use_empty()) {
-          Value::use_iterator UseIt = Inst->use_begin();
-          Instruction *UseInst = dyn_cast<Instruction>(*UseIt);
-          assert(UseInst && "Use of sext is not an Instruction!");
-          UseInst->setOperand(UseIt->getOperandNo(), SExt);
+          Use &U = *Inst->use_begin();
+          Instruction *User = dyn_cast<Instruction>(U.getUser());
+          assert(User && "User of sext is not an Instruction!");
+          User->setOperand(U.getOperandNo(), SExt);
         }
         ToRemove.insert(Inst);
         SExt->setOperand(0, Inst->getOperand(0));
@@ -385,11 +384,11 @@ void AArch64AddressTypePromotion::mergeSExts(ValueToInsts &ValToSExtendedUses,
       if (ToRemove.count(Inst))
         continue;
       bool inserted = false;
-      for (auto Pt : CurPts) {
+      for (auto &Pt : CurPts) {
         if (DT.dominates(Inst, Pt)) {
           DEBUG(dbgs() << "Replace all uses of:\n" << *Pt << "\nwith:\n"
                        << *Inst << '\n');
-          (Pt)->replaceAllUsesWith(Inst);
+          Pt->replaceAllUsesWith(Inst);
           ToRemove.insert(Pt);
           Pt = Inst;
           inserted = true;
@@ -436,7 +435,7 @@ void AArch64AddressTypePromotion::analyzeSExtension(Instructions &SExtInsts) {
 
       bool insert = false;
       // #1.
-      for (const Use &U : SExt->uses()) {
+      for (const User *U : SExt->users()) {
         const Instruction *Inst = dyn_cast<GetElementPtrInst>(U);
         if (Inst && Inst->getNumOperands() > 2) {
           DEBUG(dbgs() << "Interesting use in GetElementPtrInst\n" << *Inst
diff --git a/lib/Target/AArch64/AArch64AsmPrinter.cpp b/lib/Target/AArch64/AArch64AsmPrinter.cpp
index c3ee9bb..cd94e24 100644
--- a/lib/Target/AArch64/AArch64AsmPrinter.cpp
+++ b/lib/Target/AArch64/AArch64AsmPrinter.cpp
@@ -211,7 +211,7 @@ void AArch64AsmPrinter::printOperand(const MachineInstr *MI, unsigned OpNum,
   const MachineOperand &MO = MI->getOperand(OpNum);
   switch (MO.getType()) {
   default:
-    assert(0 && "<unknown operand type>");
+    llvm_unreachable("<unknown operand type>");
   case MachineOperand::MO_Register: {
     unsigned Reg = MO.getReg();
     assert(TargetRegisterInfo::isPhysicalRegister(Reg));
diff --git a/lib/Target/AArch64/AArch64BranchRelaxation.cpp b/lib/Target/AArch64/AArch64BranchRelaxation.cpp
index 5209452..484e7e8 100644
--- a/lib/Target/AArch64/AArch64BranchRelaxation.cpp
+++ b/lib/Target/AArch64/AArch64BranchRelaxation.cpp
@@ -291,7 +291,7 @@ static bool isConditionalBranch(unsigned Opc) {
 static MachineBasicBlock *getDestBlock(MachineInstr *MI) {
   switch (MI->getOpcode()) {
   default:
-    assert(0 && "unexpected opcode!");
+    llvm_unreachable("unexpected opcode!");
   case AArch64::TBZW:
   case AArch64::TBNZW:
   case AArch64::TBZX:
@@ -309,7 +309,7 @@ static MachineBasicBlock *getDestBlock(MachineInstr *MI) {
 static unsigned getOppositeConditionOpcode(unsigned Opc) {
   switch (Opc) {
   default:
-    assert(0 && "unexpected opcode!");
+    llvm_unreachable("unexpected opcode!");
   case AArch64::TBNZW:   return AArch64::TBZW;
   case AArch64::TBNZX:   return AArch64::TBZX;
   case AArch64::TBZW:    return AArch64::TBNZW;
@@ -325,7 +325,7 @@ static unsigned getOppositeConditionOpcode(unsigned Opc) {
 static unsigned getBranchDisplacementBits(unsigned Opc) {
   switch (Opc) {
   default:
-    assert(0 && "unexpected opcode!");
+    llvm_unreachable("unexpected opcode!");
   case AArch64::TBNZW:
   case AArch64::TBZW:
   case AArch64::TBNZX:
diff --git a/lib/Target/AArch64/AArch64CallingConvention.td b/lib/Target/AArch64/AArch64CallingConvention.td
index ded2e17..8e8bd3d 100644
--- a/lib/Target/AArch64/AArch64CallingConvention.td
+++ b/lib/Target/AArch64/AArch64CallingConvention.td
@@ -18,9 +18,6 @@ class CCIfAlign<string Align, CCAction A> :
 class CCIfBigEndian<CCAction A> :
   CCIf<"State.getTarget().getDataLayout()->isBigEndian()", A>;
 
-class CCIfUnallocated<string Reg, CCAction A> :
-  CCIf<"!State.isAllocated(AArch64::" # Reg # ")", A>;
-
 //===----------------------------------------------------------------------===//
 // ARM AAPCS64 Calling Convention
 //===----------------------------------------------------------------------===//
@@ -45,7 +42,7 @@ def CC_AArch64_AAPCS : CallingConv<[
 
   // Handle i1, i8, i16, i32, i64, f32, f64 and v2f64 by passing in registers,
   // up to eight each of GPR and FPR.
-  CCIfType<[i1, i8, i16], CCIfUnallocated<"X7", CCPromoteToType<i32>>>,
+  CCIfType<[i1, i8, i16], CCPromoteToType<i32>>,
   CCIfType<[i32], CCAssignToRegWithShadow<[W0, W1, W2, W3, W4, W5, W6, W7],
                                           [X0, X1, X2, X3, X4, X5, X6, X7]>>,
   // i128 is split to two i64s, we can't fit half to register X7.
@@ -120,7 +117,7 @@ def CC_AArch64_DarwinPCS : CallingConv<[
 
   // Handle i1, i8, i16, i32, i64, f32, f64 and v2f64 by passing in registers,
   // up to eight each of GPR and FPR.
-  CCIfType<[i1, i8, i16], CCIfUnallocated<"X7", CCPromoteToType<i32>>>,
+  CCIfType<[i1, i8, i16], CCPromoteToType<i32>>,
   CCIfType<[i32], CCAssignToRegWithShadow<[W0, W1, W2, W3, W4, W5, W6, W7],
                                           [X0, X1, X2, X3, X4, X5, X6, X7]>>,
   // i128 is split to two i64s, we can't fit half to register X7.
@@ -143,8 +140,8 @@ def CC_AArch64_DarwinPCS : CallingConv<[
            CCAssignToReg<[Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
 
   // If more than will fit in registers, pass them on the stack instead.
-  CCIfType<[i1, i8], CCAssignToStack<1, 1>>,
-  CCIfType<[i16], CCAssignToStack<2, 2>>,
+  CCIf<"ValVT == MVT::i1 || ValVT == MVT::i8", CCAssignToStack<1, 1>>,
+  CCIf<"ValVT == MVT::i16", CCAssignToStack<2, 2>>,
   CCIfType<[i32, f32], CCAssignToStack<4, 4>>,
   CCIfType<[i64, f64, v1f64, v2f32, v1i64, v2i32, v4i16, v8i8],
            CCAssignToStack<8, 8>>,
@@ -172,12 +169,11 @@ def CC_AArch64_DarwinPCS_VarArg : CallingConv<[
 // 32bit quantity as undef.
 def CC_AArch64_WebKit_JS : CallingConv<[
   // Handle i1, i8, i16, i32, and i64 passing in register X0 (W0).
-  CCIfType<[i1, i8, i16], CCIfUnallocated<"X0", CCPromoteToType<i32>>>,
+  CCIfType<[i1, i8, i16], CCPromoteToType<i32>>,
   CCIfType<[i32], CCAssignToRegWithShadow<[W0], [X0]>>,
   CCIfType<[i64], CCAssignToRegWithShadow<[X0], [W0]>>,
 
   // Pass the remaining arguments on the stack instead.
-  CCIfType<[i1, i8, i16], CCAssignToStack<4, 4>>,
   CCIfType<[i32, f32], CCAssignToStack<4, 4>>,
   CCIfType<[i64, f64], CCAssignToStack<8, 8>>
 ]>;
diff --git a/lib/Target/AArch64/AArch64FastISel.cpp b/lib/Target/AArch64/AArch64FastISel.cpp
index c3b5369..2164d77 100644
--- a/lib/Target/AArch64/AArch64FastISel.cpp
+++ b/lib/Target/AArch64/AArch64FastISel.cpp
@@ -240,21 +240,15 @@ unsigned AArch64FastISel::AArch64MaterializeFP(const ConstantFP *CFP, MVT VT) {
 }
 
 unsigned AArch64FastISel::AArch64MaterializeGV(const GlobalValue *GV) {
-  // We can't handle thread-local variables quickly yet. Unfortunately we have
-  // to peer through any aliases to find out if that rule applies.
-  const GlobalValue *TLSGV = GV;
-  if (const GlobalAlias *GA = dyn_cast<GlobalAlias>(GV))
-    TLSGV = GA->getAliasee();
+  // We can't handle thread-local variables quickly yet.
+  if (GV->isThreadLocal())
+    return 0;
 
   // MachO still uses GOT for large code-model accesses, but ELF requires
   // movz/movk sequences, which FastISel doesn't handle yet.
   if (TM.getCodeModel() != CodeModel::Small && !Subtarget->isTargetMachO())
     return 0;
 
-  if (const GlobalVariable *GVar = dyn_cast<GlobalVariable>(TLSGV))
-    if (GVar->isThreadLocal())
-      return 0;
-
   unsigned char OpFlags = Subtarget->ClassifyGlobalReference(GV, TM);
 
   EVT DestEVT = TLI.getValueType(GV->getType(), true);
@@ -469,11 +463,18 @@ bool AArch64FastISel::SimplifyAddress(Address &Addr, MVT VT,
     break;
   }
 
-  // FIXME: If this is a stack pointer and the offset needs to be simplified
-  // then put the alloca address into a register, set the base type back to
-  // register and continue. This should almost never happen.
+  //If this is a stack pointer and the offset needs to be simplified then put
+  // the alloca address into a register, set the base type back to register and
+  // continue. This should almost never happen.
   if (needsLowering && Addr.getKind() == Address::FrameIndexBase) {
-    return false;
+    unsigned ResultReg = createResultReg(&AArch64::GPR64RegClass);
+    BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::ADDXri),
+            ResultReg)
+        .addFrameIndex(Addr.getFI())
+        .addImm(0)
+        .addImm(0);
+    Addr.setKind(Address::RegBase);
+    Addr.setReg(ResultReg);
   }
 
   // Since the offset is too large for the load/store instruction get the
@@ -1224,7 +1225,6 @@ bool AArch64FastISel::ProcessCallArgs(
       Arg = EmitIntExt(SrcVT, Arg, DestVT, /*isZExt*/ false);
       if (Arg == 0)
         return false;
-      ArgVT = DestVT;
       break;
     }
     case CCValAssign::AExt:
@@ -1235,7 +1235,6 @@ bool AArch64FastISel::ProcessCallArgs(
       Arg = EmitIntExt(SrcVT, Arg, DestVT, /*isZExt*/ true);
       if (Arg == 0)
         return false;
-      ArgVT = DestVT;
       break;
     }
     default:
@@ -1254,7 +1253,7 @@ bool AArch64FastISel::ProcessCallArgs(
       assert(VA.isMemLoc() && "Assuming store on stack.");
 
       // Need to store on the stack.
-      unsigned ArgSize = VA.getLocVT().getSizeInBits() / 8;
+      unsigned ArgSize = (ArgVT.getSizeInBits() + 7) / 8;
 
       unsigned BEAlign = 0;
       if (ArgSize < 8 && !Subtarget->isLittleEndian())
@@ -1468,10 +1467,12 @@ bool AArch64FastISel::TryEmitSmallMemCpy(Address Dest, Address Src,
     bool RV;
     unsigned ResultReg;
     RV = EmitLoad(VT, ResultReg, Src);
-    assert(RV == true && "Should be able to handle this load.");
+    if (!RV)
+      return false;
+
     RV = EmitStore(VT, ResultReg, Dest);
-    assert(RV == true && "Should be able to handle this store.");
-    (void)RV;
+    if (!RV)
+      return false;
 
     int64_t Size = VT.getSizeInBits() / 8;
     Len -= Size;
@@ -1749,6 +1750,17 @@ unsigned AArch64FastISel::Emiti1Ext(unsigned SrcReg, MVT DestVT, bool isZExt) {
 unsigned AArch64FastISel::EmitIntExt(MVT SrcVT, unsigned SrcReg, MVT DestVT,
                                      bool isZExt) {
   assert(DestVT != MVT::i1 && "ZeroExt/SignExt an i1?");
+
+  // FastISel does not have plumbing to deal with extensions where the SrcVT or
+  // DestVT are odd things, so test to make sure that they are both types we can
+  // handle (i1/i8/i16/i32 for SrcVT and i8/i16/i32/i64 for DestVT), otherwise
+  // bail out to SelectionDAG.
+  if (((DestVT != MVT::i8) && (DestVT != MVT::i16) &&
+       (DestVT != MVT::i32) && (DestVT != MVT::i64)) ||
+      ((SrcVT !=  MVT::i1) && (SrcVT !=  MVT::i8) &&
+       (SrcVT !=  MVT::i16) && (SrcVT !=  MVT::i32)))
+    return 0;
+
   unsigned Opc;
   unsigned Imm = 0;
 
@@ -1895,6 +1907,7 @@ bool AArch64FastISel::SelectMul(const Instruction *I) {
   case MVT::i32:
     ZReg = AArch64::WZR;
     Opc = AArch64::MADDWrrr;
+    SrcVT = MVT::i32;
     break;
   case MVT::i64:
     ZReg = AArch64::XZR;
diff --git a/lib/Target/AArch64/AArch64FrameLowering.cpp b/lib/Target/AArch64/AArch64FrameLowering.cpp
index deb306a..9c33717 100644
--- a/lib/Target/AArch64/AArch64FrameLowering.cpp
+++ b/lib/Target/AArch64/AArch64FrameLowering.cpp
@@ -158,7 +158,7 @@ void AArch64FrameLowering::emitCalleeSavedFrameMoves(
   MachineFrameInfo *MFI = MF.getFrameInfo();
   MachineModuleInfo &MMI = MF.getMMI();
   const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo();
-  const AArch64InstrInfo *TII = TM.getInstrInfo();
+  const TargetInstrInfo *TII = MF.getTarget().getInstrInfo();
   DebugLoc DL = MBB.findDebugLoc(MBBI);
 
   // Add callee saved registers to move list.
@@ -204,8 +204,9 @@ void AArch64FrameLowering::emitPrologue(MachineFunction &MF) const {
   MachineBasicBlock::iterator MBBI = MBB.begin();
   const MachineFrameInfo *MFI = MF.getFrameInfo();
   const Function *Fn = MF.getFunction();
-  const AArch64RegisterInfo *RegInfo = TM.getRegisterInfo();
-  const AArch64InstrInfo *TII = TM.getInstrInfo();
+  const AArch64RegisterInfo *RegInfo = static_cast<const AArch64RegisterInfo *>(
+      MF.getTarget().getRegisterInfo());
+  const TargetInstrInfo *TII = MF.getTarget().getInstrInfo();
   MachineModuleInfo &MMI = MF.getMMI();
   AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();
   bool needsFrameMoves = MMI.hasDebugInfo() || Fn->needsUnwindTableEntry();
diff --git a/lib/Target/AArch64/AArch64FrameLowering.h b/lib/Target/AArch64/AArch64FrameLowering.h
index 0e00d16..7686e6f 100644
--- a/lib/Target/AArch64/AArch64FrameLowering.h
+++ b/lib/Target/AArch64/AArch64FrameLowering.h
@@ -18,18 +18,11 @@
 
 namespace llvm {
 
-class AArch64Subtarget;
-class AArch64TargetMachine;
-
 class AArch64FrameLowering : public TargetFrameLowering {
-  const AArch64TargetMachine &TM;
-
 public:
-  explicit AArch64FrameLowering(const AArch64TargetMachine &TM,
-                              const AArch64Subtarget &STI)
+  explicit AArch64FrameLowering()
       : TargetFrameLowering(StackGrowsDown, 16, 0, 16,
-                            false /*StackRealignable*/),
-        TM(TM) {}
+                            false /*StackRealignable*/) {}
 
   void emitCalleeSavedFrameMoves(MachineBasicBlock &MBB,
                                  MachineBasicBlock::iterator MBBI,
diff --git a/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp b/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
index 7007ffc..3f49fab 100644
--- a/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
+++ b/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
@@ -153,9 +153,6 @@ public:
   SDNode *SelectStoreLane(SDNode *N, unsigned NumVecs, unsigned Opc);
   SDNode *SelectPostStoreLane(SDNode *N, unsigned NumVecs, unsigned Opc);
 
-  SDNode *SelectSIMDAddSubNarrowing(unsigned IntNo, SDNode *Node);
-  SDNode *SelectSIMDXtnNarrowing(unsigned IntNo, SDNode *Node);
-
   SDNode *SelectBitfieldExtractOp(SDNode *N);
   SDNode *SelectBitfieldInsertOp(SDNode *N);
 
@@ -596,8 +593,9 @@ bool AArch64DAGToDAGISel::SelectAddrModeIndexed(SDValue N, unsigned Size,
     const GlobalValue *GV = GAN->getGlobal();
     unsigned Alignment = GV->getAlignment();
     const DataLayout *DL = TLI->getDataLayout();
-    if (Alignment == 0 && !Subtarget->isTargetDarwin())
-      Alignment = DL->getABITypeAlignment(GV->getType()->getElementType());
+    Type *Ty = GV->getType()->getElementType();
+    if (Alignment == 0 && Ty->isSized() && !Subtarget->isTargetDarwin())
+      Alignment = DL->getABITypeAlignment(Ty);
 
     if (Alignment >= Size)
       return true;
@@ -2111,7 +2109,7 @@ SDNode *AArch64DAGToDAGISel::Select(SDNode *Node) {
                 .getVectorElementType()
                 .getSizeInBits()) {
     default:
-      assert(0 && "Unexpected vector element type!");
+      llvm_unreachable("Unexpected vector element type!");
     case 64:
       SubReg = AArch64::dsub;
       break;
diff --git a/lib/Target/AArch64/AArch64ISelLowering.cpp b/lib/Target/AArch64/AArch64ISelLowering.cpp
index 80d6669..28d0035 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -67,15 +67,15 @@ EnableAArch64SlrGeneration("aarch64-shift-insert-generation", cl::Hidden,
 //===----------------------------------------------------------------------===//
 // AArch64 Lowering public interface.
 //===----------------------------------------------------------------------===//
-static TargetLoweringObjectFile *createTLOF(TargetMachine &TM) {
-  if (TM.getSubtarget<AArch64Subtarget>().isTargetDarwin())
+static TargetLoweringObjectFile *createTLOF(const Triple &TT) {
+  if (TT.isOSBinFormatMachO())
     return new AArch64_MachoTargetObjectFile();
 
   return new AArch64_ELFTargetObjectFile();
 }
 
-AArch64TargetLowering::AArch64TargetLowering(AArch64TargetMachine &TM)
-    : TargetLowering(TM, createTLOF(TM)) {
+AArch64TargetLowering::AArch64TargetLowering(TargetMachine &TM)
+    : TargetLowering(TM, createTLOF(Triple(TM.getTargetTriple()))) {
   Subtarget = &TM.getSubtarget<AArch64Subtarget>();
 
   // AArch64 doesn't have comparisons which set GPRs or setcc instructions, so
@@ -627,7 +627,7 @@ MVT AArch64TargetLowering::getScalarShiftAmountTy(EVT LHSTy) const {
 
 unsigned AArch64TargetLowering::getMaximalGlobalOffset() const {
   // FIXME: On AArch64, this depends on the type.
-  // Basically, the addressable offsets are o to 4095 * Ty.getSizeInBytes().
+  // Basically, the addressable offsets are up to 4095 * Ty.getSizeInBytes().
   // and the offset has to be a multiple of the related size in bytes.
   return 4095;
 }
@@ -823,8 +823,7 @@ AArch64TargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
 #ifndef NDEBUG
     MI->dump();
 #endif
-    assert(0 && "Unexpected instruction for custom inserter!");
-    break;
+    llvm_unreachable("Unexpected instruction for custom inserter!");
 
   case AArch64::F128CSEL:
     return EmitF128CSEL(MI, BB);
@@ -833,7 +832,6 @@ AArch64TargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
   case TargetOpcode::PATCHPOINT:
     return emitPatchPoint(MI, BB);
   }
-  llvm_unreachable("Unexpected instruction for custom inserter!");
 }
 
 //===----------------------------------------------------------------------===//
@@ -1273,7 +1271,7 @@ static SDValue LowerADDC_ADDE_SUBC_SUBE(SDValue Op, SelectionDAG &DAG) {
   bool ExtraOp = false;
   switch (Op.getOpcode()) {
   default:
-    assert(0 && "Invalid code");
+    llvm_unreachable("Invalid code");
   case ISD::ADDC:
     Opc = AArch64ISD::ADDS;
     break;
@@ -1387,24 +1385,22 @@ static SDValue LowerVectorFP_TO_INT(SDValue Op, SelectionDAG &DAG) {
   EVT InVT = Op.getOperand(0).getValueType();
   EVT VT = Op.getValueType();
 
-  // FP_TO_XINT conversion from the same type are legal.
-  if (VT.getSizeInBits() == InVT.getSizeInBits())
-    return Op;
-
-  if (InVT == MVT::v2f64 || InVT == MVT::v4f32) {
+  if (VT.getSizeInBits() < InVT.getSizeInBits()) {
     SDLoc dl(Op);
     SDValue Cv =
         DAG.getNode(Op.getOpcode(), dl, InVT.changeVectorElementTypeToInteger(),
                     Op.getOperand(0));
     return DAG.getNode(ISD::TRUNCATE, dl, VT, Cv);
-  } else if (InVT == MVT::v2f32) {
+  }
+
+  if (VT.getSizeInBits() > InVT.getSizeInBits()) {
     SDLoc dl(Op);
     SDValue Ext = DAG.getNode(ISD::FP_EXTEND, dl, MVT::v2f64, Op.getOperand(0));
     return DAG.getNode(Op.getOpcode(), dl, VT, Ext);
   }
 
   // Type changing conversions are illegal.
-  return SDValue();
+  return Op;
 }
 
 SDValue AArch64TargetLowering::LowerFP_TO_INT(SDValue Op,
@@ -1440,32 +1436,23 @@ static SDValue LowerVectorINT_TO_FP(SDValue Op, SelectionDAG &DAG) {
   SDValue In = Op.getOperand(0);
   EVT InVT = In.getValueType();
 
-  // v2i32 to v2f32 is legal.
-  if (VT == MVT::v2f32 && InVT == MVT::v2i32)
-    return Op;
-
-  // This function only handles v2f64 outputs.
-  if (VT == MVT::v2f64) {
-    // Extend the input argument to a v2i64 that we can feed into the
-    // floating point conversion. Zero or sign extend based on whether
-    // we're doing a signed or unsigned float conversion.
-    unsigned Opc =
-        Op.getOpcode() == ISD::UINT_TO_FP ? ISD::ZERO_EXTEND : ISD::SIGN_EXTEND;
-    assert(Op.getNumOperands() == 1 && "FP conversions take one argument");
-    SDValue Promoted = DAG.getNode(Opc, dl, MVT::v2i64, Op.getOperand(0));
-    return DAG.getNode(Op.getOpcode(), dl, Op.getValueType(), Promoted);
+  if (VT.getSizeInBits() < InVT.getSizeInBits()) {
+    MVT CastVT =
+        MVT::getVectorVT(MVT::getFloatingPointVT(InVT.getScalarSizeInBits()),
+                         InVT.getVectorNumElements());
+    In = DAG.getNode(Op.getOpcode(), dl, CastVT, In);
+    return DAG.getNode(ISD::FP_ROUND, dl, VT, In, DAG.getIntPtrConstant(0));
   }
 
