disas/libvixl: Update to upstream VIXL 1.7

Update our copy of libvixl to upstream's 1.7 release.
This includes upstream's fix for the issue we had a local
patch for in commit 94cc44a9e.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1422274779-13359-2-git-send-email-peter.maydell@linaro.org

11 files changed, 537 insertions, 210 deletions

diff --git a/disas/libvixl/README b/disas/libvixl/README
index cba31b4..58db41c 100644
--- a/disas/libvixl/README
+++ b/disas/libvixl/README
@@ -2,7 +2,7 @@
 The code in this directory is a subset of libvixl:
  https://github.com/armvixl/vixl
 (specifically, it is the set of files needed for disassembly only,
-taken from libvixl 1.6).
+taken from libvixl 1.7).
 Bugfixes should preferably be sent upstream initially.
 
 The disassembler does not currently support the entire A64 instruction
diff --git a/disas/libvixl/a64/assembler-a64.h b/disas/libvixl/a64/assembler-a64.h
index 16a704b..35aaf20 100644
--- a/disas/libvixl/a64/assembler-a64.h
+++ b/disas/libvixl/a64/assembler-a64.h
@@ -151,21 +151,21 @@ class CPURegister {
     return Aliases(other) && (size_ == other.size_);
   }
 
-  inline bool IsZero() const {
+  bool IsZero() const {
     VIXL_ASSERT(IsValid());
     return IsRegister() && (code_ == kZeroRegCode);
   }
 
-  inline bool IsSP() const {
+  bool IsSP() const {
     VIXL_ASSERT(IsValid());
     return IsRegister() && (code_ == kSPRegInternalCode);
   }
 
-  inline bool IsRegister() const {
+  bool IsRegister() const {
     return type_ == kRegister;
   }
 
-  inline bool IsFPRegister() const {
+  bool IsFPRegister() const {
     return type_ == kFPRegister;
   }
 
@@ -179,7 +179,7 @@ class CPURegister {
   const FPRegister& S() const;
   const FPRegister& D() const;
 
-  inline bool IsSameSizeAndType(const CPURegister& other) const {
+  bool IsSameSizeAndType(const CPURegister& other) const {
     return (size_ == other.size_) && (type_ == other.type_);
   }
 
@@ -198,7 +198,7 @@ class CPURegister {
 class Register : public CPURegister {
  public:
   Register() : CPURegister() {}
-  inline explicit Register(const CPURegister& other)
+  explicit Register(const CPURegister& other)
       : CPURegister(other.code(), other.size(), other.type()) {
     VIXL_ASSERT(IsValidRegister());
   }
@@ -213,10 +213,6 @@ class Register : public CPURegister {
   static const Register& WRegFromCode(unsigned code);
   static const Register& XRegFromCode(unsigned code);
 
-  // V8 compatibility.
-  static const int kNumRegisters = kNumberOfRegisters;
-  static const int kNumAllocatableRegisters = kNumberOfRegisters - 1;
-
  private:
   static const Register wregisters[];
   static const Register xregisters[];
@@ -225,12 +221,12 @@ class Register : public CPURegister {
 
 class FPRegister : public CPURegister {
  public:
-  inline FPRegister() : CPURegister() {}
-  inline explicit FPRegister(const CPURegister& other)
+  FPRegister() : CPURegister() {}
+  explicit FPRegister(const CPURegister& other)
       : CPURegister(other.code(), other.size(), other.type()) {
     VIXL_ASSERT(IsValidFPRegister());
   }
-  inline FPRegister(unsigned code, unsigned size)
+  FPRegister(unsigned code, unsigned size)
       : CPURegister(code, size, kFPRegister) {}
 
   bool IsValid() const {
@@ -241,10 +237,6 @@ class FPRegister : public CPURegister {
   static const FPRegister& SRegFromCode(unsigned code);
   static const FPRegister& DRegFromCode(unsigned code);
 
-  // V8 compatibility.
-  static const int kNumRegisters = kNumberOfFPRegisters;
-  static const int kNumAllocatableRegisters = kNumberOfFPRegisters - 1;
-
  private:
   static const FPRegister sregisters[];
   static const FPRegister dregisters[];
@@ -312,23 +304,23 @@ bool AreSameSizeAndType(const CPURegister& reg1,
 // Lists of registers.
 class CPURegList {
  public:
-  inline explicit CPURegList(CPURegister reg1,
-                             CPURegister reg2 = NoCPUReg,
-                             CPURegister reg3 = NoCPUReg,
-                             CPURegister reg4 = NoCPUReg)
+  explicit CPURegList(CPURegister reg1,
+                      CPURegister reg2 = NoCPUReg,
+                      CPURegister reg3 = NoCPUReg,
+                      CPURegister reg4 = NoCPUReg)
       : list_(reg1.Bit() | reg2.Bit() | reg3.Bit() | reg4.Bit()),
         size_(reg1.size()), type_(reg1.type()) {
     VIXL_ASSERT(AreSameSizeAndType(reg1, reg2, reg3, reg4));
     VIXL_ASSERT(IsValid());
   }
 
-  inline CPURegList(CPURegister::RegisterType type, unsigned size, RegList list)
+  CPURegList(CPURegister::RegisterType type, unsigned size, RegList list)
       : list_(list), size_(size), type_(type) {
     VIXL_ASSERT(IsValid());
   }
 
-  inline CPURegList(CPURegister::RegisterType type, unsigned size,
-                    unsigned first_reg, unsigned last_reg)
+  CPURegList(CPURegister::RegisterType type, unsigned size,
+             unsigned first_reg, unsigned last_reg)
       : size_(size), type_(type) {
     VIXL_ASSERT(((type == CPURegister::kRegister) &&
                  (last_reg < kNumberOfRegisters)) ||
@@ -340,7 +332,7 @@ class CPURegList {
     VIXL_ASSERT(IsValid());
   }
 
-  inline CPURegister::RegisterType type() const {
+  CPURegister::RegisterType type() const {
     VIXL_ASSERT(IsValid());
     return type_;
   }
@@ -366,13 +358,13 @@ class CPURegList {
   }
 
   // Variants of Combine and Remove which take a single register.
-  inline void Combine(const CPURegister& other) {
+  void Combine(const CPURegister& other) {
     VIXL_ASSERT(other.type() == type_);
     VIXL_ASSERT(other.size() == size_);
     Combine(other.code());
   }
 