-  // Scalarize v2i64 to v2f32 conversions.
-  std::vector<SDValue> BuildVectorOps;
-  for (unsigned i = 0; i < VT.getVectorNumElements(); ++i) {
-    SDValue Sclr = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i64, In,
-                               DAG.getConstant(i, MVT::i64));
-    Sclr = DAG.getNode(Op->getOpcode(), dl, MVT::f32, Sclr);
-    BuildVectorOps.push_back(Sclr);
+  if (VT.getSizeInBits() > InVT.getSizeInBits()) {
+    unsigned CastOpc =
+        Op.getOpcode() == ISD::SINT_TO_FP ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
+    EVT CastVT = VT.changeVectorElementTypeToInteger();
+    In = DAG.getNode(CastOpc, dl, CastVT, In);
+    return DAG.getNode(Op.getOpcode(), dl, VT, In);
   }
 
-  return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, BuildVectorOps);
+  return Op;
 }
 
 SDValue AArch64TargetLowering::LowerINT_TO_FP(SDValue Op,
@@ -1516,7 +1503,7 @@ SDValue AArch64TargetLowering::LowerFSINCOS(SDValue Op,
   StructType *RetTy = StructType::get(ArgTy, ArgTy, NULL);
   TargetLowering::CallLoweringInfo CLI(DAG);
   CLI.setDebugLoc(dl).setChain(DAG.getEntryNode())
-    .setCallee(CallingConv::Fast, RetTy, Callee, &Args, 0);
+    .setCallee(CallingConv::Fast, RetTy, Callee, std::move(Args), 0);
 
   std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
   return CallResult.first;
@@ -1711,7 +1698,9 @@ SDValue AArch64TargetLowering::LowerFormalArguments(
       InVals.push_back(FrameIdxN);
 
       continue;
-    } if (VA.isRegLoc()) {
+    }
+    
+    if (VA.isRegLoc()) {
       // Arguments stored in registers.
       EVT RegVT = VA.getLocVT();
 
@@ -1772,10 +1761,16 @@ SDValue AArch64TargetLowering::LowerFormalArguments(
       SDValue FIN = DAG.getFrameIndex(FI, getPointerTy());
       SDValue ArgValue;
 
+      // For NON_EXTLOAD, generic code in getLoad assert(ValVT == MemVT)
       ISD::LoadExtType ExtType = ISD::NON_EXTLOAD;
+      MVT MemVT = VA.getValVT();
+
       switch (VA.getLocInfo()) {
       default:
         break;
+      case CCValAssign::BCvt:
+        MemVT = VA.getLocVT();
+        break;
       case CCValAssign::SExt:
         ExtType = ISD::SEXTLOAD;
         break;
@@ -1787,10 +1782,9 @@ SDValue AArch64TargetLowering::LowerFormalArguments(
         break;
       }
 
-      ArgValue = DAG.getExtLoad(ExtType, DL, VA.getValVT(), Chain, FIN,
+      ArgValue = DAG.getExtLoad(ExtType, DL, VA.getLocVT(), Chain, FIN,
                                 MachinePointerInfo::getFixedStack(FI),
-                                VA.getLocVT(),
-                                false, false, false, 0);
+                                MemVT, false, false, false, nullptr);
 
       InVals.push_back(ArgValue);
     }
@@ -2339,11 +2333,9 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
         // Since we pass i1/i8/i16 as i1/i8/i16 on stack and Arg is already
         // promoted to a legal register type i32, we should truncate Arg back to
         // i1/i8/i16.
-        if (Arg.getValueType().isSimple() &&
-            Arg.getValueType().getSimpleVT() == MVT::i32 &&
-            (VA.getLocVT() == MVT::i1 || VA.getLocVT() == MVT::i8 ||
-             VA.getLocVT() == MVT::i16))
-          Arg = DAG.getNode(ISD::TRUNCATE, DL, VA.getLocVT(), Arg);
+        if (VA.getValVT() == MVT::i1 || VA.getValVT() == MVT::i8 ||
+            VA.getValVT() == MVT::i16)
+          Arg = DAG.getNode(ISD::TRUNCATE, DL, VA.getValVT(), Arg);
 
         SDValue Store =
             DAG.getStore(Chain, DL, Arg, DstAddr, DstInfo, false, false, 0);
@@ -4116,6 +4108,7 @@ static SDValue NarrowVector(SDValue V128Reg, SelectionDAG &DAG) {
 // shuffle in combination with VEXTs.
 SDValue AArch64TargetLowering::ReconstructShuffle(SDValue Op,
                                                   SelectionDAG &DAG) const {
+  assert(Op.getOpcode() == ISD::BUILD_VECTOR && "Unknown opcode!");
   SDLoc dl(Op);
   EVT VT = Op.getValueType();
   unsigned NumElts = VT.getVectorNumElements();
@@ -4164,35 +4157,47 @@ SDValue AArch64TargetLowering::ReconstructShuffle(SDValue Op,
 
   SDValue ShuffleSrcs[2] = { DAG.getUNDEF(VT), DAG.getUNDEF(VT) };
   int VEXTOffsets[2] = { 0, 0 };
+  int OffsetMultipliers[2] = { 1, 1 };
 
   // This loop extracts the usage patterns of the source vectors
   // and prepares appropriate SDValues for a shuffle if possible.
   for (unsigned i = 0; i < SourceVecs.size(); ++i) {
-    if (SourceVecs[i].getValueType() == VT) {
+    unsigned NumSrcElts = SourceVecs[i].getValueType().getVectorNumElements();
+    SDValue CurSource = SourceVecs[i];
+    if (SourceVecs[i].getValueType().getVectorElementType() !=
+        VT.getVectorElementType()) {
+      // It may hit this case if SourceVecs[i] is AssertSext/AssertZext.
+      // Then bitcast it to the vector which holds asserted element type,
+      // and record the multiplier of element width between SourceVecs and
+      // Build_vector which is needed to extract the correct lanes later.
+      EVT CastVT =
+          EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
+                           SourceVecs[i].getValueSizeInBits() /
+                               VT.getVectorElementType().getSizeInBits());
+
+      CurSource = DAG.getNode(ISD::BITCAST, dl, CastVT, SourceVecs[i]);
+      OffsetMultipliers[i] = CastVT.getVectorNumElements() / NumSrcElts;
+      NumSrcElts *= OffsetMultipliers[i];
+      MaxElts[i] *= OffsetMultipliers[i];
+      MinElts[i] *= OffsetMultipliers[i];
+    }
+
+    if (CurSource.getValueType() == VT) {
       // No VEXT necessary
-      ShuffleSrcs[i] = SourceVecs[i];
+      ShuffleSrcs[i] = CurSource;
       VEXTOffsets[i] = 0;
       continue;
-    } else if (SourceVecs[i].getValueType().getVectorNumElements() < NumElts) {
+    } else if (NumSrcElts < NumElts) {
       // We can pad out the smaller vector for free, so if it's part of a
       // shuffle...
-      ShuffleSrcs[i] = DAG.getNode(ISD::CONCAT_VECTORS, dl, VT, SourceVecs[i],
-                                   DAG.getUNDEF(SourceVecs[i].getValueType()));
+      ShuffleSrcs[i] = DAG.getNode(ISD::CONCAT_VECTORS, dl, VT, CurSource,
+                                   DAG.getUNDEF(CurSource.getValueType()));
       continue;
     }
 
-    // Don't attempt to extract subvectors from BUILD_VECTOR sources
-    // that expand or trunc the original value.
-    // TODO: We can try to bitcast and ANY_EXTEND the result but
-    // we need to consider the cost of vector ANY_EXTEND, and the
-    // legality of all the types.
-    if (SourceVecs[i].getValueType().getVectorElementType() !=
-        VT.getVectorElementType())
-      return SDValue();
-
     // Since only 64-bit and 128-bit vectors are legal on ARM and
     // we've eliminated the other cases...
-    assert(SourceVecs[i].getValueType().getVectorNumElements() == 2 * NumElts &&
+    assert(NumSrcElts == 2 * NumElts &&
            "unexpected vector sizes in ReconstructShuffle");
 
     if (MaxElts[i] - MinElts[i] >= NumElts) {
@@ -4203,22 +4208,20 @@ SDValue AArch64TargetLowering::ReconstructShuffle(SDValue Op,
     if (MinElts[i] >= NumElts) {
       // The extraction can just take the second half
       VEXTOffsets[i] = NumElts;
-      ShuffleSrcs[i] =
-          DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, SourceVecs[i],
-                      DAG.getIntPtrConstant(NumElts));
+      ShuffleSrcs[i] = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, CurSource,
+                                   DAG.getIntPtrConstant(NumElts));
     } else if (MaxElts[i] < NumElts) {
       // The extraction can just take the first half
       VEXTOffsets[i] = 0;
-      ShuffleSrcs[i] = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT,
-                                   SourceVecs[i], DAG.getIntPtrConstant(0));
+      ShuffleSrcs[i] = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, CurSource,
+                                   DAG.getIntPtrConstant(0));
     } else {
       // An actual VEXT is needed
       VEXTOffsets[i] = MinElts[i];
-      SDValue VEXTSrc1 = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT,
-                                     SourceVecs[i], DAG.getIntPtrConstant(0));
-      SDValue VEXTSrc2 =
-          DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, SourceVecs[i],
-                      DAG.getIntPtrConstant(NumElts));
+      SDValue VEXTSrc1 = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, CurSource,
+                                     DAG.getIntPtrConstant(0));
+      SDValue VEXTSrc2 = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, VT, CurSource,
+                                     DAG.getIntPtrConstant(NumElts));
       unsigned Imm = VEXTOffsets[i] * getExtFactor(VEXTSrc1);
       ShuffleSrcs[i] = DAG.getNode(AArch64ISD::EXT, dl, VT, VEXTSrc1, VEXTSrc2,
                                    DAG.getConstant(Imm, MVT::i32));
@@ -4238,9 +4241,10 @@ SDValue AArch64TargetLowering::ReconstructShuffle(SDValue Op,
     int ExtractElt =
         cast<ConstantSDNode>(Op.getOperand(i).getOperand(1))->getSExtValue();
     if (ExtractVec == SourceVecs[0]) {
-      Mask.push_back(ExtractElt - VEXTOffsets[0]);
+      Mask.push_back(ExtractElt * OffsetMultipliers[0] - VEXTOffsets[0]);
     } else {
-      Mask.push_back(ExtractElt + NumElts - VEXTOffsets[1]);
+      Mask.push_back(ExtractElt * OffsetMultipliers[1] + NumElts -
+                     VEXTOffsets[1]);
     }
   }
 
@@ -5177,11 +5181,37 @@ FailedModImm:
   return Op;
 }
 
+// Normalize the operands of BUILD_VECTOR. The value of constant operands will
+// be truncated to fit element width.
+static SDValue NormalizeBuildVector(SDValue Op,
+                                    SelectionDAG &DAG) {
+  assert(Op.getOpcode() == ISD::BUILD_VECTOR && "Unknown opcode!");
+  SDLoc dl(Op);
+  EVT VT = Op.getValueType();
+  EVT EltTy= VT.getVectorElementType();
+
+  if (EltTy.isFloatingPoint() || EltTy.getSizeInBits() > 16)
+    return Op;
+
+  SmallVector<SDValue, 16> Ops;
+  for (unsigned I = 0, E = VT.getVectorNumElements(); I != E; ++I) {
+    SDValue Lane = Op.getOperand(I);
+    if (Lane.getOpcode() == ISD::Constant) {
+      APInt LowBits(EltTy.getSizeInBits(),
+                    cast<ConstantSDNode>(Lane)->getZExtValue());
+      Lane = DAG.getConstant(LowBits.getZExtValue(), MVT::i32);
+    }
+    Ops.push_back(Lane);
+  }
+  return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, Ops);
+}
+
 SDValue AArch64TargetLowering::LowerBUILD_VECTOR(SDValue Op,
                                                  SelectionDAG &DAG) const {
-  BuildVectorSDNode *BVN = cast<BuildVectorSDNode>(Op.getNode());
   SDLoc dl(Op);
   EVT VT = Op.getValueType();
+  Op = NormalizeBuildVector(Op, DAG);
+  BuildVectorSDNode *BVN = cast<BuildVectorSDNode>(Op.getNode());
 
   APInt CnstBits(VT.getSizeInBits(), 0);
   APInt UndefBits(VT.getSizeInBits(), 0);
@@ -6047,18 +6077,14 @@ bool AArch64TargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const {
     return false;
   unsigned NumBits1 = Ty1->getPrimitiveSizeInBits();
   unsigned NumBits2 = Ty2->getPrimitiveSizeInBits();
-  if (NumBits1 <= NumBits2)
-    return false;
-  return true;
+  return NumBits1 > NumBits2;
 }
 bool AArch64TargetLowering::isTruncateFree(EVT VT1, EVT VT2) const {
-  if (!VT1.isInteger() || !VT2.isInteger())
+  if (VT1.isVector() || VT2.isVector() || !VT1.isInteger() || !VT2.isInteger())
     return false;
   unsigned NumBits1 = VT1.getSizeInBits();
   unsigned NumBits2 = VT2.getSizeInBits();
-  if (NumBits1 <= NumBits2)
-    return false;
-  return true;
+  return NumBits1 > NumBits2;
 }
 
 // All 32-bit GPR operations implicitly zero the high-half of the corresponding
@@ -6068,18 +6094,14 @@ bool AArch64TargetLowering::isZExtFree(Type *Ty1, Type *Ty2) const {
     return false;
   unsigned NumBits1 = Ty1->getPrimitiveSizeInBits();
   unsigned NumBits2 = Ty2->getPrimitiveSizeInBits();
-  if (NumBits1 == 32 && NumBits2 == 64)
-    return true;
-  return false;
+  return NumBits1 == 32 && NumBits2 == 64;
 }
 bool AArch64TargetLowering::isZExtFree(EVT VT1, EVT VT2) const {
-  if (!VT1.isInteger() || !VT2.isInteger())
+  if (VT1.isVector() || VT2.isVector() || !VT1.isInteger() || !VT2.isInteger())
     return false;
   unsigned NumBits1 = VT1.getSizeInBits();
   unsigned NumBits2 = VT2.getSizeInBits();
-  if (NumBits1 == 32 && NumBits2 == 64)
-    return true;
-  return false;
+  return NumBits1 == 32 && NumBits2 == 64;
 }
 
 bool AArch64TargetLowering::isZExtFree(SDValue Val, EVT VT2) const {
@@ -6092,8 +6114,9 @@ bool AArch64TargetLowering::isZExtFree(SDValue Val, EVT VT2) const {
     return false;
 
   // 8-, 16-, and 32-bit integer loads all implicitly zero-extend.
-  return (VT1.isSimple() && VT1.isInteger() && VT2.isSimple() &&
-          VT2.isInteger() && VT1.getSizeInBits() <= 32);
+  return (VT1.isSimple() && !VT1.isVector() && VT1.isInteger() &&
+          VT2.isSimple() && !VT2.isVector() && VT2.isInteger() &&
+          VT1.getSizeInBits() <= 32);
 }
 
 bool AArch64TargetLowering::hasPairedLoad(Type *LoadedType,
@@ -6346,23 +6369,45 @@ static SDValue performMulCombine(SDNode *N, SelectionDAG &DAG,
   if (ConstantSDNode *C = dyn_cast<ConstantSDNode>(N->getOperand(1))) {
     APInt Value = C->getAPIntValue();
     EVT VT = N->getValueType(0);
-    APInt VP1 = Value + 1;
-    if (VP1.isPowerOf2()) {
-      // Multiplying by one less than a power of two, replace with a shift
-      // and a subtract.
-      SDValue ShiftedVal =
-          DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
-                      DAG.getConstant(VP1.logBase2(), MVT::i64));
-      return DAG.getNode(ISD::SUB, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
-    }
-    APInt VM1 = Value - 1;
-    if (VM1.isPowerOf2()) {
-      // Multiplying by one more than a power of two, replace with a shift
-      // and an add.
-      SDValue ShiftedVal =
-          DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
-                      DAG.getConstant(VM1.logBase2(), MVT::i64));
-      return DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
+    if (Value.isNonNegative()) {
+      // (mul x, 2^N + 1) => (add (shl x, N), x)
+      APInt VM1 = Value - 1;
+      if (VM1.isPowerOf2()) {
+        SDValue ShiftedVal =
+            DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+                        DAG.getConstant(VM1.logBase2(), MVT::i64));
+        return DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal,
+                           N->getOperand(0));
+      }
+      // (mul x, 2^N - 1) => (sub (shl x, N), x)
+      APInt VP1 = Value + 1;
+      if (VP1.isPowerOf2()) {
+        SDValue ShiftedVal =
+            DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+                        DAG.getConstant(VP1.logBase2(), MVT::i64));
+        return DAG.getNode(ISD::SUB, SDLoc(N), VT, ShiftedVal,
+                           N->getOperand(0));
+      }
+    } else {
+      // (mul x, -(2^N + 1)) => - (add (shl x, N), x)
+      APInt VNM1 = -Value - 1;
+      if (VNM1.isPowerOf2()) {
+        SDValue ShiftedVal =
+            DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+                        DAG.getConstant(VNM1.logBase2(), MVT::i64));
+        SDValue Add =
+            DAG.getNode(ISD::ADD, SDLoc(N), VT, ShiftedVal, N->getOperand(0));
+        return DAG.getNode(ISD::SUB, SDLoc(N), VT, DAG.getConstant(0, VT), Add);
+      }
+      // (mul x, -(2^N - 1)) => (sub x, (shl x, N))
+      APInt VNP1 = -Value + 1;
+      if (VNP1.isPowerOf2()) {
+        SDValue ShiftedVal =
+            DAG.getNode(ISD::SHL, SDLoc(N), VT, N->getOperand(0),
+                        DAG.getConstant(VNP1.logBase2(), MVT::i64));
+        return DAG.getNode(ISD::SUB, SDLoc(N), VT, N->getOperand(0),
+                           ShiftedVal);
+      }
     }
   }
   return SDValue();
@@ -6687,7 +6732,7 @@ static SDValue tryCombineFixedPointConvert(SDNode *N,
     else if (Vec.getValueType() == MVT::v2i64)
       VecResTy = MVT::v2f64;
     else
-      assert(0 && "unexpected vector type!");
+      llvm_unreachable("unexpected vector type!");
 
     SDValue Convert =
         DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, VecResTy, IID, Vec, Shift);
@@ -7020,7 +7065,7 @@ static SDValue tryCombineShiftImm(unsigned IID, SDNode *N, SelectionDAG &DAG) {
   if (IsRightShift && ShiftAmount <= -1 && ShiftAmount >= -(int)ElemBits)
     return DAG.getNode(Opcode, SDLoc(N), N->getValueType(0), N->getOperand(1),
                        DAG.getConstant(-ShiftAmount, MVT::i32));
-  else if (!IsRightShift && ShiftAmount >= 0 && ShiftAmount <= ElemBits)
+  else if (!IsRightShift && ShiftAmount >= 0 && ShiftAmount < ElemBits)
     return DAG.getNode(Opcode, SDLoc(N), N->getValueType(0), N->getOperand(1),
                        DAG.getConstant(ShiftAmount, MVT::i32));
 
@@ -7867,6 +7912,18 @@ bool AArch64TargetLowering::shouldExpandAtomicInIR(Instruction *Inst) const {
   return Inst->getType()->getPrimitiveSizeInBits() <= 128;
 }
 
+TargetLoweringBase::LegalizeTypeAction
+AArch64TargetLowering::getPreferredVectorAction(EVT VT) const {
+  MVT SVT = VT.getSimpleVT();
+  // During type legalization, we prefer to widen v1i8, v1i16, v1i32  to v8i8,
+  // v4i16, v2i32 instead of to promote.
+  if (SVT == MVT::v1i8 || SVT == MVT::v1i16 || SVT == MVT::v1i32
+      || SVT == MVT::v1f32)
+    return TypeWidenVector;
+
+  return TargetLoweringBase::getPreferredVectorAction(VT);
+}
+
 Value *AArch64TargetLowering::emitLoadLinked(IRBuilder<> &Builder, Value *Addr,
                                              AtomicOrdering Ord) const {
   Module *M = Builder.GetInsertBlock()->getParent()->getParent();
diff --git a/lib/Target/AArch64/AArch64ISelLowering.h b/lib/Target/AArch64/AArch64ISelLowering.h
index de16c4d..cb0b9ef 100644
--- a/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/lib/Target/AArch64/AArch64ISelLowering.h
@@ -197,7 +197,7 @@ class AArch64TargetLowering : public TargetLowering {
   bool RequireStrictAlign;
 
 public:
-  explicit AArch64TargetLowering(AArch64TargetMachine &TM);
+  explicit AArch64TargetLowering(TargetMachine &TM);
 
   /// Selects the correct CCAssignFn for a the given CallingConvention
   /// value.
@@ -324,6 +324,9 @@ public:
 
   bool shouldExpandAtomicInIR(Instruction *Inst) const override;
 
+  TargetLoweringBase::LegalizeTypeAction
+  getPreferredVectorAction(EVT VT) const override;
+
 private:
   /// Subtarget - Keep a pointer to the AArch64Subtarget around so that we can
   /// make the right decision when generating code for different targets.
diff --git a/lib/Target/AArch64/AArch64InstrFormats.td b/lib/Target/AArch64/AArch64InstrFormats.td
index d455d7e..5007172 100644
--- a/lib/Target/AArch64/AArch64InstrFormats.td
+++ b/lib/Target/AArch64/AArch64InstrFormats.td
@@ -448,13 +448,19 @@ def logical_imm64_XFORM : SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant(enc, MVT::i32);
 }]>;
 
-def LogicalImm32Operand : AsmOperandClass {
-  let Name = "LogicalImm32";
-  let DiagnosticType = "LogicalSecondSource";
-}
-def LogicalImm64Operand : AsmOperandClass {
-  let Name = "LogicalImm64";
-  let DiagnosticType = "LogicalSecondSource";
+let DiagnosticType = "LogicalSecondSource" in {
+  def LogicalImm32Operand : AsmOperandClass {
+    let Name = "LogicalImm32";
+  }
+  def LogicalImm64Operand : AsmOperandClass {
+    let Name = "LogicalImm64";
+  }
+  def LogicalImm32NotOperand : AsmOperandClass {
+    let Name = "LogicalImm32Not";
+  }
+  def LogicalImm64NotOperand : AsmOperandClass {
+    let Name = "LogicalImm64Not";
+  }
 }
 def logical_imm32 : Operand<i32>, PatLeaf<(imm), [{
   return AArch64_AM::isLogicalImmediate(N->getZExtValue(), 32);
@@ -468,6 +474,12 @@ def logical_imm64 : Operand<i64>, PatLeaf<(imm), [{
   let PrintMethod = "printLogicalImm64";
   let ParserMatchClass = LogicalImm64Operand;
 }
+def logical_imm32_not : Operand<i32> {
+  let ParserMatchClass = LogicalImm32NotOperand;
+}
+def logical_imm64_not : Operand<i64> {
+  let ParserMatchClass = LogicalImm64NotOperand;
+}
 
 // imm0_65535 predicate - True if the immediate is in the range [0,65535].
 def Imm0_65535Operand : AsmImmRange<0, 65535>;
@@ -963,8 +975,14 @@ def ccode : Operand<i32> {
   let ParserMatchClass = CondCode;
 }
 def inv_ccode : Operand<i32> {
+  // AL and NV are invalid in the aliases which use inv_ccode
   let PrintMethod = "printInverseCondCode";
   let ParserMatchClass = CondCode;
+  let MCOperandPredicate = [{
+    return MCOp.isImm() &&
+           MCOp.getImm() != AArch64CC::AL &&
+           MCOp.getImm() != AArch64CC::NV;
+  }];
 }
 
 // Conditional branch target. 19-bit immediate. The low two bits of the target
@@ -1323,13 +1341,13 @@ class BaseMulAccum<bit isSub, bits<3> opc, RegisterClass multype,
 multiclass MulAccum<bit isSub, string asm, SDNode AccNode> {
   def Wrrr : BaseMulAccum<isSub, 0b000, GPR32, GPR32, asm,
       [(set GPR32:$Rd, (AccNode GPR32:$Ra, (mul GPR32:$Rn, GPR32:$Rm)))]>,
-      Sched<[WriteIM32, ReadIMA, ReadIM, ReadIM]> {
+      Sched<[WriteIM32, ReadIM, ReadIM, ReadIMA]> {
     let Inst{31} = 0;
   }
 
   def Xrrr : BaseMulAccum<isSub, 0b000, GPR64, GPR64, asm,
       [(set GPR64:$Rd, (AccNode GPR64:$Ra, (mul GPR64:$Rn, GPR64:$Rm)))]>,
-      Sched<[WriteIM64, ReadIMA, ReadIM, ReadIM]> {
+      Sched<[WriteIM64, ReadIM, ReadIM, ReadIMA]> {
     let Inst{31} = 1;
   }
 }
@@ -1339,7 +1357,7 @@ class WideMulAccum<bit isSub, bits<3> opc, string asm,
   : BaseMulAccum<isSub, opc, GPR32, GPR64, asm,
     [(set GPR64:$Rd, (AccNode GPR64:$Ra,
                             (mul (ExtNode GPR32:$Rn), (ExtNode GPR32:$Rm))))]>,
-    Sched<[WriteIM32, ReadIMA, ReadIM, ReadIM]> {
+    Sched<[WriteIM32, ReadIM, ReadIM, ReadIMA]> {
   let Inst{31} = 1;
 }
 
@@ -1738,6 +1756,10 @@ multiclass AddSubS<bit isSub, string mnemonic, SDNode OpNode, string cmp> {
                   WZR, GPR32:$src1, GPR32:$src2, 0), 5>;
   def : InstAlias<cmp#" $src1, $src2", (!cast<Instruction>(NAME#"Xrs")
                   XZR, GPR64:$src1, GPR64:$src2, 0), 5>;
+  def : InstAlias<cmp#" $src1, $src2", (!cast<Instruction>(NAME#"Wrx")
+                  WZR, GPR32sponly:$src1, GPR32:$src2, 16), 5>;
+  def : InstAlias<cmp#" $src1, $src2", (!cast<Instruction>(NAME#"Xrx64")
+                  XZR, GPR64sponly:$src1, GPR64:$src2, 24), 5>;
 
   // Register/register aliases with no shift when SP is not used.
   def : AddSubRegAlias<mnemonic, !cast<Instruction>(NAME#"Wrs"),
@@ -1925,22 +1947,32 @@ class LogicalRegAlias<string asm, Instruction inst, RegisterClass regtype>
     : InstAlias<asm#" $dst, $src1, $src2",
                 (inst regtype:$dst, regtype:$src1, regtype:$src2, 0)>;
 
-let AddedComplexity = 6 in
-multiclass LogicalImm<bits<2> opc, string mnemonic, SDNode OpNode> {
+multiclass LogicalImm<bits<2> opc, string mnemonic, SDNode OpNode,
+                      string Alias> {
+  let AddedComplexity = 6 in
   def Wri : BaseLogicalImm<opc, GPR32sp, GPR32, logical_imm32, mnemonic,
                            [(set GPR32sp:$Rd, (OpNode GPR32:$Rn,
                                                logical_imm32:$imm))]> {
     let Inst{31} = 0;
     let Inst{22} = 0; // 64-bit version has an additional bit of immediate.
   }
+  let AddedComplexity = 6 in
   def Xri : BaseLogicalImm<opc, GPR64sp, GPR64, logical_imm64, mnemonic,
                            [(set GPR64sp:$Rd, (OpNode GPR64:$Rn,
                                                logical_imm64:$imm))]> {
     let Inst{31} = 1;
   }
+
+  def : InstAlias<Alias # " $Rd, $Rn, $imm",
+                  (!cast<Instruction>(NAME # "Wri") GPR32sp:$Rd, GPR32:$Rn,
+                      logical_imm32_not:$imm), 0>;
+  def : InstAlias<Alias # " $Rd, $Rn, $imm",
+                  (!cast<Instruction>(NAME # "Xri") GPR64sp:$Rd, GPR64:$Rn,
+                       logical_imm64_not:$imm), 0>;
 }
 
-multiclass LogicalImmS<bits<2> opc, string mnemonic, SDNode OpNode> {
+multiclass LogicalImmS<bits<2> opc, string mnemonic, SDNode OpNode,
+                       string Alias> {
   let isCompare = 1, Defs = [NZCV] in {
   def Wri  : BaseLogicalImm<opc, GPR32, GPR32, logical_imm32, mnemonic,
       [(set GPR32:$Rd, (OpNode GPR32:$Rn, logical_imm32:$imm))]> {
@@ -1952,6 +1984,13 @@ multiclass LogicalImmS<bits<2> opc, string mnemonic, SDNode OpNode> {
     let Inst{31} = 1;
   }
   } // end Defs = [NZCV]
+
+  def : InstAlias<Alias # " $Rd, $Rn, $imm",
+                  (!cast<Instruction>(NAME # "Wri") GPR32:$Rd, GPR32:$Rn,
+                      logical_imm32_not:$imm), 0>;
+  def : InstAlias<Alias # " $Rd, $Rn, $imm",
+                  (!cast<Instruction>(NAME # "Xri") GPR64:$Rd, GPR64:$Rn,
+                       logical_imm64_not:$imm), 0>;
 }
 
 class BaseLogicalRegPseudo<RegisterClass regtype, SDPatternOperator OpNode>
diff --git a/lib/Target/AArch64/AArch64InstrInfo.cpp b/lib/Target/AArch64/AArch64InstrInfo.cpp
index ff115c0..ce85b2c 100644
--- a/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -35,8 +35,14 @@ AArch64InstrInfo::AArch64InstrInfo(const AArch64Subtarget &STI)
 /// GetInstSize - Return the number of bytes of code the specified
 /// instruction may be.  This returns the maximum number of bytes.
 unsigned AArch64InstrInfo::GetInstSizeInBytes(const MachineInstr *MI) const {
-  const MCInstrDesc &Desc = MI->getDesc();
+  const MachineBasicBlock &MBB = *MI->getParent();
+  const MachineFunction *MF = MBB.getParent();
+  const MCAsmInfo *MAI = MF->getTarget().getMCAsmInfo();
+
+  if (MI->getOpcode() == AArch64::INLINEASM)
+    return getInlineAsmLength(MI->getOperand(0).getSymbolName(), *MAI);
 
+  const MCInstrDesc &Desc = MI->getDesc();
   switch (Desc.getOpcode()) {
   default:
     // Anything not explicitly designated otherwise is a nomal 4-byte insn.
@@ -1224,7 +1230,7 @@ void AArch64InstrInfo::copyPhysRegTuple(
     MachineBasicBlock &MBB, MachineBasicBlock::iterator I, DebugLoc DL,
     unsigned DestReg, unsigned SrcReg, bool KillSrc, unsigned Opcode,
     llvm::ArrayRef<unsigned> Indices) const {
-  assert(getSubTarget().hasNEON() &&
+  assert(Subtarget.hasNEON() &&
          "Unexpected register copy without NEON");
   const TargetRegisterInfo *TRI = &getRegisterInfo();
   uint16_t DestEncoding = TRI->getEncodingValue(DestReg);
@@ -1385,7 +1391,7 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   if (AArch64::FPR128RegClass.contains(DestReg) &&
       AArch64::FPR128RegClass.contains(SrcReg)) {
-    if(getSubTarget().hasNEON()) {
+    if(Subtarget.hasNEON()) {
       BuildMI(MBB, I, DL, get(AArch64::ORRv16i8), DestReg)
           .addReg(SrcReg)
           .addReg(SrcReg, getKillRegState(KillSrc));
@@ -1406,7 +1412,7 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   if (AArch64::FPR64RegClass.contains(DestReg) &&
       AArch64::FPR64RegClass.contains(SrcReg)) {
-    if(getSubTarget().hasNEON()) {
+    if(Subtarget.hasNEON()) {
       DestReg = RI.getMatchingSuperReg(DestReg, AArch64::dsub,
                                        &AArch64::FPR128RegClass);
       SrcReg = RI.getMatchingSuperReg(SrcReg, AArch64::dsub,
@@ -1423,7 +1429,7 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   if (AArch64::FPR32RegClass.contains(DestReg) &&
       AArch64::FPR32RegClass.contains(SrcReg)) {
-    if(getSubTarget().hasNEON()) {
+    if(Subtarget.hasNEON()) {
       DestReg = RI.getMatchingSuperReg(DestReg, AArch64::ssub,
                                        &AArch64::FPR128RegClass);
       SrcReg = RI.getMatchingSuperReg(SrcReg, AArch64::ssub,
@@ -1440,7 +1446,7 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   if (AArch64::FPR16RegClass.contains(DestReg) &&
       AArch64::FPR16RegClass.contains(SrcReg)) {
-    if(getSubTarget().hasNEON()) {
+    if(Subtarget.hasNEON()) {
       DestReg = RI.getMatchingSuperReg(DestReg, AArch64::hsub,
                                        &AArch64::FPR128RegClass);
       SrcReg = RI.getMatchingSuperReg(SrcReg, AArch64::hsub,
@@ -1461,7 +1467,7 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   if (AArch64::FPR8RegClass.contains(DestReg) &&
       AArch64::FPR8RegClass.contains(SrcReg)) {
-    if(getSubTarget().hasNEON()) {
+    if(Subtarget.hasNEON()) {
       DestReg = RI.getMatchingSuperReg(DestReg, AArch64::bsub,
                                        &AArch64::FPR128RegClass);
       SrcReg = RI.getMatchingSuperReg(SrcReg, AArch64::bsub,
@@ -1577,39 +1583,39 @@ void AArch64InstrInfo::storeRegToStackSlot(
     if (AArch64::FPR128RegClass.hasSubClassEq(RC))
       Opc = AArch64::STRQui;
     else if (AArch64::DDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Twov1d, Offset = false;
     }
     break;
   case 24:
     if (AArch64::DDDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Threev1d, Offset = false;
     }
     break;
   case 32:
     if (AArch64::DDDDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Fourv1d, Offset = false;
     } else if (AArch64::QQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Twov2d, Offset = false;
     }
     break;
   case 48:
     if (AArch64::QQQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Threev2d, Offset = false;
     }
     break;
   case 64:
     if (AArch64::QQQQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register store without NEON");
       Opc = AArch64::ST1Fourv2d, Offset = false;
     }
@@ -1675,39 +1681,39 @@ void AArch64InstrInfo::loadRegFromStackSlot(
     if (AArch64::FPR128RegClass.hasSubClassEq(RC))
       Opc = AArch64::LDRQui;
     else if (AArch64::DDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Twov1d, Offset = false;
     }
     break;
   case 24:
     if (AArch64::DDDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Threev1d, Offset = false;
     }
     break;
   case 32:
     if (AArch64::DDDDRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Fourv1d, Offset = false;
     } else if (AArch64::QQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Twov2d, Offset = false;
     }
     break;
   case 48:
     if (AArch64::QQQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Threev2d, Offset = false;
     }
     break;
   case 64:
     if (AArch64::QQQQRegClass.hasSubClassEq(RC)) {
-      assert(getSubTarget().hasNEON() &&
+      assert(Subtarget.hasNEON() &&
              "Unexpected register load without NEON");
       Opc = AArch64::LD1Fourv2d, Offset = false;
     }
@@ -1726,7 +1732,7 @@ void AArch64InstrInfo::loadRegFromStackSlot(
 void llvm::emitFrameOffset(MachineBasicBlock &MBB,
                            MachineBasicBlock::iterator MBBI, DebugLoc DL,
                            unsigned DestReg, unsigned SrcReg, int Offset,
-                           const AArch64InstrInfo *TII,
+                           const TargetInstrInfo *TII,
                            MachineInstr::MIFlag Flag, bool SetNZCV) {
   if (DestReg == SrcReg && Offset == 0)
     return;
@@ -1835,7 +1841,7 @@ int llvm::isAArch64FrameOffsetLegal(const MachineInstr &MI, int &Offset,
     *OutUnscaledOp = 0;
   switch (MI.getOpcode()) {
   default:
-    assert(0 && "unhandled opcode in rewriteAArch64FrameIndex");
+    llvm_unreachable("unhandled opcode in rewriteAArch64FrameIndex");
   // Vector spills/fills can't take an immediate offset.
   case AArch64::LD1Twov2d:
   case AArch64::LD1Threev2d:
diff --git a/lib/Target/AArch64/AArch64InstrInfo.h b/lib/Target/AArch64/AArch64InstrInfo.h
index 90ce75f..f70b82b 100644
--- a/lib/Target/AArch64/AArch64InstrInfo.h
+++ b/lib/Target/AArch64/AArch64InstrInfo.h
@@ -44,8 +44,6 @@ public:
   /// always be able to get register info as well (through this method).
   const AArch64RegisterInfo &getRegisterInfo() const { return RI; }
 
-  const AArch64Subtarget &getSubTarget() const { return Subtarget; }
-
   unsigned GetInstSizeInBytes(const MachineInstr *MI) const;
 
   bool isCoalescableExtInstr(const MachineInstr &MI, unsigned &SrcReg,
@@ -168,7 +166,7 @@ private:
 /// if necessary, to be replaced by the scavenger at the end of PEI.
 void emitFrameOffset(MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI,
                      DebugLoc DL, unsigned DestReg, unsigned SrcReg, int Offset,
-                     const AArch64InstrInfo *TII,
+                     const TargetInstrInfo *TII,
                      MachineInstr::MIFlag = MachineInstr::NoFlags,
                      bool SetNZCV = false);
 
diff --git a/lib/Target/AArch64/AArch64InstrInfo.td b/lib/Target/AArch64/AArch64InstrInfo.td
index 9ad36e8..1211fba 100644
--- a/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/lib/Target/AArch64/AArch64InstrInfo.td
@@ -323,7 +323,7 @@ def : Pat<(AArch64LOADgot tconstpool:$addr),
 // System instructions.
 //===----------------------------------------------------------------------===//
 
-def HINT  : HintI<"hint">;
+def HINT : HintI<"hint">;
 def : InstAlias<"nop",  (HINT 0b000)>;
 def : InstAlias<"yield",(HINT 0b001)>;
 def : InstAlias<"wfe",  (HINT 0b010)>;
@@ -671,10 +671,10 @@ def CRC32CXrr : BaseCRC32<1, 0b11, 1, GPR64, int_aarch64_crc32cx, "crc32cx">;
 //===----------------------------------------------------------------------===//
 
 // (immediate)
-defm ANDS : LogicalImmS<0b11, "ands", AArch64and_flag>;
-defm AND  : LogicalImm<0b00, "and", and>;
-defm EOR  : LogicalImm<0b10, "eor", xor>;
-defm ORR  : LogicalImm<0b01, "orr", or>;
+defm ANDS : LogicalImmS<0b11, "ands", AArch64and_flag, "bics">;
+defm AND  : LogicalImm<0b00, "and", and, "bic">;
+defm EOR  : LogicalImm<0b10, "eor", xor, "eon">;
+defm ORR  : LogicalImm<0b01, "orr", or, "orn">;
 
 // FIXME: these aliases *are* canonical sometimes (when movz can't be
 // used). Actually, it seems to be working right now, but putting logical_immXX
@@ -737,6 +737,10 @@ def : Pat<(not GPR64:$Xm), (ORNXrr XZR, GPR64:$Xm)>;
 defm CLS    : OneOperandData<0b101, "cls">;
 defm CLZ    : OneOperandData<0b100, "clz", ctlz>;
 defm RBIT   : OneOperandData<0b000, "rbit">;
+
+def : Pat<(int_aarch64_rbit GPR32:$Rn), (RBITWr $Rn)>;
+def : Pat<(int_aarch64_rbit GPR64:$Rn), (RBITXr $Rn)>;
+
 def  REV16Wr : OneWRegData<0b001, "rev16",
                                   UnOpFrag<(rotr (bswap node:$LHS), (i64 16))>>;
 def  REV16Xr : OneXRegData<0b001, "rev16", null_frag>;
@@ -2238,6 +2242,81 @@ def : Pat<(f32_to_f16 FPR32:$Rn),
 def FCVTSHpseudo : Pseudo<(outs FPR32:$Rd), (ins FPR32:$Rn),
                           [(set (f32 FPR32:$Rd), (f16_to_f32 i32:$Rn))]>;
 
+// When converting from f16 coming directly from a load, make sure we
+// load into the FPR16 registers rather than going through the GPRs.
+//   f16->f32
+def : Pat<(f32 (f16_to_f32 (i32
+                (zextloadi16 (ro_Windexed16 GPR64sp:$Rn, GPR32:$Rm,
+                                    ro_Wextend16:$extend))))),
+          (FCVTSHr (LDRHroW GPR64sp:$Rn, GPR32:$Rm, ro_Wextend16:$extend))>;
+def : Pat<(f32 (f16_to_f32 (i32
+                (zextloadi16 (ro_Xindexed16 GPR64sp:$Rn, GPR64:$Rm,
+                                    ro_Xextend16:$extend))))),
+          (FCVTSHr (LDRHroX GPR64sp:$Rn, GPR64:$Rm, ro_Xextend16:$extend))>;
+def : Pat <(f32 (f16_to_f32 (i32
+                  (zextloadi16 (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset))))),
+           (FCVTSHr (LDRHui GPR64sp:$Rn, uimm12s2:$offset))>;
+def : Pat <(f32 (f16_to_f32 (i32
+                  (zextloadi16 (am_unscaled16 GPR64sp:$Rn, simm9:$offset))))),
+           (FCVTSHr (LDURHi GPR64sp:$Rn, simm9:$offset))>;
+
+//   f16->f64
+def : Pat<(f64 (fextend (f32 (f16_to_f32 (i32
+                (zextloadi16 (ro_Windexed16 GPR64sp:$Rn, GPR32:$Rm,
+                                    ro_Wextend16:$extend))))))),
+          (FCVTDHr (LDRHroW GPR64sp:$Rn, GPR32:$Rm, ro_Wextend16:$extend))>;
+def : Pat<(f64 (fextend (f32 (f16_to_f32 (i32
+                (zextloadi16 (ro_Xindexed16 GPR64sp:$Rn, GPR64:$Rm,
+                                    ro_Xextend16:$extend))))))),
+          (FCVTDHr (LDRHroX GPR64sp:$Rn, GPR64:$Rm, ro_Xextend16:$extend))>;
+def : Pat <(f64 (fextend (f32 (f16_to_f32 (i32
+                  (zextloadi16 (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset))))))),
+           (FCVTDHr (LDRHui GPR64sp:$Rn, uimm12s2:$offset))>;
+def : Pat <(f64 (fextend (f32 (f16_to_f32 (i32
+                  (zextloadi16 (am_unscaled16 GPR64sp:$Rn, simm9:$offset))))))),
+           (FCVTDHr (LDURHi GPR64sp:$Rn, simm9:$offset))>;
+
+// When converting to f16 going directly to a store, make sure we use the
+// appropriate direct conversion instructions and store via the FPR16
+// registers rather than going through the GPRs.
+let AddedComplexity = 10 in {
+// f32->f16
+def : Pat< (truncstorei16 (assertzext (i32 (f32_to_f16 FPR32:$Rt))),
+                          (ro_Windexed16 GPR64sp:$Rn, GPR32:$Rm,
+                                         ro_Wextend16:$extend)),
+           (STRHroW (FCVTHSr FPR32:$Rt), GPR64sp:$Rn, GPR32:$Rm,
+                                         ro_Wextend16:$extend)>;
+def : Pat< (truncstorei16 (assertzext (i32 (f32_to_f16 FPR32:$Rt))),
+                          (ro_Xindexed16 GPR64sp:$Rn, GPR64:$Rm,
+                                         ro_Xextend16:$extend)),
+           (STRHroX (FCVTHSr FPR32:$Rt), GPR64sp:$Rn, GPR64:$Rm,
+                                         ro_Xextend16:$extend)>;
+def : Pat <(truncstorei16 (assertzext (i32 (f32_to_f16 FPR32:$Rt))),
+              (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset)),
+           (STRHui (FCVTHSr FPR32:$Rt), GPR64sp:$Rn, uimm12s2:$offset)>;
+def : Pat <(truncstorei16 (assertzext (i32 (f32_to_f16 FPR32:$Rt))),
+              (am_unscaled16 GPR64sp:$Rn, simm9:$offset)),
+           (STURHi (FCVTHSr FPR32:$Rt), GPR64sp:$Rn, simm9:$offset)>;
+// f64->f16
+def : Pat< (truncstorei16 (assertzext (i32 (f32_to_f16 (f32 (fround FPR64:$Rt))))),
+                          (ro_Windexed16 GPR64sp:$Rn, GPR32:$Rm,
+                                         ro_Wextend16:$extend)),
+           (STRHroW (FCVTHDr FPR64:$Rt), GPR64sp:$Rn, GPR32:$Rm,
+                                         ro_Wextend16:$extend)>;
+def : Pat< (truncstorei16 (assertzext (i32 (f32_to_f16 (f32 (fround FPR64:$Rt))))),
+                          (ro_Xindexed16 GPR64sp:$Rn, GPR64:$Rm,
+                                         ro_Xextend16:$extend)),
+           (STRHroX (FCVTHDr FPR64:$Rt), GPR64sp:$Rn, GPR64:$Rm,
+                                         ro_Xextend16:$extend)>;
+def : Pat <(truncstorei16 (assertzext (i32 (f32_to_f16 (f32 (fround FPR64:$Rt))))),
+              (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset)),
+           (STRHui (FCVTHDr FPR64:$Rt), GPR64sp:$Rn, uimm12s2:$offset)>;
+def : Pat <(truncstorei16 (assertzext (i32 (f32_to_f16 (f32 (fround FPR64:$Rt))))),
+              (am_unscaled16 GPR64sp:$Rn, simm9:$offset)),
+           (STURHi (FCVTHDr FPR64:$Rt), GPR64sp:$Rn, simm9:$offset)>;
+}
+
+
 //===----------------------------------------------------------------------===//
 // Floating point single operand instructions.
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index e7454be..3df9c4f 100644
--- a/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -40,14 +40,13 @@ STATISTIC(NumPreFolded, "Number of pre-index updates folded");
 STATISTIC(NumUnscaledPairCreated,
           "Number of load/store from unscaled generated");
 
-static cl::opt<unsigned> ScanLimit("aarch64-load-store-scan-limit", cl::init(20),
-                                   cl::Hidden);
+static cl::opt<unsigned> ScanLimit("aarch64-load-store-scan-limit",
+                                   cl::init(20), cl::Hidden);
 
 // Place holder while testing unscaled load/store combining
-static cl::opt<bool>
-EnableAArch64UnscaledMemOp("aarch64-unscaled-mem-op", cl::Hidden,
-                         cl::desc("Allow AArch64 unscaled load/store combining"),
-                         cl::init(true));
+static cl::opt<bool> EnableAArch64UnscaledMemOp(
+    "aarch64-unscaled-mem-op", cl::Hidden,
+    cl::desc("Allow AArch64 unscaled load/store combining"), cl::init(true));
 
 namespace {
 struct AArch64LoadStoreOpt : public MachineFunctionPass {
@@ -60,19 +59,19 @@ struct AArch64LoadStoreOpt : public MachineFunctionPass {
   // Scan the instructions looking for a load/store that can be combined
   // with the current instruction into a load/store pair.
   // Return the matching instruction if one is found, else MBB->end().
-  // If a matching instruction is found, mergeForward is set to true if the
+  // If a matching instruction is found, MergeForward is set to true if the
   // merge is to remove the first instruction and replace the second with
   // a pair-wise insn, and false if the reverse is true.
   MachineBasicBlock::iterator findMatchingInsn(MachineBasicBlock::iterator I,
-                                               bool &mergeForward,
+                                               bool &MergeForward,
                                                unsigned Limit);
   // Merge the two instructions indicated into a single pair-wise instruction.
-  // If mergeForward is true, erase the first instruction and fold its
+  // If MergeForward is true, erase the first instruction and fold its
   // operation into the second. If false, the reverse. Return the instruction
   // following the first instruction (which may change during processing).
   MachineBasicBlock::iterator
   mergePairedInsns(MachineBasicBlock::iterator I,
-                   MachineBasicBlock::iterator Paired, bool mergeForward);
+                   MachineBasicBlock::iterator Paired, bool MergeForward);
 