-  inline void Remove(const CPURegister& other) {
+  void Remove(const CPURegister& other) {
     VIXL_ASSERT(other.type() == type_);
     VIXL_ASSERT(other.size() == size_);
     Remove(other.code());
@@ -380,24 +372,51 @@ class CPURegList {
 
   // Variants of Combine and Remove which take a single register by its code;
   // the type and size of the register is inferred from this list.
-  inline void Combine(int code) {
+  void Combine(int code) {
     VIXL_ASSERT(IsValid());
     VIXL_ASSERT(CPURegister(code, size_, type_).IsValid());
     list_ |= (UINT64_C(1) << code);
   }
 
-  inline void Remove(int code) {
+  void Remove(int code) {
     VIXL_ASSERT(IsValid());
     VIXL_ASSERT(CPURegister(code, size_, type_).IsValid());
     list_ &= ~(UINT64_C(1) << code);
   }
 
-  inline RegList list() const {
+  static CPURegList Union(const CPURegList& list_1, const CPURegList& list_2) {
+    VIXL_ASSERT(list_1.type_ == list_2.type_);
+    VIXL_ASSERT(list_1.size_ == list_2.size_);
+    return CPURegList(list_1.type_, list_1.size_, list_1.list_ | list_2.list_);
+  }
+  static CPURegList Union(const CPURegList& list_1,
+                          const CPURegList& list_2,
+                          const CPURegList& list_3);
+  static CPURegList Union(const CPURegList& list_1,
+                          const CPURegList& list_2,
+                          const CPURegList& list_3,
+                          const CPURegList& list_4);
+
+  static CPURegList Intersection(const CPURegList& list_1,
+                                 const CPURegList& list_2) {
+    VIXL_ASSERT(list_1.type_ == list_2.type_);
+    VIXL_ASSERT(list_1.size_ == list_2.size_);
+    return CPURegList(list_1.type_, list_1.size_, list_1.list_ & list_2.list_);
+  }
+  static CPURegList Intersection(const CPURegList& list_1,
+                                 const CPURegList& list_2,
+                                 const CPURegList& list_3);
+  static CPURegList Intersection(const CPURegList& list_1,
+                                 const CPURegList& list_2,
+                                 const CPURegList& list_3,
+                                 const CPURegList& list_4);
+
+  RegList list() const {
     VIXL_ASSERT(IsValid());
     return list_;
   }
 
-  inline void set_list(RegList new_list) {
+  void set_list(RegList new_list) {
     VIXL_ASSERT(IsValid());
     list_ = new_list;
   }
@@ -417,38 +436,38 @@ class CPURegList {
   static CPURegList GetCallerSaved(unsigned size = kXRegSize);
   static CPURegList GetCallerSavedFP(unsigned size = kDRegSize);
 
-  inline bool IsEmpty() const {
+  bool IsEmpty() const {
     VIXL_ASSERT(IsValid());
     return list_ == 0;
   }
 
-  inline bool IncludesAliasOf(const CPURegister& other) const {
+  bool IncludesAliasOf(const CPURegister& other) const {
     VIXL_ASSERT(IsValid());
     return (type_ == other.type()) && ((other.Bit() & list_) != 0);
   }
 
-  inline bool IncludesAliasOf(int code) const {
+  bool IncludesAliasOf(int code) const {
     VIXL_ASSERT(IsValid());
     return ((code & list_) != 0);
   }
 
-  inline int Count() const {
+  int Count() const {
     VIXL_ASSERT(IsValid());
     return CountSetBits(list_, kRegListSizeInBits);
   }
 
-  inline unsigned RegisterSizeInBits() const {
+  unsigned RegisterSizeInBits() const {
     VIXL_ASSERT(IsValid());
     return size_;
   }
 
-  inline unsigned RegisterSizeInBytes() const {
+  unsigned RegisterSizeInBytes() const {
     int size_in_bits = RegisterSizeInBits();
     VIXL_ASSERT((size_in_bits % 8) == 0);
     return size_in_bits / 8;
   }
 
-  inline unsigned TotalSizeInBytes() const {
+  unsigned TotalSizeInBytes() const {
     VIXL_ASSERT(IsValid());
     return RegisterSizeInBytes() * Count();
   }
@@ -587,8 +606,10 @@ class Label {
     VIXL_ASSERT(!IsLinked() || IsBound());
   }
 
-  inline bool IsBound() const { return location_ >= 0; }
-  inline bool IsLinked() const { return !links_.empty(); }
+  bool IsBound() const { return location_ >= 0; }
+  bool IsLinked() const { return !links_.empty(); }
+
+  ptrdiff_t location() const { return location_; }
 
  private:
   // The list of linked instructions is stored in a stack-like structure. We
@@ -647,22 +668,20 @@ class Label {
     std::stack<ptrdiff_t> * links_extended_;
   };
 
-  inline ptrdiff_t location() const { return location_; }
-
-  inline void Bind(ptrdiff_t location) {
+  void Bind(ptrdiff_t location) {
     // Labels can only be bound once.
     VIXL_ASSERT(!IsBound());
     location_ = location;
   }
 
-  inline void AddLink(ptrdiff_t instruction) {
+  void AddLink(ptrdiff_t instruction) {
     // If a label is bound, the assembler already has the information it needs
     // to write the instruction, so there is no need to add it to links_.
     VIXL_ASSERT(!IsBound());
     links_.push(instruction);
   }
 
-  inline ptrdiff_t GetAndRemoveNextLink() {
+  ptrdiff_t GetAndRemoveNextLink() {
     VIXL_ASSERT(IsLinked());
     ptrdiff_t link = links_.top();
     links_.pop();
@@ -845,14 +864,14 @@ class Assembler {
 
   // Return the address of an offset in the buffer.
   template <typename T>
-  inline T GetOffsetAddress(ptrdiff_t offset) {
+  T GetOffsetAddress(ptrdiff_t offset) {
     VIXL_STATIC_ASSERT(sizeof(T) >= sizeof(uintptr_t));
     return buffer_->GetOffsetAddress<T>(offset);
   }
 
   // Return the address of a bound label.
   template <typename T>
-  inline T GetLabelAddress(const Label * label) {
+  T GetLabelAddress(const Label * label) {
     VIXL_ASSERT(label->IsBound());
     VIXL_STATIC_ASSERT(sizeof(T) >= sizeof(uintptr_t));
     return GetOffsetAddress<T>(label->location());
@@ -860,14 +879,14 @@ class Assembler {
 