   // Scan the instruction list to find a base register update that can
   // be combined with the current instruction (a load or store) using
@@ -142,7 +141,7 @@ static bool isUnscaledLdst(unsigned Opc) {
 int AArch64LoadStoreOpt::getMemSize(MachineInstr *MemMI) {
   switch (MemMI->getOpcode()) {
   default:
-    llvm_unreachable("Opcode has has unknown size!");
+    llvm_unreachable("Opcode has unknown size!");
   case AArch64::STRSui:
   case AArch64::STURSi:
     return 4;
@@ -217,16 +216,26 @@ static unsigned getPreIndexedOpcode(unsigned Opc) {
   switch (Opc) {
   default:
     llvm_unreachable("Opcode has no pre-indexed equivalent!");
-  case AArch64::STRSui:    return AArch64::STRSpre;
-  case AArch64::STRDui:    return AArch64::STRDpre;
-  case AArch64::STRQui:    return AArch64::STRQpre;
-  case AArch64::STRWui:    return AArch64::STRWpre;
-  case AArch64::STRXui:    return AArch64::STRXpre;
-  case AArch64::LDRSui:    return AArch64::LDRSpre;
-  case AArch64::LDRDui:    return AArch64::LDRDpre;
-  case AArch64::LDRQui:    return AArch64::LDRQpre;
-  case AArch64::LDRWui:    return AArch64::LDRWpre;
-  case AArch64::LDRXui:    return AArch64::LDRXpre;
+  case AArch64::STRSui:
+    return AArch64::STRSpre;
+  case AArch64::STRDui:
+    return AArch64::STRDpre;
+  case AArch64::STRQui:
+    return AArch64::STRQpre;
+  case AArch64::STRWui:
+    return AArch64::STRWpre;
+  case AArch64::STRXui:
+    return AArch64::STRXpre;
+  case AArch64::LDRSui:
+    return AArch64::LDRSpre;
+  case AArch64::LDRDui:
+    return AArch64::LDRDpre;
+  case AArch64::LDRQui:
+    return AArch64::LDRQpre;
+  case AArch64::LDRWui:
+    return AArch64::LDRWpre;
+  case AArch64::LDRXui:
+    return AArch64::LDRXpre;
   }
 }
 
@@ -260,7 +269,7 @@ static unsigned getPostIndexedOpcode(unsigned Opc) {
 MachineBasicBlock::iterator
 AArch64LoadStoreOpt::mergePairedInsns(MachineBasicBlock::iterator I,
                                       MachineBasicBlock::iterator Paired,
-                                      bool mergeForward) {
+                                      bool MergeForward) {
   MachineBasicBlock::iterator NextI = I;
   ++NextI;
   // If NextI is the second of the two instructions to be merged, we need
@@ -276,12 +285,12 @@ AArch64LoadStoreOpt::mergePairedInsns(MachineBasicBlock::iterator I,
 
   unsigned NewOpc = getMatchingPairOpcode(I->getOpcode());
   // Insert our new paired instruction after whichever of the paired
-  // instructions mergeForward indicates.
-  MachineBasicBlock::iterator InsertionPoint = mergeForward ? Paired : I;
-  // Also based on mergeForward is from where we copy the base register operand
+  // instructions MergeForward indicates.
+  MachineBasicBlock::iterator InsertionPoint = MergeForward ? Paired : I;
+  // Also based on MergeForward is from where we copy the base register operand
   // so we get the flags compatible with the input code.
   MachineOperand &BaseRegOp =
-      mergeForward ? Paired->getOperand(1) : I->getOperand(1);
+      MergeForward ? Paired->getOperand(1) : I->getOperand(1);
 
   // Which register is Rt and which is Rt2 depends on the offset order.
   MachineInstr *RtMI, *Rt2MI;
@@ -355,8 +364,8 @@ static bool inBoundsForPair(bool IsUnscaled, int Offset, int OffsetStride) {
   if (IsUnscaled) {
     // Convert the byte-offset used by unscaled into an "element" offset used
     // by the scaled pair load/store instructions.
-    int elemOffset = Offset / OffsetStride;
-    if (elemOffset > 63 || elemOffset < -64)
+    int ElemOffset = Offset / OffsetStride;
+    if (ElemOffset > 63 || ElemOffset < -64)
       return false;
   }
   return true;
@@ -374,14 +383,14 @@ static int alignTo(int Num, int PowOf2) {
 /// be combined with the current instruction into a load/store pair.
 MachineBasicBlock::iterator
 AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
-                                      bool &mergeForward, unsigned Limit) {
+                                      bool &MergeForward, unsigned Limit) {
   MachineBasicBlock::iterator E = I->getParent()->end();
   MachineBasicBlock::iterator MBBI = I;
   MachineInstr *FirstMI = I;
   ++MBBI;
 
   int Opc = FirstMI->getOpcode();
-  bool mayLoad = FirstMI->mayLoad();
+  bool MayLoad = FirstMI->mayLoad();
   bool IsUnscaled = isUnscaledLdst(Opc);
   unsigned Reg = FirstMI->getOperand(0).getReg();
   unsigned BaseReg = FirstMI->getOperand(1).getReg();
@@ -453,7 +462,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
         // If the destination register of the loads is the same register, bail
         // and keep looking. A load-pair instruction with both destination
         // registers the same is UNPREDICTABLE and will result in an exception.
-        if (mayLoad && Reg == MI->getOperand(0).getReg()) {
+        if (MayLoad && Reg == MI->getOperand(0).getReg()) {
           trackRegDefsUses(MI, ModifiedRegs, UsedRegs, TRI);
           continue;
         }
@@ -462,7 +471,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
         // the two instructions, we can combine the second into the first.
         if (!ModifiedRegs[MI->getOperand(0).getReg()] &&
             !UsedRegs[MI->getOperand(0).getReg()]) {
-          mergeForward = false;
+          MergeForward = false;
           return MBBI;
         }
 
@@ -471,7 +480,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
         // second.
         if (!ModifiedRegs[FirstMI->getOperand(0).getReg()] &&
             !UsedRegs[FirstMI->getOperand(0).getReg()]) {
-          mergeForward = true;
+          MergeForward = true;
           return MBBI;
         }
         // Unable to combine these instructions due to interference in between.
@@ -798,14 +807,14 @@ bool AArch64LoadStoreOpt::optimizeBlock(MachineBasicBlock &MBB) {
         break;
       }
       // Look ahead up to ScanLimit instructions for a pairable instruction.
-      bool mergeForward = false;
+      bool MergeForward = false;
       MachineBasicBlock::iterator Paired =
-          findMatchingInsn(MBBI, mergeForward, ScanLimit);
+          findMatchingInsn(MBBI, MergeForward, ScanLimit);
       if (Paired != E) {
         // Merge the loads into a pair. Keeping the iterator straight is a
         // pain, so we let the merge routine tell us what the next instruction
         // is after it's done mucking about.
-        MBBI = mergePairedInsns(MBBI, Paired, mergeForward);
+        MBBI = mergePairedInsns(MBBI, Paired, MergeForward);
 
         Modified = true;
         ++NumPairCreated;
diff --git a/lib/Target/AArch64/AArch64MCInstLower.cpp b/lib/Target/AArch64/AArch64MCInstLower.cpp
index ab6d375..75a17b9 100644
--- a/lib/Target/AArch64/AArch64MCInstLower.cpp
+++ b/lib/Target/AArch64/AArch64MCInstLower.cpp
@@ -51,7 +51,7 @@ MCOperand AArch64MCInstLower::lowerSymbolOperandDarwin(const MachineOperand &MO,
              AArch64II::MO_PAGEOFF)
       RefKind = MCSymbolRefExpr::VK_GOTPAGEOFF;
     else
-      assert(0 && "Unexpected target flags with MO_GOT on GV operand");
+      llvm_unreachable("Unexpected target flags with MO_GOT on GV operand");
   } else if ((MO.getTargetFlags() & AArch64II::MO_TLS) != 0) {
     if ((MO.getTargetFlags() & AArch64II::MO_FRAGMENT) == AArch64II::MO_PAGE)
       RefKind = MCSymbolRefExpr::VK_TLVPPAGE;
@@ -154,7 +154,7 @@ bool AArch64MCInstLower::lowerOperand(const MachineOperand &MO,
                                       MCOperand &MCOp) const {
   switch (MO.getType()) {
   default:
-    assert(0 && "unknown operand type");
+    llvm_unreachable("unknown operand type");
   case MachineOperand::MO_Register:
     // Ignore all implicit register operands.
     if (MO.isImplicit())
diff --git a/lib/Target/AArch64/AArch64RegisterInfo.td b/lib/Target/AArch64/AArch64RegisterInfo.td
index 21c927f..a30e4ad 100644
--- a/lib/Target/AArch64/AArch64RegisterInfo.td
+++ b/lib/Target/AArch64/AArch64RegisterInfo.td
@@ -175,7 +175,7 @@ def GPR64all : RegisterClass<"AArch64", [i64], 64, (add GPR64common, XZR, SP)>;
 // This is for indirect tail calls to store the address of the destination.
 def tcGPR64 : RegisterClass<"AArch64", [i64], 64, (sub GPR64common, X19, X20, X21,
                                                      X22, X23, X24, X25, X26,
-                                                     X27, X28)>;
+                                                     X27, X28, FP, LR)>;
 
 // GPR register classes for post increment amount of vector load/store that
 // has alternate printing when Rm=31 and prints a constant immediate value
diff --git a/lib/Target/AArch64/AArch64SchedA53.td b/lib/Target/AArch64/AArch64SchedA53.td
index 0c3949e..d709bee 100644
--- a/lib/Target/AArch64/AArch64SchedA53.td
+++ b/lib/Target/AArch64/AArch64SchedA53.td
@@ -148,9 +148,9 @@ def : ReadAdvance<ReadVLD, 0>;
 
 // ALU - Most operands in the ALU pipes are not needed for two cycles. Shiftable
 //       operands are needed one cycle later if and only if they are to be
-//       shifted. Otherwise, they too are needed two cycle later. This same
+//       shifted. Otherwise, they too are needed two cycles later. This same
 //       ReadAdvance applies to Extended registers as well, even though there is
-//       a seperate SchedPredicate for them.
+//       a separate SchedPredicate for them.
 def : ReadAdvance<ReadI, 2, [WriteImm,WriteI,
                              WriteISReg, WriteIEReg,WriteIS,
                              WriteID32,WriteID64,
diff --git a/lib/Target/AArch64/AArch64SchedA57.td b/lib/Target/AArch64/AArch64SchedA57.td
new file mode 100644
index 0000000..8209f96
--- /dev/null
+++ b/lib/Target/AArch64/AArch64SchedA57.td
@@ -0,0 +1,304 @@
+//=- AArch64SchedA57.td - ARM Cortex-A57 Scheduling Defs -----*- tablegen -*-=//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the machine model for ARM Cortex-A57 to support
+// instruction scheduling and other instruction cost heuristics.
+//
+//===----------------------------------------------------------------------===//
+
+def CortexA57Model : SchedMachineModel {
+  let IssueWidth        =   8; // 3-way decode and 8-way issue
+  let MicroOpBufferSize = 128; // 128 micro-op re-order buffer
+  let LoadLatency       =   4; // Optimistic load latency
+  let MispredictPenalty =  14; // Fetch + Decode/Rename/Dispatch + Branch
+}
+
+//===----------------------------------------------------------------------===//
+// Define each kind of processor resource and number available on Cortex-A57.
+// Cortex A-57 has 8 pipelines that each has its own 8-entry queue where
+// micro-ops wait for their operands and then issue out-of-order.
+
+def A57UnitB : ProcResource<1> { let BufferSize = 8; }  // Type B micro-ops
+def A57UnitI : ProcResource<2> { let BufferSize = 8; }  // Type I micro-ops
+def A57UnitM : ProcResource<1> { let BufferSize = 8; }  // Type M micro-ops
+def A57UnitL : ProcResource<1> { let BufferSize = 8; }  // Type L micro-ops
+def A57UnitS : ProcResource<1> { let BufferSize = 8; }  // Type S micro-ops
+def A57UnitX : ProcResource<1> { let BufferSize = 8; }  // Type X micro-ops
+def A57UnitW : ProcResource<1> { let BufferSize = 8; }  // Type W micro-ops
+let SchedModel = CortexA57Model in {
+  def A57UnitV : ProcResGroup<[A57UnitX, A57UnitW]>;    // Type V micro-ops
+}
+
+
+let SchedModel = CortexA57Model in {
+
+//===----------------------------------------------------------------------===//
+// Define customized scheduler read/write types specific to the Cortex-A57.
+
+include "AArch64SchedA57WriteRes.td"
+
+//===----------------------------------------------------------------------===//
+// Map the target-defined scheduler read/write resources and latency for
+// Cortex-A57. The Cortex-A57 types are directly associated with resources, so
+// defining the aliases precludes the need for mapping them using WriteRes. The
+// aliases are sufficient for creating a coarse, working model. As the model
+// evolves, InstRWs will be used to override these SchedAliases.
+
+def : SchedAlias<WriteImm,   A57Write_1cyc_1I>;
+def : SchedAlias<WriteI,     A57Write_1cyc_1I>;
+def : SchedAlias<WriteISReg, A57Write_2cyc_1M>;
+def : SchedAlias<WriteIEReg, A57Write_2cyc_1M>;
+def : SchedAlias<WriteExtr,  A57Write_1cyc_1I>;
+def : SchedAlias<WriteIS,    A57Write_1cyc_1I>;
+def : SchedAlias<WriteID32,  A57Write_19cyc_1M>;
+def : SchedAlias<WriteID64,  A57Write_35cyc_1M>;
+def : SchedAlias<WriteIM32,  A57Write_3cyc_1M>;
+def : SchedAlias<WriteIM64,  A57Write_5cyc_1M>;
+def : SchedAlias<WriteBr,    A57Write_1cyc_1B>;
+def : SchedAlias<WriteBrReg, A57Write_1cyc_1B>;
+def : SchedAlias<WriteLD,    A57Write_4cyc_1L>;
+def : SchedAlias<WriteST,    A57Write_1cyc_1S>;
+def : SchedAlias<WriteSTP,   A57Write_1cyc_1S>;
+def : SchedAlias<WriteAdr,   A57Write_1cyc_1I>;
+def : SchedAlias<WriteLDIdx, A57Write_4cyc_1I_1L>;
+def : SchedAlias<WriteSTIdx, A57Write_1cyc_1I_1S>;
+def : SchedAlias<WriteF,     A57Write_3cyc_1V>;
+def : SchedAlias<WriteFCmp,  A57Write_3cyc_1V>;
+def : SchedAlias<WriteFCvt,  A57Write_5cyc_1V>;
+def : SchedAlias<WriteFCopy, A57Write_3cyc_1V>;
+def : SchedAlias<WriteFImm,  A57Write_3cyc_1V>;
+def : SchedAlias<WriteFMul,  A57Write_5cyc_1V>;
+def : SchedAlias<WriteFDiv,  A57Write_18cyc_1X>;
+def : SchedAlias<WriteV,     A57Write_3cyc_1V>;
+def : SchedAlias<WriteVLD,   A57Write_5cyc_1L>;
+def : SchedAlias<WriteVST,   A57Write_1cyc_1S>;
+
+def : WriteRes<WriteSys,     []> { let Latency = 1; }
+def : WriteRes<WriteBarrier, []> { let Latency = 1; }
+def : WriteRes<WriteHint,    []> { let Latency = 1; }
+
+def : WriteRes<WriteLDHi,    []> { let Latency = 4; }
+
+// Forwarding logic is not [yet] explicitly modeled beyond what is captured
+// in the latencies of the A57 Generic SchedWriteRes's.
+def : ReadAdvance<ReadI,       0>;
+def : ReadAdvance<ReadISReg,   0>;
+def : ReadAdvance<ReadIEReg,   0>;
+def : ReadAdvance<ReadIM,      0>;
+def : ReadAdvance<ReadIMA,     0>;
+def : ReadAdvance<ReadID,      0>;
+def : ReadAdvance<ReadExtrHi,  0>;
+def : ReadAdvance<ReadAdrBase, 0>;
+def : ReadAdvance<ReadVLD,     0>;
+
+
+//===----------------------------------------------------------------------===//
+// Specialize the coarse model by associating instruction groups with the
+// subtarget-defined types. As the modeled is refined, this will override most
+// of the above ShchedAlias mappings.
+
+// Miscellaneous
+// -----------------------------------------------------------------------------
+
+def : InstRW<[WriteI], (instrs COPY)>;
+
+
+// Branch Instructions
+// -----------------------------------------------------------------------------
+
+def : InstRW<[A57Write_1cyc_1B_1I], (instrs BL)>;
+def : InstRW<[A57Write_2cyc_1B_1I], (instrs BLR)>;
+
+
+// Divide and Multiply Instructions
+// -----------------------------------------------------------------------------
+
+// Multiply high
+def : InstRW<[A57Write_6cyc_1M], (instrs SMULHrr, UMULHrr)>;
+
+
+// Miscellaneous Data-Processing Instructions
+// -----------------------------------------------------------------------------
+
+def : InstRW<[A57Write_1cyc_1I],    (instrs EXTRWrri)>;
+def : InstRW<[A57Write_3cyc_1I_1M], (instrs EXTRXrri)>;
+def : InstRW<[A57Write_2cyc_1M],    (instregex "BFM")>;
+
+
+// Cryptography Extensions
+// -----------------------------------------------------------------------------
+
+def : InstRW<[A57Write_3cyc_1W], (instregex "CRC32")>;
+
+
+// Vector Load
+// -----------------------------------------------------------------------------
+
+def : InstRW<[A57Write_8cyc_1L_1V],           (instregex "LD1i(8|16|32)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr], (instregex "LD1i(8|16|32)_POST$")>;
+def : InstRW<[A57Write_5cyc_1L],            (instregex "LD1i(64)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],  (instregex "LD1i(64)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_1V],           (instregex "LD1Rv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr], (instregex "LD1Rv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_5cyc_1L],            (instregex "LD1Rv(1d)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],  (instregex "LD1Rv(1d)_POST$")>;
+def : InstRW<[A57Write_8cyc_1L_1V],           (instregex "LD1Rv(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr], (instregex "LD1Rv(16b|8h|4s|2d)_POST$")>;
+
+def : InstRW<[A57Write_5cyc_1L],              (instregex "LD1Onev(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],    (instregex "LD1Onev(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_5cyc_1L],              (instregex "LD1Onev(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],    (instregex "LD1Onev(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_5cyc_1L],              (instregex "LD1Twov(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],    (instregex "LD1Twov(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],             (instregex "LD1Twov(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],   (instregex "LD1Twov(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],             (instregex "LD1Threev(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],   (instregex "LD1Threev(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_7cyc_3L],            (instregex "LD1Threev(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_7cyc_3L, WriteAdr],  (instregex "LD1Threev(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],             (instregex "LD1Fourv(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],   (instregex "LD1Fourv(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_8cyc_4L],           (instregex "LD1Fourv(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_8cyc_4L, WriteAdr], (instregex "LD1Fourv(16b|8h|4s|2d)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_2V],           (instregex "LD2i(8|16)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr], (instregex "LD2i(8|16)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],            (instregex "LD2i(32)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],  (instregex "LD2i(32)_POST$")>;
+def : InstRW<[A57Write_8cyc_1L_1V],            (instregex "LD2i(64)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr],  (instregex "LD2i(64)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_1V],            (instregex "LD2Rv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr],  (instregex "LD2Rv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_5cyc_1L],             (instregex "LD2Rv(1d)$")>;
+def : InstRW<[A57Write_5cyc_1L, WriteAdr],   (instregex "LD2Rv(1d)_POST$")>;
+def : InstRW<[A57Write_8cyc_1L_2V],           (instregex "LD2Rv(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr], (instregex "LD2Rv(16b|8h|4s|2d)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_1V],             (instregex "LD2Twov(8b|4h|2s)$")>;
+def : InstRW<[A57Write_8cyc_1L_1V, WriteAdr],   (instregex "LD2Twov(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_9cyc_2L_2V],           (instregex "LD2Twov(16b|8h|4s)$")>;
+def : InstRW<[A57Write_9cyc_2L_2V, WriteAdr], (instregex "LD2Twov(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],             (instregex "LD2Twov(2d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],   (instregex "LD2Twov(2d)_POST$")>;
+
+def : InstRW<[A57Write_9cyc_1L_3V],           (instregex "LD3i(8|16)$")>;
+def : InstRW<[A57Write_9cyc_1L_3V, WriteAdr], (instregex "LD3i(8|16)_POST$")>;
+def : InstRW<[A57Write_8cyc_1L_2V],            (instregex "LD3i(32)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr],  (instregex "LD3i(32)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],             (instregex "LD3i(64)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],   (instregex "LD3i(64)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_2V],             (instregex "LD3Rv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr],   (instregex "LD3Rv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],              (instregex "LD3Rv(1d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],    (instregex "LD3Rv(1d)_POST$")>;
+def : InstRW<[A57Write_9cyc_1L_3V],            (instregex "LD3Rv(16b|8h|4s)$")>;
+def : InstRW<[A57Write_9cyc_1L_3V, WriteAdr],  (instregex "LD3Rv(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_9cyc_2L_3V],           (instregex "LD3Rv(2d)$")>;
+def : InstRW<[A57Write_9cyc_2L_3V, WriteAdr], (instregex "LD3Rv(2d)_POST$")>;
+
+def : InstRW<[A57Write_9cyc_2L_2V],               (instregex "LD3Threev(8b|4h|2s)$")>;
+def : InstRW<[A57Write_9cyc_2L_2V, WriteAdr],     (instregex "LD3Threev(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_10cyc_3L_4V],           (instregex "LD3Threev(16b|8h|4s)$")>;
+def : InstRW<[A57Write_10cyc_3L_4V, WriteAdr], (instregex "LD3Threev(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_8cyc_4L],               (instregex "LD3Threev(2d)$")>;
+def : InstRW<[A57Write_8cyc_4L, WriteAdr],     (instregex "LD3Threev(2d)_POST$")>;
+
+def : InstRW<[A57Write_9cyc_2L_3V],           (instregex "LD4i(8|16)$")>;
+def : InstRW<[A57Write_9cyc_2L_3V, WriteAdr], (instregex "LD4i(8|16)_POST$")>;
+def : InstRW<[A57Write_8cyc_1L_2V],             (instregex "LD4i(32)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr],   (instregex "LD4i(32)_POST$")>;
+def : InstRW<[A57Write_9cyc_2L_3V],           (instregex "LD4i(64)$")>;
+def : InstRW<[A57Write_9cyc_2L_3V, WriteAdr], (instregex "LD4i(64)_POST$")>;
+
+def : InstRW<[A57Write_8cyc_1L_2V],              (instregex "LD4Rv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_8cyc_1L_2V, WriteAdr],    (instregex "LD4Rv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_6cyc_2L],               (instregex "LD4Rv(1d)$")>;
+def : InstRW<[A57Write_6cyc_2L, WriteAdr],     (instregex "LD4Rv(1d)_POST$")>;
+def : InstRW<[A57Write_9cyc_2L_3V],            (instregex "LD4Rv(16b|8h|4s)$")>;
+def : InstRW<[A57Write_9cyc_2L_3V, WriteAdr],  (instregex "LD4Rv(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_9cyc_2L_4V],           (instregex "LD4Rv(2d)$")>;
+def : InstRW<[A57Write_9cyc_2L_4V, WriteAdr], (instregex "LD4Rv(2d)_POST$")>;
+
+def : InstRW<[A57Write_9cyc_2L_2V],                (instregex "LD4Fourv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_9cyc_2L_2V, WriteAdr],      (instregex "LD4Fourv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_11cyc_4L_4V],           (instregex "LD4Fourv(16b|8h|4s)$")>;
+def : InstRW<[A57Write_11cyc_4L_4V, WriteAdr], (instregex "LD4Fourv(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_8cyc_4L],                (instregex "LD4Fourv(2d)$")>;
+def : InstRW<[A57Write_8cyc_4L, WriteAdr],      (instregex "LD4Fourv(2d)_POST$")>;
+
+// Vector Store
+// -----------------------------------------------------------------------------
+
+def : InstRW<[A57Write_1cyc_1S],            (instregex "ST1i(8|16|32)$")>;
+def : InstRW<[A57Write_1cyc_1S, WriteAdr],  (instregex "ST1i(8|16|32)_POST$")>;
+def : InstRW<[A57Write_3cyc_1S_1V],           (instregex "ST1i(64)$")>;
+def : InstRW<[A57Write_3cyc_1S_1V, WriteAdr], (instregex "ST1i(64)_POST$")>;
+
+def : InstRW<[A57Write_1cyc_1S],                  (instregex "ST1Onev(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_1cyc_1S, WriteAdr],        (instregex "ST1Onev(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_2cyc_2S],                 (instregex "ST1Onev(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_2cyc_2S, WriteAdr],       (instregex "ST1Onev(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_2cyc_2S],                 (instregex "ST1Twov(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_2cyc_2S, WriteAdr],       (instregex "ST1Twov(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_4cyc_4S],               (instregex "ST1Twov(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_4cyc_4S, WriteAdr],     (instregex "ST1Twov(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_3cyc_3S],                (instregex "ST1Threev(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_3cyc_3S, WriteAdr],      (instregex "ST1Threev(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_6cyc_6S],             (instregex "ST1Threev(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_6cyc_6S, WriteAdr],   (instregex "ST1Threev(16b|8h|4s|2d)_POST$")>;
+def : InstRW<[A57Write_4cyc_4S],               (instregex "ST1Fourv(8b|4h|2s|1d)$")>;
+def : InstRW<[A57Write_4cyc_4S, WriteAdr],     (instregex "ST1Fourv(8b|4h|2s|1d)_POST$")>;
+def : InstRW<[A57Write_8cyc_8S],           (instregex "ST1Fourv(16b|8h|4s|2d)$")>;
+def : InstRW<[A57Write_8cyc_8S, WriteAdr], (instregex "ST1Fourv(16b|8h|4s|2d)_POST$")>;
+
+def : InstRW<[A57Write_3cyc_1S_1V],           (instregex "ST2i(8|16|32)$")>;
+def : InstRW<[A57Write_3cyc_1S_1V, WriteAdr], (instregex "ST2i(8|16|32)_POST$")>;
+def : InstRW<[A57Write_2cyc_2S],           (instregex "ST2i(64)$")>;
+def : InstRW<[A57Write_2cyc_2S, WriteAdr], (instregex "ST2i(64)_POST$")>;
+
+def : InstRW<[A57Write_3cyc_2S_1V],              (instregex "ST2Twov(8b|4h|2s)$")>;
+def : InstRW<[A57Write_3cyc_2S_1V, WriteAdr],    (instregex "ST2Twov(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_4cyc_4S_2V],           (instregex "ST2Twov(16b|8h|4s)$")>;
+def : InstRW<[A57Write_4cyc_4S_2V, WriteAdr], (instregex "ST2Twov(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_4cyc_4S],             (instregex "ST2Twov(2d)$")>;
+def : InstRW<[A57Write_4cyc_4S, WriteAdr],   (instregex "ST2Twov(2d)_POST$")>;
+
+def : InstRW<[A57Write_3cyc_1S_1V],            (instregex "ST3i(8|16)$")>;
+def : InstRW<[A57Write_3cyc_1S_1V, WriteAdr],  (instregex "ST3i(8|16)_POST$")>;
+def : InstRW<[A57Write_3cyc_3S],           (instregex "ST3i(32)$")>;
+def : InstRW<[A57Write_3cyc_3S, WriteAdr], (instregex "ST3i(32)_POST$")>;
+def : InstRW<[A57Write_3cyc_2S_1V],           (instregex "ST3i(64)$")>;
+def : InstRW<[A57Write_3cyc_2S_1V, WriteAdr], (instregex "ST3i(64)_POST$")>;
+
+def : InstRW<[A57Write_3cyc_3S_2V],                 (instregex "ST3Threev(8b|4h|2s)$")>;
+def : InstRW<[A57Write_3cyc_3S_2V, WriteAdr],       (instregex "ST3Threev(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_6cyc_6S_4V],           (instregex "ST3Threev(16b|8h|4s)$")>;
+def : InstRW<[A57Write_6cyc_6S_4V, WriteAdr], (instregex "ST3Threev(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_6cyc_6S],                (instregex "ST3Threev(2d)$")>;
+def : InstRW<[A57Write_6cyc_6S, WriteAdr],      (instregex "ST3Threev(2d)_POST$")>;
+
+def : InstRW<[A57Write_3cyc_1S_1V],             (instregex "ST4i(8|16)$")>;
+def : InstRW<[A57Write_3cyc_1S_1V, WriteAdr],   (instregex "ST4i(8|16)_POST$")>;
+def : InstRW<[A57Write_4cyc_4S],           (instregex "ST4i(32)$")>;
+def : InstRW<[A57Write_4cyc_4S, WriteAdr], (instregex "ST4i(32)_POST$")>;
+def : InstRW<[A57Write_3cyc_2S_1V],            (instregex "ST4i(64)$")>;
+def : InstRW<[A57Write_3cyc_2S_1V, WriteAdr],  (instregex "ST4i(64)_POST$")>;
+
+def : InstRW<[A57Write_4cyc_4S_2V],                  (instregex "ST4Fourv(8b|4h|2s)$")>;
+def : InstRW<[A57Write_4cyc_4S_2V, WriteAdr],        (instregex "ST4Fourv(8b|4h|2s)_POST$")>;
+def : InstRW<[A57Write_8cyc_8S_4V],           (instregex "ST4Fourv(16b|8h|4s)$")>;
+def : InstRW<[A57Write_8cyc_8S_4V, WriteAdr], (instregex "ST4Fourv(16b|8h|4s)_POST$")>;
+def : InstRW<[A57Write_8cyc_8S],                (instregex "ST4Fourv(2d)$")>;
+def : InstRW<[A57Write_8cyc_8S, WriteAdr],      (instregex "ST4Fourv(2d)_POST$")>;
+
+} // SchedModel = CortexA57Model
diff --git a/lib/Target/AArch64/AArch64SchedA57WriteRes.td b/lib/Target/AArch64/AArch64SchedA57WriteRes.td
new file mode 100644
index 0000000..a8f421b
--- /dev/null
+++ b/lib/Target/AArch64/AArch64SchedA57WriteRes.td
@@ -0,0 +1,512 @@
+//=- AArch64SchedA57WriteRes.td - ARM Cortex-A57 Write Res ---*- tablegen -*-=//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Contains all of the Cortex-A57 specific SchedWriteRes types. The approach
+// below is to define a generic SchedWriteRes for every combination of
+// latency and microOps. The naming conventions is to use a prefix, one field
+// for latency, and one or more microOp count/type designators.
+//   Prefix: A57Write
+//   Latency: #cyc
+//   MicroOp Count/Types: #(B|I|M|L|S|X|W|V)
+//
+// e.g. A57Write_6cyc_1I_6S_4V means the total latency is 6 and there are
+//      11 micro-ops to be issued down one I pipe, six S pipes and four V pipes.
+//
+//===----------------------------------------------------------------------===//
+
+//===----------------------------------------------------------------------===//
+// Define Generic 1 micro-op types
+
+def A57Write_5cyc_1L  : SchedWriteRes<[A57UnitL]> { let Latency = 5;  }
+def A57Write_5cyc_1M  : SchedWriteRes<[A57UnitM]> { let Latency = 5;  }
+def A57Write_5cyc_1V  : SchedWriteRes<[A57UnitV]> { let Latency = 5;  }
+def A57Write_5cyc_1W  : SchedWriteRes<[A57UnitW]> { let Latency = 5;  }
+def A57Write_10cyc_1V : SchedWriteRes<[A57UnitV]> { let Latency = 10; }
+def A57Write_18cyc_1X : SchedWriteRes<[A57UnitX]> { let Latency = 18; }
+def A57Write_19cyc_1M : SchedWriteRes<[A57UnitM]> { let Latency = 19; }
+def A57Write_1cyc_1B  : SchedWriteRes<[A57UnitB]> { let Latency = 1;  }
+def A57Write_1cyc_1I  : SchedWriteRes<[A57UnitI]> { let Latency = 1;  }
+def A57Write_1cyc_1S  : SchedWriteRes<[A57UnitS]> { let Latency = 1;  }
+def A57Write_2cyc_1M  : SchedWriteRes<[A57UnitM]> { let Latency = 2;  }
+def A57Write_32cyc_1X : SchedWriteRes<[A57UnitX]> { let Latency = 32; }
+def A57Write_35cyc_1M : SchedWriteRes<[A57UnitM]> { let Latency = 35; }
+def A57Write_3cyc_1M  : SchedWriteRes<[A57UnitM]> { let Latency = 3;  }
+def A57Write_3cyc_1V  : SchedWriteRes<[A57UnitV]> { let Latency = 3;  }
+def A57Write_3cyc_1W  : SchedWriteRes<[A57UnitW]> { let Latency = 3;  }
+def A57Write_3cyc_1X  : SchedWriteRes<[A57UnitX]> { let Latency = 3;  }
+def A57Write_4cyc_1L  : SchedWriteRes<[A57UnitL]> { let Latency = 4;  }
+def A57Write_4cyc_1X  : SchedWriteRes<[A57UnitX]> { let Latency = 4;  }
+def A57Write_9cyc_1V  : SchedWriteRes<[A57UnitV]> { let Latency = 9;  }
+def A57Write_6cyc_1M  : SchedWriteRes<[A57UnitM]> { let Latency = 6;  }
+def A57Write_6cyc_1V  : SchedWriteRes<[A57UnitV]> { let Latency = 6;  }
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 2 micro-op types
+
+def A57Write_64cyc_2X    : SchedWriteRes<[A57UnitX, A57UnitX]> {
+  let Latency     = 64;
+  let NumMicroOps = 2;
+}
+def A57Write_6cyc_1I_1L  : SchedWriteRes<[A57UnitI,
+                                          A57UnitL]> {
+  let Latency     = 6;
+  let NumMicroOps = 2;
+}
+def A57Write_7cyc_1V_1X  : SchedWriteRes<[A57UnitV,
+                                          A57UnitX]> {
+  let Latency     = 7;
+  let NumMicroOps = 2;
+}
+def A57Write_8cyc_1L_1V  : SchedWriteRes<[A57UnitL,
+                                          A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 2;
+}
+def A57Write_9cyc_2V     : SchedWriteRes<[A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 2;
+}
+def A57Write_8cyc_2X     : SchedWriteRes<[A57UnitX, A57UnitX]> {
+  let Latency     = 8;
+  let NumMicroOps = 2;
+}
+def A57Write_6cyc_2L     : SchedWriteRes<[A57UnitL, A57UnitL]> {
+  let Latency     = 6;
+  let NumMicroOps = 2;
+}
+def A57Write_6cyc_2V     : SchedWriteRes<[A57UnitV, A57UnitV]> {
+  let Latency     = 6;
+  let NumMicroOps = 2;
+}
+def A57Write_6cyc_2W     : SchedWriteRes<[A57UnitW, A57UnitW]> {
+  let Latency     = 6;
+  let NumMicroOps = 2;
+}
+def A57Write_5cyc_1I_1L  : SchedWriteRes<[A57UnitI,
+                                          A57UnitL]> {
+  let Latency     = 5;
+  let NumMicroOps = 2;
+}
+def A57Write_5cyc_2V     : SchedWriteRes<[A57UnitV, A57UnitV]> {
+  let Latency     = 5;
+  let NumMicroOps = 2;
+}
+def A57Write_5cyc_2X     : SchedWriteRes<[A57UnitX, A57UnitX]> {
+  let Latency     = 5;
+  let NumMicroOps = 2;
+}
+def A57Write_10cyc_1L_1V : SchedWriteRes<[A57UnitL,
+                                          A57UnitV]> {
+  let Latency     = 10;
+  let NumMicroOps = 2;
+}
+def A57Write_10cyc_2V    : SchedWriteRes<[A57UnitV, A57UnitV]> {
+  let Latency     = 10;
+  let NumMicroOps = 2;
+}
+def A57Write_1cyc_1B_1I  : SchedWriteRes<[A57UnitB,
+                                          A57UnitI]> {
+  let Latency     = 1;
+  let NumMicroOps = 2;
+}
+def A57Write_1cyc_1I_1S  : SchedWriteRes<[A57UnitI,
+                                          A57UnitS]> {
+  let Latency     = 1;
+  let NumMicroOps = 2;
+}
+def A57Write_2cyc_1B_1I  : SchedWriteRes<[A57UnitB,
+                                          A57UnitI]> {
+  let Latency     = 2;
+  let NumMicroOps = 2;
+}
+def A57Write_2cyc_2S     : SchedWriteRes<[A57UnitS, A57UnitS]> {
+  let Latency     = 2;
+  let NumMicroOps = 2;
+}
+def A57Write_2cyc_2V     : SchedWriteRes<[A57UnitV, A57UnitV]> {
+  let Latency     = 2;
+  let NumMicroOps = 2;
+}
+def A57Write_36cyc_2X    : SchedWriteRes<[A57UnitX, A57UnitX]> {
+  let Latency     = 36;
+  let NumMicroOps = 2;
+}
+def A57Write_3cyc_1I_1M  : SchedWriteRes<[A57UnitI,
+                                          A57UnitM]> {
+  let Latency     = 3;
+  let NumMicroOps = 2;
+}
+def A57Write_3cyc_1I_1S  : SchedWriteRes<[A57UnitI,
+                                          A57UnitS]> {
+  let Latency     = 3;
+  let NumMicroOps = 2;
+}
+def A57Write_3cyc_1S_1V  : SchedWriteRes<[A57UnitS,
+                                          A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 2;
+}
+def A57Write_4cyc_1I_1L  : SchedWriteRes<[A57UnitI,
+                                          A57UnitL]> {
+  let Latency     = 4;
+  let NumMicroOps = 2;
+}
+def A57Write_4cyc_2X     : SchedWriteRes<[A57UnitX, A57UnitX]> {
+  let Latency     = 4;
+  let NumMicroOps = 2;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 3 micro-op types
+
+def A57Write_10cyc_3V       : SchedWriteRes<[A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 10;
+  let NumMicroOps = 3;
+}
+def A57Write_2cyc_1I_2S     : SchedWriteRes<[A57UnitI,
+                                             A57UnitS, A57UnitS]> {
+  let Latency     = 2;
+  let NumMicroOps = 3;
+}
+def A57Write_3cyc_1I_1S_1V  : SchedWriteRes<[A57UnitI,
+                                             A57UnitS,
+                                             A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 3;
+}
+def A57Write_3cyc_1M_2S     : SchedWriteRes<[A57UnitM,
+                                             A57UnitS, A57UnitS]> {
+  let Latency     = 3;
+  let NumMicroOps = 3;
+}
+def A57Write_3cyc_3S        : SchedWriteRes<[A57UnitS, A57UnitS, A57UnitS]> {
+  let Latency     = 3;
+  let NumMicroOps = 3;
+}
+def A57Write_3cyc_2S_1V     : SchedWriteRes<[A57UnitS, A57UnitS,
+                                             A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 3;
+}
+def A57Write_5cyc_1I_2L     : SchedWriteRes<[A57UnitI,
+                                             A57UnitL, A57UnitL]> {
+  let Latency     = 5;
+  let NumMicroOps = 3;
+}
+def A57Write_6cyc_1I_2L     : SchedWriteRes<[A57UnitI,
+                                             A57UnitL, A57UnitL]> {
+  let Latency     = 6;
+  let NumMicroOps = 3;
+}
+def A57Write_6cyc_3V        : SchedWriteRes<[A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 6;
+  let NumMicroOps = 3;
+}
+def A57Write_7cyc_3L        : SchedWriteRes<[A57UnitL, A57UnitL, A57UnitL]> {
+  let Latency     = 7;
+  let NumMicroOps = 3;
+}
+def A57Write_8cyc_1I_1L_1V  : SchedWriteRes<[A57UnitI,
+                                             A57UnitL,
+                                             A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 3;
+}
+def A57Write_8cyc_1L_2V     : SchedWriteRes<[A57UnitL,
+                                             A57UnitV, A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 3;
+}
+def A57Write_8cyc_3V        : SchedWriteRes<[A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 3;
+}
+def A57Write_9cyc_3V        : SchedWriteRes<[A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 3;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 4 micro-op types
+
+def A57Write_2cyc_2I_2S    : SchedWriteRes<[A57UnitI, A57UnitI,
+                                            A57UnitS, A57UnitS]> {
+  let Latency     = 2;
+  let NumMicroOps = 4;
+}
+def A57Write_3cyc_2I_2S    : SchedWriteRes<[A57UnitI, A57UnitI,
+                                            A57UnitS, A57UnitS]> {
+  let Latency     = 3;
+  let NumMicroOps = 4;
+}
+def A57Write_3cyc_1I_3S    : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS, A57UnitS]> {
+  let Latency     = 3;
+  let NumMicroOps = 4;
+}
+def A57Write_3cyc_1I_2S_1V : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS,
+                                            A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 4;
+}
+def A57Write_4cyc_4S       : SchedWriteRes<[A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS]> {
+  let Latency     = 4;
+  let NumMicroOps = 4;
+}
+def A57Write_7cyc_1I_3L    : SchedWriteRes<[A57UnitI,
+                                            A57UnitL, A57UnitL, A57UnitL]> {
+  let Latency     = 7;
+  let NumMicroOps = 4;
+}
+def A57Write_5cyc_2I_2L    : SchedWriteRes<[A57UnitI, A57UnitI,
+                                            A57UnitL, A57UnitL]> {
+  let Latency     = 5;
+  let NumMicroOps = 4;
+}
+def A57Write_8cyc_1I_1L_2V : SchedWriteRes<[A57UnitI,
+                                            A57UnitL,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 4;
+}
+def A57Write_8cyc_4L       : SchedWriteRes<[A57UnitL, A57UnitL,
+                                            A57UnitL, A57UnitL]> {
+  let Latency     = 8;
+  let NumMicroOps = 4;
+}
+def A57Write_9cyc_2L_2V    : SchedWriteRes<[A57UnitL, A57UnitL,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 4;
+}
+def A57Write_9cyc_1L_3V    : SchedWriteRes<[A57UnitL,
+                                            A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 4;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 5 micro-op types
+
+def A57Write_3cyc_3S_2V    : SchedWriteRes<[A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 5;
+}
+def A57Write_8cyc_1I_4L    : SchedWriteRes<[A57UnitI,
+                                            A57UnitL, A57UnitL,
+                                            A57UnitL, A57UnitL]> {
+  let Latency     = 8;
+  let NumMicroOps = 5;
+}
+def A57Write_4cyc_1I_4S    : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS]> {
+  let Latency     = 4;
+  let NumMicroOps = 5;
+}
+def A57Write_9cyc_1I_2L_2V : SchedWriteRes<[A57UnitI,
+                                            A57UnitL, A57UnitL,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 5;
+}
+def A57Write_9cyc_1I_1L_3V : SchedWriteRes<[A57UnitI,
+                                            A57UnitL,
+                                            A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 5;
+}
+def A57Write_9cyc_2L_3V    : SchedWriteRes<[A57UnitL, A57UnitL,
+                                            A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 5;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 6 micro-op types
+
+def A57Write_3cyc_1I_3S_2V : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 3;
+  let NumMicroOps = 6;
+}
+def A57Write_4cyc_2I_4S    : SchedWriteRes<[A57UnitI, A57UnitI,
+                                            A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS]> {
+  let Latency     = 4;
+  let NumMicroOps = 6;
+}
+def A57Write_4cyc_4S_2V    : SchedWriteRes<[A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 4;
+  let NumMicroOps = 6;
+}
+def A57Write_6cyc_6S       : SchedWriteRes<[A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS, A57UnitS]> {
+  let Latency     = 6;
+  let NumMicroOps = 6;
+}
+def A57Write_9cyc_1I_2L_3V : SchedWriteRes<[A57UnitI,
+                                            A57UnitL, A57UnitL,
+                                            A57UnitV, A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 6;
+}
+def A57Write_9cyc_1I_1L_4V : SchedWriteRes<[A57UnitI,
+                                            A57UnitL,
+                                            A57UnitV, A57UnitV,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 6;
+}
+def A57Write_9cyc_2L_4V    : SchedWriteRes<[A57UnitL, A57UnitL,
+                                            A57UnitV, A57UnitV,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 6;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 7 micro-op types
+
+def A57Write_10cyc_3L_4V : SchedWriteRes<[A57UnitL, A57UnitL, A57UnitL,
+                                          A57UnitV, A57UnitV,
+                                          A57UnitV, A57UnitV]> {
+  let Latency     = 10;
+  let NumMicroOps = 7;
+}
+def A57Write_4cyc_1I_4S_2V  : SchedWriteRes<[A57UnitI,
+                                             A57UnitS, A57UnitS,
+                                             A57UnitS, A57UnitS,
+                                             A57UnitV, A57UnitV]> {
+  let Latency     = 4;
+  let NumMicroOps = 7;
+}
+def A57Write_6cyc_1I_6S  : SchedWriteRes<[A57UnitI,
+                                          A57UnitS, A57UnitS, A57UnitS,
+                                          A57UnitS, A57UnitS, A57UnitS]> {
+  let Latency     = 6;
+  let NumMicroOps = 7;
+}
+def A57Write_9cyc_1I_2L_4V  : SchedWriteRes<[A57UnitI,
+                                             A57UnitL, A57UnitL,
+                                             A57UnitV, A57UnitV,
+                                             A57UnitV, A57UnitV]> {
+  let Latency     = 9;
+  let NumMicroOps = 7;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 8 micro-op types
+
+def A57Write_10cyc_1I_3L_4V : SchedWriteRes<[A57UnitI,
+                                             A57UnitL, A57UnitL, A57UnitL,
+                                             A57UnitV, A57UnitV,
+                                             A57UnitV, A57UnitV]> {
+  let Latency     = 10;
+  let NumMicroOps = 8;
+}
+def A57Write_11cyc_4L_4V : SchedWriteRes<[A57UnitL, A57UnitL,
+                                          A57UnitL, A57UnitL,
+                                          A57UnitV, A57UnitV,
+                                          A57UnitV, A57UnitV]> {
+  let Latency     = 11;
+  let NumMicroOps = 8;
+}
+def A57Write_8cyc_8S  : SchedWriteRes<[A57UnitS, A57UnitS,
+                                       A57UnitS, A57UnitS,
+                                       A57UnitS, A57UnitS,
+                                       A57UnitS, A57UnitS]> {
+  let Latency     = 8;
+  let NumMicroOps = 8;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 9 micro-op types
+
+def A57Write_8cyc_1I_8S  : SchedWriteRes<[A57UnitI,
+                                          A57UnitS, A57UnitS,
+                                          A57UnitS, A57UnitS,
+                                          A57UnitS, A57UnitS,
+                                          A57UnitS, A57UnitS]> {
+  let Latency     = 8;
+  let NumMicroOps = 9;
+}
+def A57Write_11cyc_1I_4L_4V : SchedWriteRes<[A57UnitI,
+                                             A57UnitL, A57UnitL,
+                                             A57UnitL, A57UnitL,
+                                             A57UnitV, A57UnitV,
+                                             A57UnitV, A57UnitV]> {
+  let Latency     = 11;
+  let NumMicroOps = 9;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 10 micro-op types
+
+def A57Write_6cyc_6S_4V : SchedWriteRes<[A57UnitS, A57UnitS, A57UnitS,
+                                         A57UnitS, A57UnitS, A57UnitS,
+                                         A57UnitV, A57UnitV,
+                                         A57UnitV, A57UnitV]> {
+  let Latency     = 6;
+  let NumMicroOps = 10;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 11 micro-op types
+
+def A57Write_6cyc_1I_6S_4V : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitV, A57UnitV,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 6;
+  let NumMicroOps = 11;
+}
+
+
+//===----------------------------------------------------------------------===//
+// Define Generic 12 micro-op types
+
+def A57Write_8cyc_8S_4V : SchedWriteRes<[A57UnitS, A57UnitS, A57UnitS, A57UnitS,
+                                         A57UnitS, A57UnitS, A57UnitS, A57UnitS,
+                                         A57UnitV, A57UnitV,
+                                         A57UnitV, A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 12;
+}
+
+//===----------------------------------------------------------------------===//
+// Define Generic 13 micro-op types
+
+def A57Write_8cyc_1I_8S_4V : SchedWriteRes<[A57UnitI,
+                                            A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS, A57UnitS,
+                                            A57UnitS, A57UnitS,
+                                            A57UnitV, A57UnitV,
+                                            A57UnitV, A57UnitV]> {
+  let Latency     = 8;
+  let NumMicroOps = 13;
+}
+
diff --git a/lib/Target/AArch64/AArch64SelectionDAGInfo.cpp b/lib/Target/AArch64/AArch64SelectionDAGInfo.cpp
index 5c65b75..1bf64fc 100644
--- a/lib/Target/AArch64/AArch64SelectionDAGInfo.cpp
+++ b/lib/Target/AArch64/AArch64SelectionDAGInfo.cpp
@@ -16,9 +16,8 @@ using namespace llvm;
 