   // Return the address of the cursor.
   template <typename T>
-  inline T GetCursorAddress() {
+  T GetCursorAddress() {
     VIXL_STATIC_ASSERT(sizeof(T) >= sizeof(uintptr_t));
     return GetOffsetAddress<T>(CursorOffset());
   }
 
   // Return the address of the start of the buffer.
   template <typename T>
-  inline T GetStartAddress() {
+  T GetStartAddress() {
     VIXL_STATIC_ASSERT(sizeof(T) >= sizeof(uintptr_t));
     return GetOffsetAddress<T>(0);
   }
@@ -1074,20 +1093,20 @@ class Assembler {
 
   // Bfm aliases.
   // Bitfield insert.
-  inline void bfi(const Register& rd,
-                  const Register& rn,
-                  unsigned lsb,
-                  unsigned width) {
+  void bfi(const Register& rd,
+           const Register& rn,
+           unsigned lsb,
+           unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     bfm(rd, rn, (rd.size() - lsb) & (rd.size() - 1), width - 1);
   }
 
   // Bitfield extract and insert low.
-  inline void bfxil(const Register& rd,
-                    const Register& rn,
-                    unsigned lsb,
-                    unsigned width) {
+  void bfxil(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     bfm(rd, rn, lsb, lsb + width - 1);
@@ -1095,92 +1114,92 @@ class Assembler {
 
   // Sbfm aliases.
   // Arithmetic shift right.
-  inline void asr(const Register& rd, const Register& rn, unsigned shift) {
+  void asr(const Register& rd, const Register& rn, unsigned shift) {
     VIXL_ASSERT(shift < rd.size());
     sbfm(rd, rn, shift, rd.size() - 1);
   }
 
   // Signed bitfield insert with zero at right.
-  inline void sbfiz(const Register& rd,
-                    const Register& rn,
-                    unsigned lsb,
-                    unsigned width) {
+  void sbfiz(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     sbfm(rd, rn, (rd.size() - lsb) & (rd.size() - 1), width - 1);
   }
 
   // Signed bitfield extract.
-  inline void sbfx(const Register& rd,
-                   const Register& rn,
-                   unsigned lsb,
-                   unsigned width) {
+  void sbfx(const Register& rd,
+            const Register& rn,
+            unsigned lsb,
+            unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     sbfm(rd, rn, lsb, lsb + width - 1);
   }
 
   // Signed extend byte.
-  inline void sxtb(const Register& rd, const Register& rn) {
+  void sxtb(const Register& rd, const Register& rn) {
     sbfm(rd, rn, 0, 7);
   }
 
   // Signed extend halfword.
-  inline void sxth(const Register& rd, const Register& rn) {
+  void sxth(const Register& rd, const Register& rn) {
     sbfm(rd, rn, 0, 15);
   }
 
   // Signed extend word.
-  inline void sxtw(const Register& rd, const Register& rn) {
+  void sxtw(const Register& rd, const Register& rn) {
     sbfm(rd, rn, 0, 31);
   }
 
   // Ubfm aliases.
   // Logical shift left.
-  inline void lsl(const Register& rd, const Register& rn, unsigned shift) {
+  void lsl(const Register& rd, const Register& rn, unsigned shift) {
     unsigned reg_size = rd.size();
     VIXL_ASSERT(shift < reg_size);
     ubfm(rd, rn, (reg_size - shift) % reg_size, reg_size - shift - 1);
   }
 
   // Logical shift right.
-  inline void lsr(const Register& rd, const Register& rn, unsigned shift) {
+  void lsr(const Register& rd, const Register& rn, unsigned shift) {
     VIXL_ASSERT(shift < rd.size());
     ubfm(rd, rn, shift, rd.size() - 1);
   }
 
   // Unsigned bitfield insert with zero at right.
-  inline void ubfiz(const Register& rd,
-                    const Register& rn,
-                    unsigned lsb,
-                    unsigned width) {
+  void ubfiz(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     ubfm(rd, rn, (rd.size() - lsb) & (rd.size() - 1), width - 1);
   }
 
   // Unsigned bitfield extract.
-  inline void ubfx(const Register& rd,
-                   const Register& rn,
-                   unsigned lsb,
-                   unsigned width) {
+  void ubfx(const Register& rd,
+            const Register& rn,
+            unsigned lsb,
+            unsigned width) {
     VIXL_ASSERT(width >= 1);
     VIXL_ASSERT(lsb + width <= rn.size());
     ubfm(rd, rn, lsb, lsb + width - 1);
   }
 
   // Unsigned extend byte.
-  inline void uxtb(const Register& rd, const Register& rn) {
+  void uxtb(const Register& rd, const Register& rn) {
     ubfm(rd, rn, 0, 7);
   }
 
   // Unsigned extend halfword.
-  inline void uxth(const Register& rd, const Register& rn) {
+  void uxth(const Register& rd, const Register& rn) {
     ubfm(rd, rn, 0, 15);
   }
 
   // Unsigned extend word.
-  inline void uxtw(const Register& rd, const Register& rn) {
+  void uxtw(const Register& rd, const Register& rn) {
     ubfm(rd, rn, 0, 31);
   }
 
@@ -1230,7 +1249,7 @@ class Assembler {
   void cneg(const Register& rd, const Register& rn, Condition cond);
 
   // Rotate right.
-  inline void ror(const Register& rd, const Register& rs, unsigned shift) {
+  void ror(const Register& rd, const Register& rs, unsigned shift) {
     extr(rd, rs, rs, shift);
   }
 
@@ -1495,6 +1514,19 @@ class Assembler {
   // Load-acquire register.
   void ldar(const Register& rt, const MemOperand& src);
 
+  // Prefetch memory.
+  void prfm(PrefetchOperation op, const MemOperand& addr,
+            LoadStoreScalingOption option = PreferScaledOffset);
+
+  // Prefetch memory (with unscaled offset).
+  void prfum(PrefetchOperation op, const MemOperand& addr,
+             LoadStoreScalingOption option = PreferUnscaledOffset);
+
+  // Prefetch memory in the literal pool.
+  void prfm(PrefetchOperation op, RawLiteral* literal);
+
+  // Prefetch from pc + imm19 << 2.
+  void prfm(PrefetchOperation op, int imm19);
 