 #define DEBUG_TYPE "aarch64-selectiondag-info"
 
-AArch64SelectionDAGInfo::AArch64SelectionDAGInfo(const TargetMachine &TM)
-    : TargetSelectionDAGInfo(TM),
-      Subtarget(&TM.getSubtarget<AArch64Subtarget>()) {}
+AArch64SelectionDAGInfo::AArch64SelectionDAGInfo(const DataLayout *DL)
+    : TargetSelectionDAGInfo(DL) {}
 
 AArch64SelectionDAGInfo::~AArch64SelectionDAGInfo() {}
 
@@ -30,7 +29,9 @@ SDValue AArch64SelectionDAGInfo::EmitTargetCodeForMemset(
   ConstantSDNode *V = dyn_cast<ConstantSDNode>(Src);
   ConstantSDNode *SizeValue = dyn_cast<ConstantSDNode>(Size);
   const char *bzeroEntry =
-      (V && V->isNullValue()) ? Subtarget->getBZeroEntry() : nullptr;
+      (V && V->isNullValue())
+          ? DAG.getTarget().getSubtarget<AArch64Subtarget>().getBZeroEntry()
+          : nullptr;
   // For small size (< 256), it is not beneficial to use bzero
   // instead of memset.
   if (bzeroEntry && (!SizeValue || SizeValue->getZExtValue() > 256)) {
@@ -50,7 +51,7 @@ SDValue AArch64SelectionDAGInfo::EmitTargetCodeForMemset(
     TargetLowering::CallLoweringInfo CLI(DAG);
     CLI.setDebugLoc(dl).setChain(Chain)
       .setCallee(CallingConv::C, Type::getVoidTy(*DAG.getContext()),
-                 DAG.getExternalSymbol(bzeroEntry, IntPtr), &Args, 0)
+                 DAG.getExternalSymbol(bzeroEntry, IntPtr), std::move(Args), 0)
       .setDiscardResult();
     std::pair<SDValue, SDValue> CallResult = TLI.LowerCallTo(CLI);
     return CallResult.second;
diff --git a/lib/Target/AArch64/AArch64SelectionDAGInfo.h b/lib/Target/AArch64/AArch64SelectionDAGInfo.h
index 8381f99..1180eea 100644
--- a/lib/Target/AArch64/AArch64SelectionDAGInfo.h
+++ b/lib/Target/AArch64/AArch64SelectionDAGInfo.h
@@ -19,12 +19,8 @@
 namespace llvm {
 
 class AArch64SelectionDAGInfo : public TargetSelectionDAGInfo {
-  /// Subtarget - Keep a pointer to the ARMSubtarget around so that we can
-  /// make the right decision when generating code for different targets.
-  const AArch64Subtarget *Subtarget;
-
 public:
-  explicit AArch64SelectionDAGInfo(const TargetMachine &TM);
+  explicit AArch64SelectionDAGInfo(const DataLayout *DL);
   ~AArch64SelectionDAGInfo();
 
   SDValue EmitTargetCodeForMemset(SelectionDAG &DAG, SDLoc dl, SDValue Chain,
diff --git a/lib/Target/AArch64/AArch64Subtarget.cpp b/lib/Target/AArch64/AArch64Subtarget.cpp
index cd69994..bb0b72c 100644
--- a/lib/Target/AArch64/AArch64Subtarget.cpp
+++ b/lib/Target/AArch64/AArch64Subtarget.cpp
@@ -30,21 +30,35 @@ static cl::opt<bool>
 EnableEarlyIfConvert("aarch64-early-ifcvt", cl::desc("Enable the early if "
                      "converter pass"), cl::init(true), cl::Hidden);
 
-AArch64Subtarget::AArch64Subtarget(const std::string &TT,
-                                   const std::string &CPU,
-                                   const std::string &FS, bool LittleEndian)
-    : AArch64GenSubtargetInfo(TT, CPU, FS), ARMProcFamily(Others),
-      HasFPARMv8(false), HasNEON(false), HasCrypto(false), HasCRC(false),
-      HasZeroCycleRegMove(false), HasZeroCycleZeroing(false), CPUString(CPU),
-      TargetTriple(TT), IsLittleEndian(LittleEndian) {
+AArch64Subtarget &
+AArch64Subtarget::initializeSubtargetDependencies(StringRef FS) {
   // Determine default and user-specified characteristics
 
   if (CPUString.empty())
     CPUString = "generic";
 
   ParseSubtargetFeatures(CPUString, FS);
+  return *this;
 }
 
+AArch64Subtarget::AArch64Subtarget(const std::string &TT,
+                                   const std::string &CPU,
+                                   const std::string &FS, TargetMachine &TM,
+                                   bool LittleEndian)
+    : AArch64GenSubtargetInfo(TT, CPU, FS), ARMProcFamily(Others),
+      HasFPARMv8(false), HasNEON(false), HasCrypto(false), HasCRC(false),
+      HasZeroCycleRegMove(false), HasZeroCycleZeroing(false), CPUString(CPU),
+      TargetTriple(TT),
+      // This nested ternary is horrible, but DL needs to be properly
+      // initialized
+      // before TLInfo is constructed.
+      DL(isTargetMachO()
+             ? "e-m:o-i64:64-i128:128-n32:64-S128"
+             : (LittleEndian ? "e-m:e-i64:64-i128:128-n32:64-S128"
+                             : "E-m:e-i64:64-i128:128-n32:64-S128")),
+      FrameLowering(), InstrInfo(initializeSubtargetDependencies(FS)),
+      TSInfo(&DL), TLInfo(TM) {}
+
 /// ClassifyGlobalReference - Find the target operand flags that describe
 /// how a global value should be referenced for the current subtarget.
 unsigned char
diff --git a/lib/Target/AArch64/AArch64Subtarget.h b/lib/Target/AArch64/AArch64Subtarget.h
index 590ea05..52124f6 100644
--- a/lib/Target/AArch64/AArch64Subtarget.h
+++ b/lib/Target/AArch64/AArch64Subtarget.h
@@ -14,8 +14,13 @@
 #ifndef AArch64SUBTARGET_H
 #define AArch64SUBTARGET_H
 
-#include "llvm/Target/TargetSubtargetInfo.h"
+#include "AArch64InstrInfo.h"
+#include "AArch64FrameLowering.h"
+#include "AArch64ISelLowering.h"
 #include "AArch64RegisterInfo.h"
+#include "AArch64SelectionDAGInfo.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/Target/TargetSubtargetInfo.h"
 #include <string>
 
 #define GET_SUBTARGETINFO_HEADER
@@ -49,15 +54,32 @@ protected:
   /// TargetTriple - What processor and OS we're targeting.
   Triple TargetTriple;
 
-  /// IsLittleEndian - Is the target little endian?
-  bool IsLittleEndian;
+  const DataLayout DL;
+  AArch64FrameLowering FrameLowering;
+  AArch64InstrInfo InstrInfo;
+  AArch64SelectionDAGInfo TSInfo;
+  AArch64TargetLowering TLInfo;
+private:
+  /// initializeSubtargetDependencies - Initializes using CPUString and the
+  /// passed in feature string so that we can use initializer lists for
+  /// subtarget initialization.
+  AArch64Subtarget &initializeSubtargetDependencies(StringRef FS);
 
 public:
   /// This constructor initializes the data members to match that
   /// of the specified triple.
   AArch64Subtarget(const std::string &TT, const std::string &CPU,
-                 const std::string &FS, bool LittleEndian);
-
+		   const std::string &FS, TargetMachine &TM, bool LittleEndian);
+
+  const AArch64SelectionDAGInfo *getSelectionDAGInfo() const { return &TSInfo; }
+  const AArch64FrameLowering *getFrameLowering() const {
+    return &FrameLowering;
+  }
+  const AArch64TargetLowering *getTargetLowering() const {
+    return &TLInfo;
+  }
+  const AArch64InstrInfo *getInstrInfo() const { return &InstrInfo; }
+  const DataLayout *getDataLayout() const { return &DL; }
   bool enableMachineScheduler() const override { return true; }
 
   bool hasZeroCycleRegMove() const { return HasZeroCycleRegMove; }
@@ -69,7 +91,7 @@ public:
   bool hasCrypto() const { return HasCrypto; }
   bool hasCRC() const { return HasCRC; }
 
-  bool isLittleEndian() const { return IsLittleEndian; }
+  bool isLittleEndian() const { return DL.isLittleEndian(); }
 
   bool isTargetDarwin() const { return TargetTriple.isOSDarwin(); }
 
diff --git a/lib/Target/AArch64/AArch64TargetMachine.cpp b/lib/Target/AArch64/AArch64TargetMachine.cpp
index 0b5dd2f..f99b90b 100644
--- a/lib/Target/AArch64/AArch64TargetMachine.cpp
+++ b/lib/Target/AArch64/AArch64TargetMachine.cpp
@@ -53,6 +53,12 @@ static cl::opt<bool>
 EnableLoadStoreOpt("aarch64-load-store-opt", cl::desc("Enable the load/store pair"
                    " optimization pass"), cl::init(true), cl::Hidden);
 
+static cl::opt<bool>
+EnableAtomicTidy("aarch64-atomic-cfg-tidy", cl::Hidden,
+                 cl::desc("Run SimplifyCFG after expanding atomic operations"
+                          " to make use of cmpxchg flow-based information"),
+                 cl::init(true));
+
 extern "C" void LLVMInitializeAArch64Target() {
   // Register the target.
   RegisterTargetMachine<AArch64leTargetMachine> X(TheAArch64leTarget);
@@ -71,16 +77,7 @@ AArch64TargetMachine::AArch64TargetMachine(const Target &T, StringRef TT,
                                            CodeGenOpt::Level OL,
                                            bool LittleEndian)
     : LLVMTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL),
-      Subtarget(TT, CPU, FS, LittleEndian),
-      // This nested ternary is horrible, but DL needs to be properly
-      // initialized
-      // before TLInfo is constructed.
-      DL(Subtarget.isTargetMachO()
-             ? "e-m:o-i64:64-i128:128-n32:64-S128"
-             : (LittleEndian ? "e-m:e-i64:64-i128:128-n32:64-S128"
-                             : "E-m:e-i64:64-i128:128-n32:64-S128")),
-      InstrInfo(Subtarget), TLInfo(*this), FrameLowering(*this, Subtarget),
-      TSInfo(*this) {
+      Subtarget(TT, CPU, FS, *this, LittleEndian) {
   initAsmInfo();
 }
 
@@ -113,6 +110,7 @@ public:
     return getTM<AArch64TargetMachine>();
   }
 
+  void addIRPasses()  override;
   bool addPreISel() override;
   bool addInstSelector() override;
   bool addILPOpts() override;
@@ -135,6 +133,20 @@ TargetPassConfig *AArch64TargetMachine::createPassConfig(PassManagerBase &PM) {
   return new AArch64PassConfig(this, PM);
 }
 
+void AArch64PassConfig::addIRPasses() {
+  // Always expand atomic operations, we don't deal with atomicrmw or cmpxchg
+  // ourselves.
+  addPass(createAtomicExpandLoadLinkedPass(TM));
+
+  // Cmpxchg instructions are often used with a subsequent comparison to
+  // determine whether it succeeded. We can exploit existing control-flow in
+  // ldrex/strex loops to simplify this, but it needs tidying up.
+  if (TM->getOptLevel() != CodeGenOpt::None && EnableAtomicTidy)
+    addPass(createCFGSimplificationPass());
+
+  TargetPassConfig::addIRPasses();
+}
+
 // Pass Pipeline Configuration
 bool AArch64PassConfig::addPreISel() {
   // Run promote constant before global merge, so that the promoted constants
@@ -146,10 +158,6 @@ bool AArch64PassConfig::addPreISel() {
   if (TM->getOptLevel() != CodeGenOpt::None)
     addPass(createAArch64AddressTypePromotionPass());
 
-  // Always expand atomic operations, we don't deal with atomicrmw or cmpxchg
-  // ourselves.
-  addPass(createAtomicExpandLoadLinkedPass(TM));
-
   return false;
 }
 
diff --git a/lib/Target/AArch64/AArch64TargetMachine.h b/lib/Target/AArch64/AArch64TargetMachine.h
index 079b19b..852cb3f 100644
--- a/lib/Target/AArch64/AArch64TargetMachine.h
+++ b/lib/Target/AArch64/AArch64TargetMachine.h
@@ -15,13 +15,9 @@
 #define AArch64TARGETMACHINE_H
 
 #include "AArch64InstrInfo.h"
-#include "AArch64ISelLowering.h"
 #include "AArch64Subtarget.h"
-#include "AArch64FrameLowering.h"
-#include "AArch64SelectionDAGInfo.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/Target/TargetMachine.h"
-#include "llvm/MC/MCStreamer.h"
 
 namespace llvm {
 
@@ -29,13 +25,6 @@ class AArch64TargetMachine : public LLVMTargetMachine {
 protected:
   AArch64Subtarget Subtarget;
 
-private:
-  const DataLayout DL;
-  AArch64InstrInfo InstrInfo;
-  AArch64TargetLowering TLInfo;
-  AArch64FrameLowering FrameLowering;
-  AArch64SelectionDAGInfo TSInfo;
-
 public:
   AArch64TargetMachine(const Target &T, StringRef TT, StringRef CPU,
                        StringRef FS, const TargetOptions &Options,
@@ -46,18 +35,22 @@ public:
     return &Subtarget;
   }
   const AArch64TargetLowering *getTargetLowering() const override {
-    return &TLInfo;
+    return getSubtargetImpl()->getTargetLowering();
+  }
+  const DataLayout *getDataLayout() const override {
+    return getSubtargetImpl()->getDataLayout();
   }
-  const DataLayout *getDataLayout() const override { return &DL; }
   const AArch64FrameLowering *getFrameLowering() const override {
-    return &FrameLowering;
+    return getSubtargetImpl()->getFrameLowering();
+  }
+  const AArch64InstrInfo *getInstrInfo() const override {
+    return getSubtargetImpl()->getInstrInfo();
   }
-  const AArch64InstrInfo *getInstrInfo() const override { return &InstrInfo; }
   const AArch64RegisterInfo *getRegisterInfo() const override {
-    return &InstrInfo.getRegisterInfo();
+    return &getInstrInfo()->getRegisterInfo();
   }
   const AArch64SelectionDAGInfo *getSelectionDAGInfo() const override {
-    return &TSInfo;
+    return getSubtargetImpl()->getSelectionDAGInfo();
   }
 
   // Pass Pipeline Configuration
diff --git a/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 33e482a..1dac14b 100644
--- a/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -306,28 +306,64 @@ unsigned AArch64TTI::getCastInstrCost(unsigned Opcode, Type *Dst,
   static const TypeConversionCostTblEntry<MVT> ConversionTbl[] = {
     // LowerVectorINT_TO_FP:
     { ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i32, 1 },
-    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i8, 1 },
-    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i16, 1 },
-    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i32, 1 },
+    { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i32, 1 },
     { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i64, 1 },
     { ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i32, 1 },
-    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i8, 1 },
-    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i16, 1 },
-    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i32, 1 },
+    { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i32, 1 },
     { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i64, 1 },
+
+    // Complex: to v2f32
+    { ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i8,  3 },
+    { ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i16, 3 },
+    { ISD::SINT_TO_FP, MVT::v2f32, MVT::v2i64, 2 },
+    { ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i8,  3 },
+    { ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i16, 3 },
+    { ISD::UINT_TO_FP, MVT::v2f32, MVT::v2i64, 2 },
+
+    // Complex: to v4f32
+    { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i8,  4 },
+    { ISD::SINT_TO_FP, MVT::v4f32, MVT::v4i16, 2 },
+    { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i8,  3 },
+    { ISD::UINT_TO_FP, MVT::v4f32, MVT::v4i16, 2 },
+
+    // Complex: to v2f64
+    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i8,  4 },
+    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i16, 4 },
+    { ISD::SINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
+    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i8,  4 },
+    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i16, 4 },
+    { ISD::UINT_TO_FP, MVT::v2f64, MVT::v2i32, 2 },
+
+
     // LowerVectorFP_TO_INT
+    { ISD::FP_TO_SINT, MVT::v2i32, MVT::v2f32, 1 },
     { ISD::FP_TO_SINT, MVT::v4i32, MVT::v4f32, 1 },
     { ISD::FP_TO_SINT, MVT::v2i64, MVT::v2f64, 1 },
+    { ISD::FP_TO_UINT, MVT::v2i32, MVT::v2f32, 1 },
     { ISD::FP_TO_UINT, MVT::v4i32, MVT::v4f32, 1 },
     { ISD::FP_TO_UINT, MVT::v2i64, MVT::v2f64, 1 },
-    { ISD::FP_TO_UINT, MVT::v2i32, MVT::v2f64, 1 },
-    { ISD::FP_TO_SINT, MVT::v2i32, MVT::v2f64, 1 },
-    { ISD::FP_TO_UINT, MVT::v2i64, MVT::v2f32, 4 },
-    { ISD::FP_TO_SINT, MVT::v2i64, MVT::v2f32, 4 },
-    { ISD::FP_TO_UINT, MVT::v4i16, MVT::v4f32, 4 },
-    { ISD::FP_TO_SINT, MVT::v4i16, MVT::v4f32, 4 },
-    { ISD::FP_TO_UINT, MVT::v2i64, MVT::v2f64, 4 },
-    { ISD::FP_TO_SINT, MVT::v2i64, MVT::v2f64, 4 },
+
+    // Complex, from v2f32: legal type is v2i32 (no cost) or v2i64 (1 ext).
+    { ISD::FP_TO_SINT, MVT::v2i64, MVT::v2f32, 2 },
+    { ISD::FP_TO_SINT, MVT::v2i16, MVT::v2f32, 1 },
+    { ISD::FP_TO_SINT, MVT::v2i8,  MVT::v2f32, 1 },
+    { ISD::FP_TO_UINT, MVT::v2i64, MVT::v2f32, 2 },
+    { ISD::FP_TO_UINT, MVT::v2i16, MVT::v2f32, 1 },
+    { ISD::FP_TO_UINT, MVT::v2i8,  MVT::v2f32, 1 },
+
+    // Complex, from v4f32: legal type is v4i16, 1 narrowing => ~2
+    { ISD::FP_TO_SINT, MVT::v4i16, MVT::v4f32, 2 },
+    { ISD::FP_TO_SINT, MVT::v4i8,  MVT::v4f32, 2 },
+    { ISD::FP_TO_UINT, MVT::v4i16, MVT::v4f32, 2 },
+    { ISD::FP_TO_UINT, MVT::v4i8,  MVT::v4f32, 2 },
+
+    // Complex, from v2f64: legal type is v2i32, 1 narrowing => ~2.
+    { ISD::FP_TO_SINT, MVT::v2i32, MVT::v2f64, 2 },
+    { ISD::FP_TO_SINT, MVT::v2i16, MVT::v2f64, 2 },
+    { ISD::FP_TO_SINT, MVT::v2i8,  MVT::v2f64, 2 },
+    { ISD::FP_TO_UINT, MVT::v2i32, MVT::v2f64, 2 },
+    { ISD::FP_TO_UINT, MVT::v2i16, MVT::v2f64, 2 },
+    { ISD::FP_TO_UINT, MVT::v2i8,  MVT::v2f64, 2 },
   };
 
   int Idx = ConvertCostTableLookup<MVT>(
diff --git a/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp b/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
index 65b77c5..c42d11e 100644
--- a/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
+++ b/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
@@ -38,14 +38,19 @@ namespace {
 class AArch64Operand;
 
 class AArch64AsmParser : public MCTargetAsmParser {
-public:
-  typedef SmallVectorImpl<MCParsedAsmOperand *> OperandVector;
-
 private:
   StringRef Mnemonic; ///< Instruction mnemonic.
   MCSubtargetInfo &STI;
   MCAsmParser &Parser;
 
+  // Map of register aliases registers via the .req directive.
+  StringMap<std::pair<bool, unsigned> > RegisterReqs;
+
+  AArch64TargetStreamer &getTargetStreamer() {
+    MCTargetStreamer &TS = *getParser().getStreamer().getTargetStreamer();
+    return static_cast<AArch64TargetStreamer &>(TS);
+  }
+
   MCAsmParser &getParser() const { return Parser; }
   MCAsmLexer &getLexer() const { return Parser.getLexer(); }
 
@@ -54,6 +59,7 @@ private:
   bool parseSysAlias(StringRef Name, SMLoc NameLoc, OperandVector &Operands);
   AArch64CC::CondCode parseCondCodeString(StringRef Cond);
   bool parseCondCode(OperandVector &Operands, bool invertCondCode);
+  unsigned matchRegisterNameAlias(StringRef Name, bool isVector);
   int tryParseRegister();
   int tryMatchVectorRegister(StringRef &Kind, bool expected);
   bool parseRegister(OperandVector &Operands);
@@ -70,6 +76,10 @@ private:
   bool parseDirectiveTLSDescCall(SMLoc L);
 
   bool parseDirectiveLOH(StringRef LOH, SMLoc L);
+  bool parseDirectiveLtorg(SMLoc L);
+
+  bool parseDirectiveReq(StringRef Name, SMLoc L);
+  bool parseDirectiveUnreq(SMLoc L);
 
   bool validateInstruction(MCInst &Inst, SmallVectorImpl<SMLoc> &Loc);
   bool MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
@@ -108,6 +118,8 @@ public:
                  const MCTargetOptions &Options)
       : MCTargetAsmParser(), STI(_STI), Parser(_Parser) {
     MCAsmParserExtension::Initialize(_Parser);
+    if (Parser.getStreamer().getTargetStreamer() == nullptr)
+      new AArch64TargetStreamer(Parser.getStreamer());
 
     // Initialize the set of available features.
     setAvailableFeatures(ComputeAvailableFeatures(STI.getFeatureBits()));
@@ -117,7 +129,7 @@ public:
                         SMLoc NameLoc, OperandVector &Operands) override;
   bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc) override;
   bool ParseDirective(AsmToken DirectiveID) override;
-  unsigned validateTargetOperandClass(MCParsedAsmOperand *Op,
+  unsigned validateTargetOperandClass(MCParsedAsmOperand &Op,
                                       unsigned Kind) override;
 
   static bool classifySymbolRef(const MCExpr *Expr,
@@ -240,10 +252,10 @@ private:
   // the add<>Operands() calls.
   MCContext &Ctx;
 
+public:
   AArch64Operand(KindTy K, MCContext &_Ctx)
       : MCParsedAsmOperand(), Kind(K), Ctx(_Ctx) {}
 
-public:
   AArch64Operand(const AArch64Operand &o) : MCParsedAsmOperand(), Ctx(o.Ctx) {
     Kind = o.Kind;
     StartLoc = o.StartLoc;
@@ -607,7 +619,11 @@ public:
     const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(getImm());
     if (!MCE)
       return false;
-    return AArch64_AM::isLogicalImmediate(MCE->getValue(), 32);
+    int64_t Val = MCE->getValue();
+    if (Val >> 32 != 0 && Val >> 32 != ~0LL)
+      return false;
+    Val &= 0xFFFFFFFF;
+    return AArch64_AM::isLogicalImmediate(Val, 32);
   }
   bool isLogicalImm64() const {
     if (!isImm())
@@ -617,6 +633,23 @@ public:
       return false;
     return AArch64_AM::isLogicalImmediate(MCE->getValue(), 64);
   }
+  bool isLogicalImm32Not() const {
+    if (!isImm())
+      return false;
+    const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(getImm());
+    if (!MCE)
+      return false;
+    int64_t Val = ~MCE->getValue() & 0xFFFFFFFF;
+    return AArch64_AM::isLogicalImmediate(Val, 32);
+  }
+  bool isLogicalImm64Not() const {
+    if (!isImm())
+      return false;
+    const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(getImm());
+    if (!MCE)
+      return false;
+    return AArch64_AM::isLogicalImmediate(~MCE->getValue(), 64);
+  }
   bool isShiftedImm() const { return Kind == k_ShiftedImm; }
   bool isAddSubImm() const {
     if (!isShiftedImm() && !isImm())
@@ -1348,7 +1381,8 @@ public:
     assert(N == 1 && "Invalid number of operands!");
     const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(getImm());
     assert(MCE && "Invalid logical immediate operand!");
-    uint64_t encoding = AArch64_AM::encodeLogicalImmediate(MCE->getValue(), 32);
+    uint64_t encoding =
+        AArch64_AM::encodeLogicalImmediate(MCE->getValue() & 0xFFFFFFFF, 32);
     Inst.addOperand(MCOperand::CreateImm(encoding));
   }
 
@@ -1360,6 +1394,22 @@ public:
     Inst.addOperand(MCOperand::CreateImm(encoding));
   }
 
+  void addLogicalImm32NotOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    const MCConstantExpr *MCE = cast<MCConstantExpr>(getImm());
+    int64_t Val = ~MCE->getValue() & 0xFFFFFFFF;
+    uint64_t encoding = AArch64_AM::encodeLogicalImmediate(Val, 32);
+    Inst.addOperand(MCOperand::CreateImm(encoding));
+  }
+
+  void addLogicalImm64NotOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    const MCConstantExpr *MCE = cast<MCConstantExpr>(getImm());
+    uint64_t encoding =
+        AArch64_AM::encodeLogicalImmediate(~MCE->getValue(), 64);
+    Inst.addOperand(MCOperand::CreateImm(encoding));
+  }
+
   void addSIMDImmType10Operands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(getImm());
@@ -1523,9 +1573,9 @@ public:
 
   void print(raw_ostream &OS) const override;
 
-  static AArch64Operand *CreateToken(StringRef Str, bool IsSuffix, SMLoc S,
-                                   MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_Token, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateToken(StringRef Str, bool IsSuffix, SMLoc S, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_Token, Ctx);
     Op->Tok.Data = Str.data();
     Op->Tok.Length = Str.size();
     Op->Tok.IsSuffix = IsSuffix;
@@ -1534,9 +1584,9 @@ public:
     return Op;
   }
 
-  static AArch64Operand *CreateReg(unsigned RegNum, bool isVector, SMLoc S,
-                                 SMLoc E, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_Register, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateReg(unsigned RegNum, bool isVector, SMLoc S, SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_Register, Ctx);
     Op->Reg.RegNum = RegNum;
     Op->Reg.isVector = isVector;
     Op->StartLoc = S;
@@ -1544,10 +1594,10 @@ public:
     return Op;
   }
 
-  static AArch64Operand *CreateVectorList(unsigned RegNum, unsigned Count,
-                                        unsigned NumElements, char ElementKind,
-                                        SMLoc S, SMLoc E, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_VectorList, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateVectorList(unsigned RegNum, unsigned Count, unsigned NumElements,
+                   char ElementKind, SMLoc S, SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_VectorList, Ctx);
     Op->VectorList.RegNum = RegNum;
     Op->VectorList.Count = Count;
     Op->VectorList.NumElements = NumElements;
@@ -1557,28 +1607,29 @@ public:
     return Op;
   }
 
-  static AArch64Operand *CreateVectorIndex(unsigned Idx, SMLoc S, SMLoc E,
-                                         MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_VectorIndex, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateVectorIndex(unsigned Idx, SMLoc S, SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_VectorIndex, Ctx);
     Op->VectorIndex.Val = Idx;
     Op->StartLoc = S;
     Op->EndLoc = E;
     return Op;
   }
 
-  static AArch64Operand *CreateImm(const MCExpr *Val, SMLoc S, SMLoc E,
-                                 MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_Immediate, Ctx);
+  static std::unique_ptr<AArch64Operand> CreateImm(const MCExpr *Val, SMLoc S,
+                                                   SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_Immediate, Ctx);
     Op->Imm.Val = Val;
     Op->StartLoc = S;
     Op->EndLoc = E;
     return Op;
   }
 
-  static AArch64Operand *CreateShiftedImm(const MCExpr *Val,
-                                          unsigned ShiftAmount, SMLoc S,
-                                          SMLoc E, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_ShiftedImm, Ctx);
+  static std::unique_ptr<AArch64Operand> CreateShiftedImm(const MCExpr *Val,
+                                                          unsigned ShiftAmount,
+                                                          SMLoc S, SMLoc E,
+                                                          MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_ShiftedImm, Ctx);
     Op->ShiftedImm .Val = Val;
     Op->ShiftedImm.ShiftAmount = ShiftAmount;
     Op->StartLoc = S;
@@ -1586,34 +1637,36 @@ public:
     return Op;
   }
 
-  static AArch64Operand *CreateCondCode(AArch64CC::CondCode Code, SMLoc S,
-                                        SMLoc E, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_CondCode, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateCondCode(AArch64CC::CondCode Code, SMLoc S, SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_CondCode, Ctx);
     Op->CondCode.Code = Code;
     Op->StartLoc = S;
     Op->EndLoc = E;
     return Op;
   }
 
-  static AArch64Operand *CreateFPImm(unsigned Val, SMLoc S, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_FPImm, Ctx);
+  static std::unique_ptr<AArch64Operand> CreateFPImm(unsigned Val, SMLoc S,
+                                                     MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_FPImm, Ctx);
     Op->FPImm.Val = Val;
     Op->StartLoc = S;
     Op->EndLoc = S;
     return Op;
   }
 
-  static AArch64Operand *CreateBarrier(unsigned Val, SMLoc S, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_Barrier, Ctx);
+  static std::unique_ptr<AArch64Operand> CreateBarrier(unsigned Val, SMLoc S,
+                                                       MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_Barrier, Ctx);
     Op->Barrier.Val = Val;
     Op->StartLoc = S;
     Op->EndLoc = S;
     return Op;
   }
 
-  static AArch64Operand *CreateSysReg(StringRef Str, SMLoc S,
-                                    uint64_t FeatureBits, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_SysReg, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateSysReg(StringRef Str, SMLoc S, uint64_t FeatureBits, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_SysReg, Ctx);
     Op->SysReg.Data = Str.data();
     Op->SysReg.Length = Str.size();
     Op->SysReg.FeatureBits = FeatureBits;
@@ -1622,27 +1675,28 @@ public:
     return Op;
   }
 
-  static AArch64Operand *CreateSysCR(unsigned Val, SMLoc S, SMLoc E,
-                                   MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_SysCR, Ctx);
+  static std::unique_ptr<AArch64Operand> CreateSysCR(unsigned Val, SMLoc S,
+                                                     SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_SysCR, Ctx);
     Op->SysCRImm.Val = Val;
     Op->StartLoc = S;
     Op->EndLoc = E;
     return Op;
   }
 
-  static AArch64Operand *CreatePrefetch(unsigned Val, SMLoc S, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_Prefetch, Ctx);
+  static std::unique_ptr<AArch64Operand> CreatePrefetch(unsigned Val, SMLoc S,
+                                                        MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_Prefetch, Ctx);
     Op->Prefetch.Val = Val;
     Op->StartLoc = S;
     Op->EndLoc = S;
     return Op;
   }
 
-  static AArch64Operand *CreateShiftExtend(AArch64_AM::ShiftExtendType ShOp,
-                                         unsigned Val, bool HasExplicitAmount,
-                                         SMLoc S, SMLoc E, MCContext &Ctx) {
-    AArch64Operand *Op = new AArch64Operand(k_ShiftExtend, Ctx);
+  static std::unique_ptr<AArch64Operand>
+  CreateShiftExtend(AArch64_AM::ShiftExtendType ShOp, unsigned Val,
+                    bool HasExplicitAmount, SMLoc S, SMLoc E, MCContext &Ctx) {
+    auto Op = make_unique<AArch64Operand>(k_ShiftExtend, Ctx);
     Op->ShiftExtend.Type = ShOp;
     Op->ShiftExtend.Amount = Val;
     Op->ShiftExtend.HasExplicitAmount = HasExplicitAmount;
@@ -1816,6 +1870,26 @@ bool AArch64AsmParser::ParseRegister(unsigned &RegNo, SMLoc &StartLoc,
   return (RegNo == (unsigned)-1);
 }
 
+// Matches a register name or register alias previously defined by '.req'
+unsigned AArch64AsmParser::matchRegisterNameAlias(StringRef Name,
+                                                  bool isVector) {
+  unsigned RegNum = isVector ? matchVectorRegName(Name)
+                             : MatchRegisterName(Name);
+
+  if (RegNum == 0) {
+    // Check for aliases registered via .req. Canonicalize to lower case.
+    // That's more consistent since register names are case insensitive, and
+    // it's how the original entry was passed in from MC/MCParser/AsmParser.
+    auto Entry = RegisterReqs.find(Name.lower());
+    if (Entry == RegisterReqs.end())
+      return 0;
+    // set RegNum if the match is the right kind of register
+    if (isVector == Entry->getValue().first)
+      RegNum = Entry->getValue().second;
+  }
+  return RegNum;
+}
+
 /// tryParseRegister - Try to parse a register name. The token must be an
 /// Identifier when called, and if it is a register name the token is eaten and
 /// the register is added to the operand list.
@@ -1824,7 +1898,7 @@ int AArch64AsmParser::tryParseRegister() {
   assert(Tok.is(AsmToken::Identifier) && "Token is not an Identifier");
 
   std::string lowerCase = Tok.getString().lower();
-  unsigned RegNum = MatchRegisterName(lowerCase);
+  unsigned RegNum = matchRegisterNameAlias(lowerCase, false);
   // Also handle a few aliases of registers.
   if (RegNum == 0)
     RegNum = StringSwitch<unsigned>(lowerCase)
@@ -1854,7 +1928,8 @@ int AArch64AsmParser::tryMatchVectorRegister(StringRef &Kind, bool expected) {
   // a '.'.
   size_t Start = 0, Next = Name.find('.');
   StringRef Head = Name.slice(Start, Next);
-  unsigned RegNum = matchVectorRegName(Head);
+  unsigned RegNum = matchRegisterNameAlias(Head, true);
+
   if (RegNum) {
     if (Next != StringRef::npos) {
       Kind = Name.slice(Next, StringRef::npos);
@@ -2183,8 +2258,11 @@ bool AArch64AsmParser::parseCondCode(OperandVector &Operands,
     return TokError("invalid condition code");
   Parser.Lex(); // Eat identifier token.
 
-  if (invertCondCode)
+  if (invertCondCode) {
+    if (CC == AArch64CC::AL || CC == AArch64CC::NV)
+      return TokError("condition codes AL and NV are invalid for this instruction");
     CC = AArch64CC::getInvertedCondCode(AArch64CC::CondCode(CC));
+  }
 
   Operands.push_back(
       AArch64Operand::CreateCondCode(CC, S, getLoc(), getContext()));
@@ -2849,7 +2927,7 @@ AArch64AsmParser::tryParseGPR64sp0Operand(OperandVector &Operands) {
   if (!Tok.is(AsmToken::Identifier))
     return MatchOperand_NoMatch;
 
-  unsigned RegNum = MatchRegisterName(Tok.getString().lower());
+  unsigned RegNum = matchRegisterNameAlias(Tok.getString().lower(), false);
 
   MCContext &Ctx = getContext();
   const MCRegisterInfo *RI = Ctx.getRegisterInfo();
@@ -3000,6 +3078,43 @@ bool AArch64AsmParser::parseOperand(OperandVector &Operands, bool isCondCode,
     Operands.push_back(AArch64Operand::CreateImm(ImmVal, S, E, getContext()));
     return false;
   }
+  case AsmToken::Equal: {
+    SMLoc Loc = Parser.getTok().getLoc();
+    if (Mnemonic != "ldr") // only parse for ldr pseudo (e.g. ldr r0, =val)
+      return Error(Loc, "unexpected token in operand");
+    Parser.Lex(); // Eat '='
+    const MCExpr *SubExprVal;
+    if (getParser().parseExpression(SubExprVal))
+      return true;
+
+    MCContext& Ctx = getContext();
+    E = SMLoc::getFromPointer(Loc.getPointer() - 1);
+    // If the op is an imm and can be fit into a mov, then replace ldr with mov.
+    if (isa<MCConstantExpr>(SubExprVal) && Operands.size() >= 2 &&
+        static_cast<AArch64Operand &>(*Operands[1]).isReg()) {
+      bool IsXReg =  AArch64MCRegisterClasses[AArch64::GPR64allRegClassID].contains(
+            Operands[1]->getReg());
+      uint64_t Imm = (cast<MCConstantExpr>(SubExprVal))->getValue();
+      uint32_t ShiftAmt = 0, MaxShiftAmt = IsXReg ? 48 : 16;
+      while(Imm > 0xFFFF && countTrailingZeros(Imm) >= 16) {
+        ShiftAmt += 16;
+        Imm >>= 16;
+      }
+      if (ShiftAmt <= MaxShiftAmt && Imm <= 0xFFFF) {
+          Operands[0] = AArch64Operand::CreateToken("movz", false, Loc, Ctx);
+          Operands.push_back(AArch64Operand::CreateImm(
+                     MCConstantExpr::Create(Imm, Ctx), S, E, Ctx));
+        if (ShiftAmt)
+          Operands.push_back(AArch64Operand::CreateShiftExtend(AArch64_AM::LSL,
+                     ShiftAmt, true, S, E, Ctx));
+        return false;
+      }
+    }
+    // If it is a label or an imm that cannot fit in a movz, put it into CP.
+    const MCExpr *CPLoc = getTargetStreamer().addConstantPoolEntry(SubExprVal);
+    Operands.push_back(AArch64Operand::CreateImm(CPLoc, S, E, Ctx));
+    return false;
+  }
   }
 }
 
@@ -3029,6 +3144,15 @@ bool AArch64AsmParser::ParseInstruction(ParseInstructionInfo &Info,
              .Case("bnv", "b.nv")
              .Default(Name);
 
+  // First check for the AArch64-specific .req directive.
+  if (Parser.getTok().is(AsmToken::Identifier) &&
+      Parser.getTok().getIdentifier() == ".req") {
+    parseDirectiveReq(Name, NameLoc);
+    // We always return 'error' for this, as we're done with this
+    // statement and don't need to match the 'instruction."
+    return true;
+  }
+
   // Create the leading tokens for the mnemonic, split by '.' characters.
   size_t Start = 0, Next = Name.find('.');
   StringRef Head = Name.slice(Start, Next);
@@ -3443,8 +3567,7 @@ bool AArch64AsmParser::showMatchError(SMLoc Loc, unsigned ErrCode) {
   case Match_MnemonicFail:
     return Error(Loc, "unrecognized instruction mnemonic");
   default:
-    assert(0 && "unexpected error code!");
-    return Error(Loc, "invalid instruction format");
+    llvm_unreachable("unexpected error code!");
   }
 }
 
@@ -3456,23 +3579,23 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
                                                unsigned &ErrorInfo,
                                                bool MatchingInlineAsm) {
   assert(!Operands.empty() && "Unexpect empty operand list!");
-  AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[0]);
-  assert(Op->isToken() && "Leading operand should always be a mnemonic!");
+  AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[0]);
+  assert(Op.isToken() && "Leading operand should always be a mnemonic!");
 
-  StringRef Tok = Op->getToken();
+  StringRef Tok = Op.getToken();
   unsigned NumOperands = Operands.size();
 
   if (NumOperands == 4 && Tok == "lsl") {
-    AArch64Operand *Op2 = static_cast<AArch64Operand *>(Operands[2]);
-    AArch64Operand *Op3 = static_cast<AArch64Operand *>(Operands[3]);
-    if (Op2->isReg() && Op3->isImm()) {
-      const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3->getImm());
+    AArch64Operand &Op2 = static_cast<AArch64Operand &>(*Operands[2]);
+    AArch64Operand &Op3 = static_cast<AArch64Operand &>(*Operands[3]);
+    if (Op2.isReg() && Op3.isImm()) {
+      const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3.getImm());
       if (Op3CE) {
         uint64_t Op3Val = Op3CE->getValue();
         uint64_t NewOp3Val = 0;
         uint64_t NewOp4Val = 0;
         if (AArch64MCRegisterClasses[AArch64::GPR32allRegClassID].contains(
-                Op2->getReg())) {
+                Op2.getReg())) {
           NewOp3Val = (32 - Op3Val) & 0x1f;
           NewOp4Val = 31 - Op3Val;
         } else {
@@ -3484,26 +3607,24 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
         const MCExpr *NewOp4 = MCConstantExpr::Create(NewOp4Val, getContext());
 
         Operands[0] = AArch64Operand::CreateToken(
-            "ubfm", false, Op->getStartLoc(), getContext());
-        Operands[3] = AArch64Operand::CreateImm(NewOp3, Op3->getStartLoc(),
-                                                Op3->getEndLoc(), getContext());
+            "ubfm", false, Op.getStartLoc(), getContext());
         Operands.push_back(AArch64Operand::CreateImm(
-            NewOp4, Op3->getStartLoc(), Op3->getEndLoc(), getContext()));
-        delete Op3;
-        delete Op;
+            NewOp4, Op3.getStartLoc(), Op3.getEndLoc(), getContext()));
+        Operands[3] = AArch64Operand::CreateImm(NewOp3, Op3.getStartLoc(),
+                                                Op3.getEndLoc(), getContext());
       }
     }
   } else if (NumOperands == 5) {
     // FIXME: Horrible hack to handle the BFI -> BFM, SBFIZ->SBFM, and
     // UBFIZ -> UBFM aliases.
     if (Tok == "bfi" || Tok == "sbfiz" || Tok == "ubfiz") {
-      AArch64Operand *Op1 = static_cast<AArch64Operand *>(Operands[1]);
-      AArch64Operand *Op3 = static_cast<AArch64Operand *>(Operands[3]);
-      AArch64Operand *Op4 = static_cast<AArch64Operand *>(Operands[4]);
+      AArch64Operand &Op1 = static_cast<AArch64Operand &>(*Operands[1]);
+      AArch64Operand &Op3 = static_cast<AArch64Operand &>(*Operands[3]);
+      AArch64Operand &Op4 = static_cast<AArch64Operand &>(*Operands[4]);
 
-      if (Op1->isReg() && Op3->isImm() && Op4->isImm()) {
-        const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3->getImm());
-        const MCConstantExpr *Op4CE = dyn_cast<MCConstantExpr>(Op4->getImm());
+      if (Op1.isReg() && Op3.isImm() && Op4.isImm()) {
+        const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3.getImm());
+        const MCConstantExpr *Op4CE = dyn_cast<MCConstantExpr>(Op4.getImm());
 
         if (Op3CE && Op4CE) {
           uint64_t Op3Val = Op3CE->getValue();
@@ -3511,21 +3632,21 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
 
           uint64_t RegWidth = 0;
           if (AArch64MCRegisterClasses[AArch64::GPR64allRegClassID].contains(
-              Op1->getReg()))
+                  Op1.getReg()))
             RegWidth = 64;
           else
             RegWidth = 32;
 
           if (Op3Val >= RegWidth)
-            return Error(Op3->getStartLoc(),
+            return Error(Op3.getStartLoc(),
                          "expected integer in range [0, 31]");
           if (Op4Val < 1 || Op4Val > RegWidth)
-            return Error(Op4->getStartLoc(),
+            return Error(Op4.getStartLoc(),
                          "expected integer in range [1, 32]");
 
           uint64_t NewOp3Val = 0;
           if (AArch64MCRegisterClasses[AArch64::GPR32allRegClassID].contains(
-                  Op1->getReg()))
+                  Op1.getReg()))
             NewOp3Val = (32 - Op3Val) & 0x1f;
           else
             NewOp3Val = (64 - Op3Val) & 0x3f;
@@ -3533,7 +3654,7 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
           uint64_t NewOp4Val = Op4Val - 1;
 
           if (NewOp3Val != 0 && NewOp4Val >= NewOp3Val)
-            return Error(Op4->getStartLoc(),
+            return Error(Op4.getStartLoc(),
                          "requested insert overflows register");
 
           const MCExpr *NewOp3 =
@@ -3541,24 +3662,20 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
           const MCExpr *NewOp4 =
               MCConstantExpr::Create(NewOp4Val, getContext());
           Operands[3] = AArch64Operand::CreateImm(
-              NewOp3, Op3->getStartLoc(), Op3->getEndLoc(), getContext());
+              NewOp3, Op3.getStartLoc(), Op3.getEndLoc(), getContext());
           Operands[4] = AArch64Operand::CreateImm(
-              NewOp4, Op4->getStartLoc(), Op4->getEndLoc(), getContext());
+              NewOp4, Op4.getStartLoc(), Op4.getEndLoc(), getContext());
           if (Tok == "bfi")
             Operands[0] = AArch64Operand::CreateToken(
-                "bfm", false, Op->getStartLoc(), getContext());
+                "bfm", false, Op.getStartLoc(), getContext());
           else if (Tok == "sbfiz")
             Operands[0] = AArch64Operand::CreateToken(
-                "sbfm", false, Op->getStartLoc(), getContext());
+                "sbfm", false, Op.getStartLoc(), getContext());
           else if (Tok == "ubfiz")
             Operands[0] = AArch64Operand::CreateToken(
-                "ubfm", false, Op->getStartLoc(), getContext());
+                "ubfm", false, Op.getStartLoc(), getContext());
           else
             llvm_unreachable("No valid mnemonic for alias?");
-
-          delete Op;
-          delete Op3;
-          delete Op4;
         }
       }
 
@@ -3566,13 +3683,13 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
       // UBFX -> UBFM aliases.
     } else if (NumOperands == 5 &&
                (Tok == "bfxil" || Tok == "sbfx" || Tok == "ubfx")) {
-      AArch64Operand *Op1 = static_cast<AArch64Operand *>(Operands[1]);
-      AArch64Operand *Op3 = static_cast<AArch64Operand *>(Operands[3]);
-      AArch64Operand *Op4 = static_cast<AArch64Operand *>(Operands[4]);
+      AArch64Operand &Op1 = static_cast<AArch64Operand &>(*Operands[1]);
+      AArch64Operand &Op3 = static_cast<AArch64Operand &>(*Operands[3]);
+      AArch64Operand &Op4 = static_cast<AArch64Operand &>(*Operands[4]);
 