   // Move instructions. The default shift of -1 indicates that the move
   // instruction will calculate an appropriate 16-bit immediate and left shift
@@ -1638,12 +1670,21 @@ class Assembler {
   // FP round to integer (nearest with ties to away).
   void frinta(const FPRegister& fd, const FPRegister& fn);
 
+  // FP round to integer (implicit rounding).
+  void frinti(const FPRegister& fd, const FPRegister& fn);
+
   // FP round to integer (toward minus infinity).
   void frintm(const FPRegister& fd, const FPRegister& fn);
 
   // FP round to integer (nearest with ties to even).
   void frintn(const FPRegister& fd, const FPRegister& fn);
 
+  // FP round to integer (toward plus infinity).
+  void frintp(const FPRegister& fd, const FPRegister& fn);
+
+  // FP round to integer (exact, implicit rounding).
+  void frintx(const FPRegister& fd, const FPRegister& fn);
+
   // FP round to integer (towards zero).
   void frintz(const FPRegister& fd, const FPRegister& fn);
 
@@ -1705,16 +1746,16 @@ class Assembler {
 
   // Emit generic instructions.
   // Emit raw instructions into the instruction stream.
-  inline void dci(Instr raw_inst) { Emit(raw_inst); }
+  void dci(Instr raw_inst) { Emit(raw_inst); }
 
   // Emit 32 bits of data into the instruction stream.
-  inline void dc32(uint32_t data) {
+  void dc32(uint32_t data) {
     VIXL_ASSERT(buffer_monitor_ > 0);
     buffer_->Emit32(data);
   }
 
   // Emit 64 bits of data into the instruction stream.
-  inline void dc64(uint64_t data) {
+  void dc64(uint64_t data) {
     VIXL_ASSERT(buffer_monitor_ > 0);
     buffer_->Emit64(data);
   }
@@ -1849,14 +1890,14 @@ class Assembler {
     }
   }
 
-  static inline Instr ImmS(unsigned imms, unsigned reg_size) {
+  static Instr ImmS(unsigned imms, unsigned reg_size) {
     VIXL_ASSERT(((reg_size == kXRegSize) && is_uint6(imms)) ||
            ((reg_size == kWRegSize) && is_uint5(imms)));
     USE(reg_size);
     return imms << ImmS_offset;
   }
 
-  static inline Instr ImmR(unsigned immr, unsigned reg_size) {
+  static Instr ImmR(unsigned immr, unsigned reg_size) {
     VIXL_ASSERT(((reg_size == kXRegSize) && is_uint6(immr)) ||
            ((reg_size == kWRegSize) && is_uint5(immr)));
     USE(reg_size);
@@ -1864,7 +1905,7 @@ class Assembler {
     return immr << ImmR_offset;
   }
 
-  static inline Instr ImmSetBits(unsigned imms, unsigned reg_size) {
+  static Instr ImmSetBits(unsigned imms, unsigned reg_size) {
     VIXL_ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
     VIXL_ASSERT(is_uint6(imms));
     VIXL_ASSERT((reg_size == kXRegSize) || is_uint6(imms + 3));
@@ -1872,7 +1913,7 @@ class Assembler {
     return imms << ImmSetBits_offset;
   }
 
-  static inline Instr ImmRotate(unsigned immr, unsigned reg_size) {
+  static Instr ImmRotate(unsigned immr, unsigned reg_size) {
     VIXL_ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
     VIXL_ASSERT(((reg_size == kXRegSize) && is_uint6(immr)) ||
            ((reg_size == kWRegSize) && is_uint5(immr)));
@@ -1880,12 +1921,12 @@ class Assembler {
     return immr << ImmRotate_offset;
   }
 
-  static inline Instr ImmLLiteral(int imm19) {
+  static Instr ImmLLiteral(int imm19) {
     VIXL_ASSERT(is_int19(imm19));
     return truncate_to_int19(imm19) << ImmLLiteral_offset;
   }
 
-  static inline Instr BitN(unsigned bitn, unsigned reg_size) {
+  static Instr BitN(unsigned bitn, unsigned reg_size) {
     VIXL_ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
     VIXL_ASSERT((reg_size == kXRegSize) || (bitn == 0));
     USE(reg_size);
@@ -1943,6 +1984,11 @@ class Assembler {
     return shift_amount << ImmShiftLS_offset;
   }
 
+  static Instr ImmPrefetchOperation(int imm5) {
+    VIXL_ASSERT(is_uint5(imm5));
+    return imm5 << ImmPrefetchOperation_offset;
+  }
+
   static Instr ImmException(int imm16) {
     VIXL_ASSERT(is_uint16(imm16));
     return imm16 << ImmException_offset;
@@ -2003,12 +2049,32 @@ class Assembler {
     return scale << FPScale_offset;
   }
 
+  // Immediate field checking helpers.
+  static bool IsImmAddSub(int64_t immediate);
+  static bool IsImmConditionalCompare(int64_t immediate);
+  static bool IsImmFP32(float imm);
+  static bool IsImmFP64(double imm);
+  static bool IsImmLogical(uint64_t value,
+                           unsigned width,
+                           unsigned* n = NULL,
+                           unsigned* imm_s = NULL,
+                           unsigned* imm_r = NULL);
+  static bool IsImmLSPair(int64_t offset, LSDataSize size);
+  static bool IsImmLSScaled(int64_t offset, LSDataSize size);
+  static bool IsImmLSUnscaled(int64_t offset);
+  static bool IsImmMovn(uint64_t imm, unsigned reg_size);
+  static bool IsImmMovz(uint64_t imm, unsigned reg_size);
+
   // Size of the code generated since label to the current position.
   size_t SizeOfCodeGeneratedSince(Label* label) const {
     VIXL_ASSERT(label->IsBound());
     return buffer_->OffsetFrom(label->location());
   }
 
+  size_t SizeOfCodeGenerated() const {
+    return buffer_->CursorOffset();
+  }
+
   size_t BufferCapacity() const { return buffer_->capacity(); }
 
   size_t RemainingBufferSpace() const { return buffer_->RemainingBytes(); }
@@ -2025,7 +2091,7 @@ class Assembler {
     }
   }
 
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
   void AcquireBuffer() {
     VIXL_ASSERT(buffer_monitor_ >= 0);
     buffer_monitor_++;
@@ -2037,16 +2103,16 @@ class Assembler {
   }
 #endif
 
-  inline PositionIndependentCodeOption pic() {
+  PositionIndependentCodeOption pic() const {
     return pic_;
   }
 
-  inline bool AllowPageOffsetDependentCode() {
+  bool AllowPageOffsetDependentCode() const {
     return (pic() == PageOffsetDependentCode) ||
            (pic() == PositionDependentCode);
   }
 
-  static inline const Register& AppropriateZeroRegFor(const CPURegister& reg) {
+  static const Register& AppropriateZeroRegFor(const CPURegister& reg) {
     return reg.Is64Bits() ? xzr : wzr;
   }
 
@@ -2056,14 +2122,15 @@ class Assembler {
                  const MemOperand& addr,
                  LoadStoreOp op,
                  LoadStoreScalingOption option = PreferScaledOffset);
-  static bool IsImmLSUnscaled(int64_t offset);
-  static bool IsImmLSScaled(int64_t offset, LSDataSize size);
 
   void LoadStorePair(const CPURegister& rt,
                      const CPURegister& rt2,
                      const MemOperand& addr,
                      LoadStorePairOp op);
-  static bool IsImmLSPair(int64_t offset, LSDataSize size);
+
+  void Prefetch(PrefetchOperation op,
+                const MemOperand& addr,
+                LoadStoreScalingOption option = PreferScaledOffset);
 
   // TODO(all): The third parameter should be passed by reference but gcc 4.8.2
   // reports a bogus uninitialised warning then.
@@ -2077,18 +2144,12 @@ class Assembler {
                         unsigned imm_s,
                         unsigned imm_r,
                         LogicalOp op);
-  static bool IsImmLogical(uint64_t value,
-                           unsigned width,
-                           unsigned* n = NULL,
-                           unsigned* imm_s = NULL,
-                           unsigned* imm_r = NULL);
 
   void ConditionalCompare(const Register& rn,
                           const Operand& operand,
                           StatusFlags nzcv,
                           Condition cond,
                           ConditionalCompareOp op);
-  static bool IsImmConditionalCompare(int64_t immediate);
 
   void AddSubWithCarry(const Register& rd,
                        const Register& rn,
@@ -2096,8 +2157,6 @@ class Assembler {
                        FlagsUpdate S,
                        AddSubWithCarryOp op);
 
-  static bool IsImmFP32(float imm);
-  static bool IsImmFP64(double imm);
 
   // Functions for emulating operands not directly supported by the instruction
   // set.
@@ -2115,7 +2174,6 @@ class Assembler {
               const Operand& operand,
               FlagsUpdate S,
               AddSubOp op);
-  static bool IsImmAddSub(int64_t immediate);
 
   // Find an appropriate LoadStoreOp or LoadStorePairOp for the specified
   // registers. Only simple loads are supported; sign- and zero-extension (such
@@ -2180,6 +2238,12 @@ class Assembler {
                                const FPRegister& fa,
                                FPDataProcessing3SourceOp op);
 
+  // Encode the specified MemOperand for the specified access size and scaling
+  // preference.
+  Instr LoadStoreMemOperand(const MemOperand& addr,
+                            LSDataSize size,
+                            LoadStoreScalingOption option);
+
   // Link the current (not-yet-emitted) instruction to the specified label, then
   // return an offset to be encoded in the instruction. If the label is not yet
   // bound, an offset of 0 is returned.
@@ -2205,7 +2269,7 @@ class Assembler {
   CodeBuffer* buffer_;
   PositionIndependentCodeOption pic_;
 
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
   int64_t buffer_monitor_;
 #endif
 };
@@ -2239,7 +2303,7 @@ class CodeBufferCheckScope {
                        AssertPolicy assert_policy = kMaximumSize)
       : assm_(assm) {
     if (check_policy == kCheck) assm->EnsureSpaceFor(size);
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
     assm->bind(&start_);
     size_ = size;
     assert_policy_ = assert_policy;
@@ -2251,7 +2315,7 @@ class CodeBufferCheckScope {
 
   // This is a shortcut for CodeBufferCheckScope(assm, 0, kNoCheck, kNoAssert).
   explicit CodeBufferCheckScope(Assembler* assm) : assm_(assm) {
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
     size_ = 0;
     assert_policy_ = kNoAssert;
     assm->AcquireBuffer();
@@ -2259,7 +2323,7 @@ class CodeBufferCheckScope {
   }
 
   ~CodeBufferCheckScope() {
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
     assm_->ReleaseBuffer();
     switch (assert_policy_) {
       case kNoAssert: break;
@@ -2277,7 +2341,7 @@ class CodeBufferCheckScope {
 
  protected:
   Assembler* assm_;
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
   Label start_;
   size_t size_;
   AssertPolicy assert_policy_;
diff --git a/disas/libvixl/a64/constants-a64.h b/disas/libvixl/a64/constants-a64.h
index 7a14f85..bc1a2c4 100644
--- a/disas/libvixl/a64/constants-a64.h
+++ b/disas/libvixl/a64/constants-a64.h
@@ -31,12 +31,6 @@ namespace vixl {
 
 const unsigned kNumberOfRegisters = 32;
 const unsigned kNumberOfFPRegisters = 32;
-// Callee saved registers are x21-x30(lr).
-const int kNumberOfCalleeSavedRegisters = 10;
-const int kFirstCalleeSavedRegisterIndex = 21;
-// Callee saved FP registers are d8-d15.
-const int kNumberOfCalleeSavedFPRegisters = 8;
-const int kFirstCalleeSavedFPRegisterIndex = 8;
 
 #define REGISTER_CODE_LIST(R)                                                  \
 R(0)  R(1)  R(2)  R(3)  R(4)  R(5)  R(6)  R(7)                                 \
@@ -53,7 +47,6 @@ V_(Ra, 14, 10, Bits)                      /* Third source register.       */   \
 V_(Rt, 4, 0, Bits)                        /* Load/store register.         */   \
 V_(Rt2, 14, 10, Bits)                     /* Load/store second register.  */   \
 V_(Rs, 20, 16, Bits)                      /* Exclusive access status.     */   \
-V_(PrefetchMode, 4, 0, Bits)                                                   \
                                                                                \
 /* Common bits */                                                              \
 V_(SixtyFourBits, 31, 31, Bits)                                                \
@@ -109,6 +102,10 @@ V_(ImmLSUnsigned, 21, 10, Bits)                                                \
 V_(ImmLSPair, 21, 15, SignedBits)                                              \
 V_(SizeLS, 31, 30, Bits)                                                       \
 V_(ImmShiftLS, 12, 12, Bits)                                                   \
+V_(ImmPrefetchOperation, 4, 0, Bits)                                           \
+V_(PrefetchHint, 4, 3, Bits)                                                   \
+V_(PrefetchTarget, 2, 1, Bits)                                                 \
+V_(PrefetchStream, 0, 0, Bits)                                                 \
                                                                                \
 /* Other immediates */                                                         \
 V_(ImmUncondBranch, 25, 0, SignedBits)                                         \
@@ -269,6 +266,29 @@ enum BarrierType {
   BarrierAll    = 3
 };
 