-      if (Op1->isReg() && Op3->isImm() && Op4->isImm()) {
-        const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3->getImm());
-        const MCConstantExpr *Op4CE = dyn_cast<MCConstantExpr>(Op4->getImm());
+      if (Op1.isReg() && Op3.isImm() && Op4.isImm()) {
+        const MCConstantExpr *Op3CE = dyn_cast<MCConstantExpr>(Op3.getImm());
+        const MCConstantExpr *Op4CE = dyn_cast<MCConstantExpr>(Op4.getImm());
 
         if (Op3CE && Op4CE) {
           uint64_t Op3Val = Op3CE->getValue();
@@ -3580,42 +3697,39 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
 
           uint64_t RegWidth = 0;
           if (AArch64MCRegisterClasses[AArch64::GPR64allRegClassID].contains(
-              Op1->getReg()))
+                  Op1.getReg()))
             RegWidth = 64;
           else
             RegWidth = 32;
 
           if (Op3Val >= RegWidth)
-            return Error(Op3->getStartLoc(),
+            return Error(Op3.getStartLoc(),
                          "expected integer in range [0, 31]");
           if (Op4Val < 1 || Op4Val > RegWidth)
-            return Error(Op4->getStartLoc(),
+            return Error(Op4.getStartLoc(),
                          "expected integer in range [1, 32]");
 
           uint64_t NewOp4Val = Op3Val + Op4Val - 1;
 
           if (NewOp4Val >= RegWidth || NewOp4Val < Op3Val)
-            return Error(Op4->getStartLoc(),
+            return Error(Op4.getStartLoc(),
                          "requested extract overflows register");
 
           const MCExpr *NewOp4 =
               MCConstantExpr::Create(NewOp4Val, getContext());
           Operands[4] = AArch64Operand::CreateImm(
-              NewOp4, Op4->getStartLoc(), Op4->getEndLoc(), getContext());
+              NewOp4, Op4.getStartLoc(), Op4.getEndLoc(), getContext());
           if (Tok == "bfxil")
             Operands[0] = AArch64Operand::CreateToken(
-                "bfm", false, Op->getStartLoc(), getContext());
+                "bfm", false, Op.getStartLoc(), getContext());
           else if (Tok == "sbfx")
             Operands[0] = AArch64Operand::CreateToken(
-                "sbfm", false, Op->getStartLoc(), getContext());
+                "sbfm", false, Op.getStartLoc(), getContext());
           else if (Tok == "ubfx")
             Operands[0] = AArch64Operand::CreateToken(
-                "ubfm", false, Op->getStartLoc(), getContext());
+                "ubfm", false, Op.getStartLoc(), getContext());
           else
             llvm_unreachable("No valid mnemonic for alias?");
-
-          delete Op;
-          delete Op4;
         }
       }
     }
@@ -3626,63 +3740,58 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
   if (NumOperands == 3 && (Tok == "sxtw" || Tok == "uxtw")) {
     // The source register can be Wn here, but the matcher expects a
     // GPR64. Twiddle it here if necessary.
-    AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[2]);
-    if (Op->isReg()) {
-      unsigned Reg = getXRegFromWReg(Op->getReg());
-      Operands[2] = AArch64Operand::CreateReg(Reg, false, Op->getStartLoc(),
-                                              Op->getEndLoc(), getContext());
-      delete Op;
+    AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[2]);
+    if (Op.isReg()) {
+      unsigned Reg = getXRegFromWReg(Op.getReg());
+      Operands[2] = AArch64Operand::CreateReg(Reg, false, Op.getStartLoc(),
+                                              Op.getEndLoc(), getContext());
     }
   }
   // FIXME: Likewise for sxt[bh] with a Xd dst operand
   else if (NumOperands == 3 && (Tok == "sxtb" || Tok == "sxth")) {
-    AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[1]);
-    if (Op->isReg() &&
+    AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[1]);
+    if (Op.isReg() &&
         AArch64MCRegisterClasses[AArch64::GPR64allRegClassID].contains(
-            Op->getReg())) {
+            Op.getReg())) {
       // The source register can be Wn here, but the matcher expects a
       // GPR64. Twiddle it here if necessary.
-      AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[2]);
-      if (Op->isReg()) {
-        unsigned Reg = getXRegFromWReg(Op->getReg());
-        Operands[2] = AArch64Operand::CreateReg(Reg, false, Op->getStartLoc(),
-                                                Op->getEndLoc(), getContext());
-        delete Op;
+      AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[2]);
+      if (Op.isReg()) {
+        unsigned Reg = getXRegFromWReg(Op.getReg());
+        Operands[2] = AArch64Operand::CreateReg(Reg, false, Op.getStartLoc(),
+                                                Op.getEndLoc(), getContext());
       }
     }
   }
   // FIXME: Likewise for uxt[bh] with a Xd dst operand
   else if (NumOperands == 3 && (Tok == "uxtb" || Tok == "uxth")) {
-    AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[1]);
-    if (Op->isReg() &&
+    AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[1]);
+    if (Op.isReg() &&
         AArch64MCRegisterClasses[AArch64::GPR64allRegClassID].contains(
-            Op->getReg())) {
+            Op.getReg())) {
       // The source register can be Wn here, but the matcher expects a
       // GPR32. Twiddle it here if necessary.
-      AArch64Operand *Op = static_cast<AArch64Operand *>(Operands[1]);
-      if (Op->isReg()) {
-        unsigned Reg = getWRegFromXReg(Op->getReg());
-        Operands[1] = AArch64Operand::CreateReg(Reg, false, Op->getStartLoc(),
-                                                Op->getEndLoc(), getContext());
-        delete Op;
+      AArch64Operand &Op = static_cast<AArch64Operand &>(*Operands[1]);
+      if (Op.isReg()) {
+        unsigned Reg = getWRegFromXReg(Op.getReg());
+        Operands[1] = AArch64Operand::CreateReg(Reg, false, Op.getStartLoc(),
+                                                Op.getEndLoc(), getContext());
       }
     }
   }
 
   // Yet another horrible hack to handle FMOV Rd, #0.0 using [WX]ZR.
   if (NumOperands == 3 && Tok == "fmov") {
-    AArch64Operand *RegOp = static_cast<AArch64Operand *>(Operands[1]);
-    AArch64Operand *ImmOp = static_cast<AArch64Operand *>(Operands[2]);
-    if (RegOp->isReg() && ImmOp->isFPImm() &&
-        ImmOp->getFPImm() == (unsigned)-1) {
+    AArch64Operand &RegOp = static_cast<AArch64Operand &>(*Operands[1]);
+    AArch64Operand &ImmOp = static_cast<AArch64Operand &>(*Operands[2]);
+    if (RegOp.isReg() && ImmOp.isFPImm() && ImmOp.getFPImm() == (unsigned)-1) {
       unsigned zreg =
           AArch64MCRegisterClasses[AArch64::FPR32RegClassID].contains(
-              RegOp->getReg())
+              RegOp.getReg())
               ? AArch64::WZR
               : AArch64::XZR;
-      Operands[2] = AArch64Operand::CreateReg(zreg, false, Op->getStartLoc(),
-                                              Op->getEndLoc(), getContext());
-      delete ImmOp;
+      Operands[2] = AArch64Operand::CreateReg(zreg, false, Op.getStartLoc(),
+                                              Op.getEndLoc(), getContext());
     }
   }
 
@@ -3735,14 +3844,14 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
       if (ErrorInfo >= Operands.size())
         return Error(IDLoc, "too few operands for instruction");
 
-      ErrorLoc = ((AArch64Operand *)Operands[ErrorInfo])->getStartLoc();
+      ErrorLoc = ((AArch64Operand &)*Operands[ErrorInfo]).getStartLoc();
       if (ErrorLoc == SMLoc())
         ErrorLoc = IDLoc;
     }
     // If the match failed on a suffix token operand, tweak the diagnostic
     // accordingly.
-    if (((AArch64Operand *)Operands[ErrorInfo])->isToken() &&
-        ((AArch64Operand *)Operands[ErrorInfo])->isTokenSuffix())
+    if (((AArch64Operand &)*Operands[ErrorInfo]).isToken() &&
+        ((AArch64Operand &)*Operands[ErrorInfo]).isTokenSuffix())
       MatchResult = Match_InvalidSuffix;
 
     return showMatchError(ErrorLoc, MatchResult);
@@ -3794,9 +3903,11 @@ bool AArch64AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
   case Match_InvalidLabel:
   case Match_MSR:
   case Match_MRS: {
+    if (ErrorInfo >= Operands.size())
+      return Error(IDLoc, "too few operands for instruction");
     // Any time we get here, there's nothing fancy to do. Just get the
     // operand SMLoc and display the diagnostic.
-    SMLoc ErrorLoc = ((AArch64Operand *)Operands[ErrorInfo])->getStartLoc();
+    SMLoc ErrorLoc = ((AArch64Operand &)*Operands[ErrorInfo]).getStartLoc();
     if (ErrorLoc == SMLoc())
       ErrorLoc = IDLoc;
     return showMatchError(ErrorLoc, MatchResult);
@@ -3819,6 +3930,10 @@ bool AArch64AsmParser::ParseDirective(AsmToken DirectiveID) {
     return parseDirectiveWord(8, Loc);
   if (IDVal == ".tlsdesccall")
     return parseDirectiveTLSDescCall(Loc);
+  if (IDVal == ".ltorg" || IDVal == ".pool")
+    return parseDirectiveLtorg(Loc);
+  if (IDVal == ".unreq")
+    return parseDirectiveUnreq(DirectiveID.getLoc());
 
   return parseDirectiveLOH(IDVal, Loc);
 }
@@ -3920,6 +4035,66 @@ bool AArch64AsmParser::parseDirectiveLOH(StringRef IDVal, SMLoc Loc) {
   return false;
 }
 
+/// parseDirectiveLtorg
+///  ::= .ltorg | .pool
+bool AArch64AsmParser::parseDirectiveLtorg(SMLoc L) {
+  getTargetStreamer().emitCurrentConstantPool();
+  return false;
+}
+
+/// parseDirectiveReq
+///  ::= name .req registername
+bool AArch64AsmParser::parseDirectiveReq(StringRef Name, SMLoc L) {
+  Parser.Lex(); // Eat the '.req' token.
+  SMLoc SRegLoc = getLoc();
+  unsigned RegNum = tryParseRegister();
+  bool IsVector = false;
+
+  if (RegNum == static_cast<unsigned>(-1)) {
+    StringRef Kind;
+    RegNum = tryMatchVectorRegister(Kind, false);
+    if (!Kind.empty()) {
+      Error(SRegLoc, "vector register without type specifier expected");
+      return false;
+    }
+    IsVector = true;
+  }
+
+  if (RegNum == static_cast<unsigned>(-1)) {
+    Parser.eatToEndOfStatement();
+    Error(SRegLoc, "register name or alias expected");
+    return false;
+  }
+
+  // Shouldn't be anything else.
+  if (Parser.getTok().isNot(AsmToken::EndOfStatement)) {
+    Error(Parser.getTok().getLoc(), "unexpected input in .req directive");
+    Parser.eatToEndOfStatement();
+    return false;
+  }
+
+  Parser.Lex(); // Consume the EndOfStatement
+
+  auto pair = std::make_pair(IsVector, RegNum);
+  if (RegisterReqs.GetOrCreateValue(Name, pair).getValue() != pair)
+    Warning(L, "ignoring redefinition of register alias '" + Name + "'");
+
+  return true;
+}
+
+/// parseDirectiveUneq
+///  ::= .unreq registername
+bool AArch64AsmParser::parseDirectiveUnreq(SMLoc L) {
+  if (Parser.getTok().isNot(AsmToken::Identifier)) {
+    Error(Parser.getTok().getLoc(), "unexpected input in .unreq directive.");
+    Parser.eatToEndOfStatement();
+    return false;
+  }
+  RegisterReqs.erase(Parser.getTok().getIdentifier().lower());
+  Parser.Lex(); // Eat the identifier.
+  return false;
+}
+
 bool
 AArch64AsmParser::classifySymbolRef(const MCExpr *Expr,
                                     AArch64MCExpr::VariantKind &ELFRefKind,
@@ -3986,9 +4161,9 @@ extern "C" void LLVMInitializeAArch64AsmParser() {
 
 // Define this matcher function after the auto-generated include so we
 // have the match class enum definitions.
-unsigned AArch64AsmParser::validateTargetOperandClass(MCParsedAsmOperand *AsmOp,
+unsigned AArch64AsmParser::validateTargetOperandClass(MCParsedAsmOperand &AsmOp,
                                                       unsigned Kind) {
-  AArch64Operand *Op = static_cast<AArch64Operand *>(AsmOp);
+  AArch64Operand &Op = static_cast<AArch64Operand &>(AsmOp);
   // If the kind is a token for a literal immediate, check if our asm
   // operand matches. This is for InstAliases which have a fixed-value
   // immediate in the syntax.
@@ -4036,9 +4211,9 @@ unsigned AArch64AsmParser::validateTargetOperandClass(MCParsedAsmOperand *AsmOp,
     ExpectedVal = 8;
     break;
   }
-  if (!Op->isImm())
+  if (!Op.isImm())
     return Match_InvalidOperand;
-  const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(Op->getImm());
+  const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(Op.getImm());
   if (!CE)
     return Match_InvalidOperand;
   if (CE->getValue() == ExpectedVal)
diff --git a/lib/Target/AArch64/Disassembler/AArch64ExternalSymbolizer.cpp b/lib/Target/AArch64/Disassembler/AArch64ExternalSymbolizer.cpp
index 2466368..2057c51 100644
--- a/lib/Target/AArch64/Disassembler/AArch64ExternalSymbolizer.cpp
+++ b/lib/Target/AArch64/Disassembler/AArch64ExternalSymbolizer.cpp
@@ -37,8 +37,7 @@ getVariant(uint64_t LLVMDisassembler_VariantKind) {
   case LLVMDisassembler_VariantKind_ARM64_TLVP:
   case LLVMDisassembler_VariantKind_ARM64_TLVOFF:
   default:
-    assert(0 && "bad LLVMDisassembler_VariantKind");
-    return MCSymbolRefExpr::VK_None;
+    llvm_unreachable("bad LLVMDisassembler_VariantKind");
   }
 }
 
diff --git a/lib/Target/AArch64/Disassembler/CMakeLists.txt b/lib/Target/AArch64/Disassembler/CMakeLists.txt
index be4ccad..d64c05b 100644
--- a/lib/Target/AArch64/Disassembler/CMakeLists.txt
+++ b/lib/Target/AArch64/Disassembler/CMakeLists.txt
@@ -4,11 +4,5 @@ add_llvm_library(LLVMAArch64Disassembler
   AArch64Disassembler.cpp
   AArch64ExternalSymbolizer.cpp
   )
-# workaround for hanging compilation on MSVC8, 9 and 10
-#if( MSVC_VERSION EQUAL 1400 OR MSVC_VERSION EQUAL 1500 OR MSVC_VERSION EQUAL 1600 )
-#set_property(
-#  SOURCE ARMDisassembler.cpp
-#  PROPERTY COMPILE_FLAGS "/Od"
-#  )
-#endif()
+
 add_dependencies(LLVMAArch64Disassembler AArch64CommonTableGen)
diff --git a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
index f484a5b..8a21f06 100644
--- a/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
+++ b/lib/Target/AArch64/InstPrinter/AArch64InstPrinter.cpp
@@ -918,7 +918,7 @@ void AArch64InstPrinter::printPostIncOperand(const MCInst *MI, unsigned OpNo,
     else
       O << getRegisterName(Reg);
   } else
-    assert(0 && "unknown operand kind in printPostIncOperand64");
+    llvm_unreachable("unknown operand kind in printPostIncOperand64");
 }
 
 void AArch64InstPrinter::printVRegOperand(const MCInst *MI, unsigned OpNo,
@@ -1109,7 +1109,7 @@ static unsigned getNextVectorRegister(unsigned Reg, unsigned Stride = 1) {
   while (Stride--) {
     switch (Reg) {
     default:
-      assert(0 && "Vector register expected!");
+      llvm_unreachable("Vector register expected!");
     case AArch64::Q0:  Reg = AArch64::Q1;  break;
     case AArch64::Q1:  Reg = AArch64::Q2;  break;
     case AArch64::Q2:  Reg = AArch64::Q3;  break;
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp
index d8900d4..a917616 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp
@@ -86,7 +86,7 @@ public:
 static unsigned getFixupKindNumBytes(unsigned Kind) {
   switch (Kind) {
   default:
-    assert(0 && "Unknown fixup kind!");
+    llvm_unreachable("Unknown fixup kind!");
 
   case AArch64::fixup_aarch64_tlsdesc_call:
     return 0;
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
index dc4a8bf..1763b40 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
@@ -96,4 +96,6 @@ AArch64MCAsmInfoELF::AArch64MCAsmInfoELF(StringRef TT) {
   ExceptionsType = ExceptionHandling::DwarfCFI;
 
   UseIntegratedAssembler = true;
+
+  HasIdentDirective = true;
 }
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64MCCodeEmitter.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64MCCodeEmitter.cpp
index 464a18c..f051357 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64MCCodeEmitter.cpp
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64MCCodeEmitter.cpp
@@ -218,13 +218,9 @@ AArch64MCCodeEmitter::getMachineOpValue(const MCInst &MI, const MCOperand &MO,
                                         const MCSubtargetInfo &STI) const {
   if (MO.isReg())
     return Ctx.getRegisterInfo()->getEncodingValue(MO.getReg());
-  else {
-    assert(MO.isImm() && "did not expect relocated expression");
-    return static_cast<unsigned>(MO.getImm());
-  }
 
-  assert(0 && "Unable to encode MCOperand!");
-  return 0;
+  assert(MO.isImm() && "did not expect relocated expression");
+  return static_cast<unsigned>(MO.getImm());
 }
 
 template<unsigned FixupKind> uint32_t
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.cpp
index 85c3ec7..42a6787 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.cpp
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.cpp
@@ -81,37 +81,8 @@ void AArch64MCExpr::PrintImpl(raw_ostream &OS) const {
   OS << *Expr;
 }
 
-// FIXME: This basically copies MCObjectStreamer::AddValueSymbols. Perhaps
-// that method should be made public?
-// FIXME: really do above: now that two backends are using it.
-static void AddValueSymbolsImpl(const MCExpr *Value, MCAssembler *Asm) {
-  switch (Value->getKind()) {
-  case MCExpr::Target:
-    llvm_unreachable("Can't handle nested target expr!");
-    break;
-
-  case MCExpr::Constant:
-    break;
-
-  case MCExpr::Binary: {
-    const MCBinaryExpr *BE = cast<MCBinaryExpr>(Value);
-    AddValueSymbolsImpl(BE->getLHS(), Asm);
-    AddValueSymbolsImpl(BE->getRHS(), Asm);
-    break;
-  }
-
-  case MCExpr::SymbolRef:
-    Asm->getOrCreateSymbolData(cast<MCSymbolRefExpr>(Value)->getSymbol());
-    break;
-
-  case MCExpr::Unary:
-    AddValueSymbolsImpl(cast<MCUnaryExpr>(Value)->getSubExpr(), Asm);
-    break;
-  }
-}
-
-void AArch64MCExpr::AddValueSymbols(MCAssembler *Asm) const {
-  AddValueSymbolsImpl(getSubExpr(), Asm);
+void AArch64MCExpr::visitUsedExpr(MCStreamer &Streamer) const {
+  Streamer.visitUsedExpr(*getSubExpr());
 }
 
 const MCSection *AArch64MCExpr::FindAssociatedSection() const {
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.h b/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.h
index e869ed0..5422f9d 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.h
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64MCExpr.h
@@ -147,7 +147,7 @@ public:
 
   void PrintImpl(raw_ostream &OS) const override;
 
-  void AddValueSymbols(MCAssembler *) const override;
+  void visitUsedExpr(MCStreamer &Streamer) const override;
 
   const MCSection *FindAssociatedSection() const override;
 
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
index 5c86189..ba95366 100644
--- a/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
@@ -75,7 +75,7 @@ bool AArch64MachObjectWriter::getAArch64FixupKindMachOInfo(
     Log2Size = llvm::Log2_32(4);
     switch (Sym->getKind()) {
     default:
-      assert(0 && "Unexpected symbol reference variant kind!");
+      llvm_unreachable("Unexpected symbol reference variant kind!");
     case MCSymbolRefExpr::VK_PAGEOFF:
       RelocType = unsigned(MachO::ARM64_RELOC_PAGEOFF12);
       return true;
diff --git a/lib/Target/AArch64/MCTargetDesc/AArch64TargetStreamer.cpp b/lib/Target/AArch64/MCTargetDesc/AArch64TargetStreamer.cpp
new file mode 100644
index 0000000..f9aeb35
--- /dev/null
+++ b/lib/Target/AArch64/MCTargetDesc/AArch64TargetStreamer.cpp
@@ -0,0 +1,40 @@
+//===- AArch64TargetStreamer.cpp - AArch64TargetStreamer class --*- C++ -*---------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the AArch64TargetStreamer class.
+//
+//===----------------------------------------------------------------------===//
+#include "llvm/ADT/MapVector.h"
+#include "llvm/MC/ConstantPools.h"
+#include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCStreamer.h"
+
+using namespace llvm;
+
+//
+// AArch64TargetStreamer Implemenation
+//
+AArch64TargetStreamer::AArch64TargetStreamer(MCStreamer &S)
+    : MCTargetStreamer(S), ConstantPools(new AssemblerConstantPools()) {}
+
+AArch64TargetStreamer::~AArch64TargetStreamer() {}
+
+// The constant pool handling is shared by all AArch64TargetStreamer
+// implementations.
+const MCExpr *AArch64TargetStreamer::addConstantPoolEntry(const MCExpr *Expr) {
+  return ConstantPools->addEntry(Streamer, Expr);
+}
+
+void AArch64TargetStreamer::emitCurrentConstantPool() {
+  ConstantPools->emitForCurrentSection(Streamer);
+}
+
+// finish() - write out any non-empty assembler constant pools.
+void AArch64TargetStreamer::finish() { ConstantPools->emitAll(Streamer); }
diff --git a/lib/Target/AArch64/MCTargetDesc/Android.mk b/lib/Target/AArch64/MCTargetDesc/Android.mk
index e9d2323..a23c0e5 100644
--- a/lib/Target/AArch64/MCTargetDesc/Android.mk
+++ b/lib/Target/AArch64/MCTargetDesc/Android.mk
@@ -14,7 +14,8 @@ aarch64_mc_desc_SRC_FILES := \
   AArch64MCAsmInfo.cpp \
   AArch64MCCodeEmitter.cpp \
   AArch64MCExpr.cpp \
-  AArch64MCTargetDesc.cpp
+  AArch64MCTargetDesc.cpp \
+  AArch64TargetStreamer.cpp
 
 # For the host
 # =====================================================
diff --git a/lib/Target/AArch64/MCTargetDesc/CMakeLists.txt b/lib/Target/AArch64/MCTargetDesc/CMakeLists.txt
index 7d5bced..6d8be5e 100644
--- a/lib/Target/AArch64/MCTargetDesc/CMakeLists.txt
+++ b/lib/Target/AArch64/MCTargetDesc/CMakeLists.txt
@@ -7,6 +7,7 @@ add_llvm_library(LLVMAArch64Desc
   AArch64MCExpr.cpp
   AArch64MCTargetDesc.cpp
   AArch64MachObjectWriter.cpp
+  AArch64TargetStreamer.cpp
 )
 add_dependencies(LLVMAArch64Desc AArch64CommonTableGen)
 
diff --git a/lib/Target/AArch64/Utils/AArch64BaseInfo.h b/lib/Target/AArch64/Utils/AArch64BaseInfo.h
index 9e4c389..9d2ce21 100644
--- a/lib/Target/AArch64/Utils/AArch64BaseInfo.h
+++ b/lib/Target/AArch64/Utils/AArch64BaseInfo.h
@@ -233,23 +233,9 @@ inline static const char *getCondCodeName(CondCode Code) {
 }
 
 inline static CondCode getInvertedCondCode(CondCode Code) {
-  switch (Code) {
-  default: llvm_unreachable("Unknown condition code");
-  case EQ:  return NE;
-  case NE:  return EQ;
-  case HS:  return LO;
-  case LO:  return HS;
-  case MI:  return PL;
-  case PL:  return MI;
-  case VS:  return VC;
-  case VC:  return VS;
-  case HI:  return LS;
-  case LS:  return HI;
-  case GE:  return LT;
-  case LT:  return GE;
-  case GT:  return LE;
-  case LE:  return GT;
-  }
+  // To reverse a condition it's necessary to only invert the low bit:
+
+  return static_cast<CondCode>(static_cast<unsigned>(Code) ^ 0x1);
 }
 
 /// Given a condition code, return NZCV flags that would satisfy that condition.
author	Stephen Hines <srhines@google.com>	2014-07-21 00:45:20 -0700
committer	Stephen Hines <srhines@google.com>	2014-07-25 00:48:57 -0700
commit	cd81d94322a39503e4a3e87b6ee03d4fcb3465fb (patch)
tree	81b7dd2bb4370a392f31d332a566c903b5744764 /lib/Target/AArch64
parent	0c5f13c0c4499eaf42ab5e9e2ceabd4e20e36861 (diff)
download	external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.zip external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.tar.gz external_llvm-cd81d94322a39503e4a3e87b6ee03d4fcb3465fb.tar.bz2