+enum PrefetchOperation {
+  PLDL1KEEP = 0x00,
+  PLDL1STRM = 0x01,
+  PLDL2KEEP = 0x02,
+  PLDL2STRM = 0x03,
+  PLDL3KEEP = 0x04,
+  PLDL3STRM = 0x05,
+
+  PLIL1KEEP = 0x08,
+  PLIL1STRM = 0x09,
+  PLIL2KEEP = 0x0a,
+  PLIL2STRM = 0x0b,
+  PLIL3KEEP = 0x0c,
+  PLIL3STRM = 0x0d,
+
+  PSTL1KEEP = 0x10,
+  PSTL1STRM = 0x11,
+  PSTL2KEEP = 0x12,
+  PSTL2STRM = 0x13,
+  PSTL3KEEP = 0x14,
+  PSTL3STRM = 0x15
+};
+
 // System/special register names.
 // This information is not encoded as one field but as the concatenation of
 // multiple fields (Op0<0>, Op1, Crn, Crm, Op2).
@@ -605,6 +625,12 @@ enum LoadStoreAnyOp {
   LoadStoreAnyFixed = 0x08000000
 };
 
+// Any load pair or store pair.
+enum LoadStorePairAnyOp {
+  LoadStorePairAnyFMask = 0x3a000000,
+  LoadStorePairAnyFixed = 0x28000000
+};
+
 #define LOAD_STORE_PAIR_OP_LIST(V)  \
   V(STP, w,   0x00000000),          \
   V(LDP, w,   0x00400000),          \
@@ -703,27 +729,28 @@ enum LoadLiteralOp {
   V(LD, R, d,   0xC4400000)
 
 
+// Load/store (post, pre, offset and unsigned.)
+enum LoadStoreOp {
+  LoadStoreOpMask = 0xC4C00000,
+  #define LOAD_STORE(A, B, C, D)  \
+  A##B##_##C = D
+  LOAD_STORE_OP_LIST(LOAD_STORE),
+  #undef LOAD_STORE
+  PRFM = 0xC0800000
+};
+
 // Load/store unscaled offset.
 enum LoadStoreUnscaledOffsetOp {
   LoadStoreUnscaledOffsetFixed = 0x38000000,
   LoadStoreUnscaledOffsetFMask = 0x3B200C00,
   LoadStoreUnscaledOffsetMask  = 0xFFE00C00,
+  PRFUM                        = LoadStoreUnscaledOffsetFixed | PRFM,
   #define LOAD_STORE_UNSCALED(A, B, C, D)  \
   A##U##B##_##C = LoadStoreUnscaledOffsetFixed | D
   LOAD_STORE_OP_LIST(LOAD_STORE_UNSCALED)
   #undef LOAD_STORE_UNSCALED
 };
 
-// Load/store (post, pre, offset and unsigned.)
-enum LoadStoreOp {
-  LoadStoreOpMask = 0xC4C00000,
-  #define LOAD_STORE(A, B, C, D)  \
-  A##B##_##C = D
-  LOAD_STORE_OP_LIST(LOAD_STORE),
-  #undef LOAD_STORE
-  PRFM = 0xC0800000
-};
-
 // Load/store post index.
 enum LoadStorePostIndex {
   LoadStorePostIndexFixed = 0x38000400,
diff --git a/disas/libvixl/a64/decoder-a64.h b/disas/libvixl/a64/decoder-a64.h
index 172594c..fd08d6c 100644
--- a/disas/libvixl/a64/decoder-a64.h
+++ b/disas/libvixl/a64/decoder-a64.h
@@ -108,7 +108,7 @@ class DecoderVisitor {
   }
 
  private:
-  VisitorConstness constness_;
+  const VisitorConstness constness_;
 };
 
 
diff --git a/disas/libvixl/a64/disasm-a64.cc b/disas/libvixl/a64/disasm-a64.cc
index e4a74aa..f7bc246 100644
--- a/disas/libvixl/a64/disasm-a64.cc
+++ b/disas/libvixl/a64/disasm-a64.cc
@@ -34,6 +34,7 @@ Disassembler::Disassembler() {
   buffer_ = reinterpret_cast<char*>(malloc(buffer_size_));
   buffer_pos_ = 0;
   own_buffer_ = true;
+  code_address_offset_ = 0;
 }
 
 
@@ -42,6 +43,7 @@ Disassembler::Disassembler(char* text_buffer, int buffer_size) {
   buffer_ = text_buffer;
   buffer_pos_ = 0;
   own_buffer_ = false;
+  code_address_offset_ = 0;
 }
 
 
@@ -739,9 +741,25 @@ void Disassembler::VisitMoveWideImmediate(const Instruction* instr) {
   // shift calculation.
   switch (instr->Mask(MoveWideImmediateMask)) {
     case MOVN_w:
-    case MOVN_x: mnemonic = "movn"; break;
+    case MOVN_x:
+      if ((instr->ImmMoveWide()) || (instr->ShiftMoveWide() == 0)) {
+        if ((instr->SixtyFourBits() == 0) && (instr->ImmMoveWide() == 0xffff)) {
+          mnemonic = "movn";
+        } else {
+          mnemonic = "mov";
+          form = "'Rd, 'IMoveNeg";
+        }
+      } else {
+        mnemonic = "movn";
+      }
+      break;
     case MOVZ_w:
-    case MOVZ_x: mnemonic = "movz"; break;
+    case MOVZ_x:
+      if ((instr->ImmMoveWide()) || (instr->ShiftMoveWide() == 0))
+        mnemonic = "mov";
+      else
+        mnemonic = "movz";
+      break;
     case MOVK_w:
     case MOVK_x: mnemonic = "movk"; form = "'Rd, 'IMoveLSL"; break;
     default: VIXL_UNREACHABLE();
@@ -806,7 +824,7 @@ void Disassembler::VisitLoadStoreUnsignedOffset(const Instruction* instr) {
     case A##_unsigned: mnemonic = B; form = C ", ['Xns'ILU]"; break;
     LOAD_STORE_LIST(LS_UNSIGNEDOFFSET)
     #undef LS_UNSIGNEDOFFSET
-    case PRFM_unsigned: mnemonic = "prfm"; form = "'PrefOp, ['Xn'ILU]";
+    case PRFM_unsigned: mnemonic = "prfm"; form = "'PrefOp, ['Xns'ILU]";
   }
   Format(instr, mnemonic, form);
 }
@@ -833,6 +851,7 @@ void Disassembler::VisitLoadStoreUnscaledOffset(const Instruction* instr) {
   const char *form_x = "'Xt, ['Xns'ILS]";
   const char *form_s = "'St, ['Xns'ILS]";
   const char *form_d = "'Dt, ['Xns'ILS]";
+  const char *form_prefetch = "'PrefOp, ['Xns'ILS]";
 
   switch (instr->Mask(LoadStoreUnscaledOffsetMask)) {
     case STURB_w:  mnemonic = "sturb"; break;
@@ -852,6 +871,7 @@ void Disassembler::VisitLoadStoreUnscaledOffset(const Instruction* instr) {
     case LDURSH_x: form = form_x;  // Fall through.
     case LDURSH_w: mnemonic = "ldursh"; break;
     case LDURSW_x: mnemonic = "ldursw"; form = form_x; break;
+    case PRFUM:    mnemonic = "prfum"; form = form_prefetch; break;
     default: form = "(LoadStoreUnscaledOffset)";
   }
   Format(instr, mnemonic, form);
@@ -872,6 +892,11 @@ void Disassembler::VisitLoadLiteral(const Instruction* instr) {
       form = "'Xt, 'ILLiteral 'LValue";
       break;
     }
+    case PRFM_lit: {
+      mnemonic = "prfm";
+      form = "'PrefOp, 'ILLiteral 'LValue";
+      break;
+    }
     default: mnemonic = "unimplemented";
   }
   Format(instr, mnemonic, form);
@@ -1344,7 +1369,7 @@ void Disassembler::AppendPCRelativeOffsetToOutput(const Instruction* instr,
 void Disassembler::AppendAddressToOutput(const Instruction* instr,
                                          const void* addr) {
   USE(instr);
-  AppendToOutput("(addr %p)", addr);
+  AppendToOutput("(addr 0x%" PRIxPTR ")", reinterpret_cast<uintptr_t>(addr));
 }
 
 
@@ -1360,6 +1385,40 @@ void Disassembler::AppendDataAddressToOutput(const Instruction* instr,
 }
 
 
+void Disassembler::AppendCodeRelativeAddressToOutput(const Instruction* instr,
+                                                     const void* addr) {
+  USE(instr);
+  int64_t rel_addr = CodeRelativeAddress(addr);
+  if (rel_addr >= 0) {
+    AppendToOutput("(addr 0x%" PRIx64 ")", rel_addr);
+  } else {
+    AppendToOutput("(addr -0x%" PRIx64 ")", -rel_addr);
+  }
+}
+
+
+void Disassembler::AppendCodeRelativeCodeAddressToOutput(
+    const Instruction* instr, const void* addr) {
+  AppendCodeRelativeAddressToOutput(instr, addr);
+}
+
+
+void Disassembler::AppendCodeRelativeDataAddressToOutput(
+    const Instruction* instr, const void* addr) {
+  AppendCodeRelativeAddressToOutput(instr, addr);
+}
+
+
+void Disassembler::MapCodeAddress(int64_t base_address,
+                                  const Instruction* instr_address) {
+  set_code_address_offset(
+      base_address - reinterpret_cast<intptr_t>(instr_address));
+}
+int64_t Disassembler::CodeRelativeAddress(const void* addr) {
+  return reinterpret_cast<intptr_t>(addr) + code_address_offset();
+}
+
+
 void Disassembler::Format(const Instruction* instr, const char* mnemonic,
                           const char* format) {
   VIXL_ASSERT(mnemonic != NULL);
@@ -1486,16 +1545,20 @@ int Disassembler::SubstituteImmediateField(const Instruction* instr,
   VIXL_ASSERT(format[0] == 'I');
 
   switch (format[1]) {
-    case 'M': {  // IMoveImm or IMoveLSL.
-      if (format[5] == 'I') {
-        uint64_t imm = instr->ImmMoveWide() << (16 * instr->ShiftMoveWide());
-        AppendToOutput("#0x%" PRIx64, imm);
-      } else {
-        VIXL_ASSERT(format[5] == 'L');
+    case 'M': {  // IMoveImm, IMoveNeg or IMoveLSL.
+      if (format[5] == 'L') {
         AppendToOutput("#0x%" PRIx64, instr->ImmMoveWide());
         if (instr->ShiftMoveWide() > 0) {
           AppendToOutput(", lsl #%" PRId64, 16 * instr->ShiftMoveWide());
         }
+      } else {
+        VIXL_ASSERT((format[5] == 'I') || (format[5] == 'N'));
+        uint64_t imm = instr->ImmMoveWide() << (16 * instr->ShiftMoveWide());
+        if (format[5] == 'N')
+          imm = ~imm;
+        if (!instr->SixtyFourBits())
+          imm &= UINT64_C(0xffffffff);
+        AppendToOutput("#0x%" PRIx64, imm);
       }
       return 8;
     }
@@ -1634,14 +1697,31 @@ int Disassembler::SubstituteLiteralField(const Instruction* instr,
   VIXL_ASSERT(strncmp(format, "LValue", 6) == 0);
   USE(format);
 
+  const void * address = instr->LiteralAddress<const void *>();
   switch (instr->Mask(LoadLiteralMask)) {
     case LDR_w_lit:
     case LDR_x_lit:
     case LDRSW_x_lit:
     case LDR_s_lit:
     case LDR_d_lit:
-      AppendDataAddressToOutput(instr, instr->LiteralAddress());
+      AppendCodeRelativeDataAddressToOutput(instr, address);
       break;
+    case PRFM_lit: {
+      // Use the prefetch hint to decide how to print the address.
+      switch (instr->PrefetchHint()) {
+        case 0x0:     // PLD: prefetch for load.
+        case 0x2:     // PST: prepare for store.
+          AppendCodeRelativeDataAddressToOutput(instr, address);
+          break;
+        case 0x1:     // PLI: preload instructions.
+          AppendCodeRelativeCodeAddressToOutput(instr, address);
+          break;
+        case 0x3:     // Unallocated hint.
+          AppendCodeRelativeAddressToOutput(instr, address);
+          break;
+      }
+      break;
+    }
     default:
       VIXL_UNREACHABLE();
   }
@@ -1701,17 +1781,22 @@ int Disassembler::SubstitutePCRelAddressField(const Instruction* instr,
               (strcmp(format, "AddrPCRelPage") == 0));    // Used by `adrp`.
 
   int64_t offset = instr->ImmPCRel();
-  const Instruction * base = instr;
 
+  // Compute the target address based on the effective address (after applying
+  // code_address_offset). This is required for correct behaviour of adrp.
+  const Instruction* base = instr + code_address_offset();
   if (format[9] == 'P') {
     offset *= kPageSize;
     base = AlignDown(base, kPageSize);
   }
+  // Strip code_address_offset before printing, so we can use the
+  // semantically-correct AppendCodeRelativeAddressToOutput.
+  const void* target =
+      reinterpret_cast<const void*>(base + offset - code_address_offset());
 
-  const void* target = reinterpret_cast<const void*>(base + offset);
   AppendPCRelativeOffsetToOutput(instr, offset);
   AppendToOutput(" ");
-  AppendAddressToOutput(instr, target);
+  AppendCodeRelativeAddressToOutput(instr, target);
   return 13;
 }
 
@@ -1738,7 +1823,7 @@ int Disassembler::SubstituteBranchTargetField(const Instruction* instr,
 
   AppendPCRelativeOffsetToOutput(instr, offset);
   AppendToOutput(" ");
-  AppendCodeAddressToOutput(instr, target_address);
+  AppendCodeRelativeCodeAddressToOutput(instr, target_address);
 
   return 8;
 }
@@ -1805,13 +1890,26 @@ int Disassembler::SubstitutePrefetchField(const Instruction* instr,
   VIXL_ASSERT(format[0] == 'P');
   USE(format);
 
-  int prefetch_mode = instr->PrefetchMode();
-
-  const char* ls = (prefetch_mode & 0x10) ? "st" : "ld";
-  int level = (prefetch_mode >> 1) + 1;
-  const char* ks = (prefetch_mode & 1) ? "strm" : "keep";
-
-  AppendToOutput("p%sl%d%s", ls, level, ks);
+  static const char* hints[] = {"ld", "li", "st"};
+  static const char* stream_options[] = {"keep", "strm"};
+
+  unsigned hint = instr->PrefetchHint();
+  unsigned target = instr->PrefetchTarget() + 1;
+  unsigned stream = instr->PrefetchStream();
+
+  if ((hint >= (sizeof(hints) / sizeof(hints[0]))) || (target > 3)) {
+    // Unallocated prefetch operations.
+    int prefetch_mode = instr->ImmPrefetchOperation();
+    AppendToOutput("#0b%c%c%c%c%c",
+                   (prefetch_mode & (1 << 4)) ? '1' : '0',
+                   (prefetch_mode & (1 << 3)) ? '1' : '0',
+                   (prefetch_mode & (1 << 2)) ? '1' : '0',
+                   (prefetch_mode & (1 << 1)) ? '1' : '0',
+                   (prefetch_mode & (1 << 0)) ? '1' : '0');
+  } else {
+    VIXL_ASSERT(stream < (sizeof(stream_options) / sizeof(stream_options[0])));
+    AppendToOutput("p%sl%d%s", hints[hint], target, stream_options[stream]);
+  }
   return 6;
 }
 
diff --git a/disas/libvixl/a64/disasm-a64.h b/disas/libvixl/a64/disasm-a64.h
index db04337..ddfe98b 100644
--- a/disas/libvixl/a64/disasm-a64.h
+++ b/disas/libvixl/a64/disasm-a64.h
@@ -43,7 +43,7 @@ class Disassembler: public DecoderVisitor {
   char* GetOutput();
 
   // Declare all Visitor functions.
-  #define DECLARE(A)  void Visit##A(const Instruction* instr);
+  #define DECLARE(A) virtual void Visit##A(const Instruction* instr);
   VISITOR_LIST(DECLARE)
   #undef DECLARE
 
@@ -65,23 +65,45 @@ class Disassembler: public DecoderVisitor {
 
   // Prints an address, in the general case. It can be code or data. This is
   // used for example to print the target address of an ADR instruction.
-  virtual void AppendAddressToOutput(const Instruction* instr,
-                                     const void* addr);
+  virtual void AppendCodeRelativeAddressToOutput(const Instruction* instr,
+                                                 const void* addr);
 
   // Prints the address of some code.
   // This is used for example to print the target address of a branch to an
   // immediate offset.
   // A sub-class can for example override this method to lookup the address and
   // print an appropriate name.
-  virtual void AppendCodeAddressToOutput(const Instruction* instr,
-                                         const void* addr);
+  virtual void AppendCodeRelativeCodeAddressToOutput(const Instruction* instr,
+                                                     const void* addr);
 
   // Prints the address of some data.
   // This is used for example to print the source address of a load literal
   // instruction.
+  virtual void AppendCodeRelativeDataAddressToOutput(const Instruction* instr,
+                                                     const void* addr);
+
+  // Same as the above, but for addresses that are not relative to the code
+  // buffer. They are currently not used by VIXL.
+  virtual void AppendAddressToOutput(const Instruction* instr,
+                                     const void* addr);
+  virtual void AppendCodeAddressToOutput(const Instruction* instr,
+                                         const void* addr);
   virtual void AppendDataAddressToOutput(const Instruction* instr,
                                          const void* addr);
 
+ public:
+  // Get/Set the offset that should be added to code addresses when printing
+  // code-relative addresses in the AppendCodeRelative<Type>AddressToOutput()
+  // helpers.
+  // Below is an example of how a branch immediate instruction in memory at
+  // address 0xb010200 would disassemble with different offsets.
+  // Base address | Disassembly
+  //          0x0 | 0xb010200:  b #+0xcc  (addr 0xb0102cc)
+  //      0x10000 | 0xb000200:  b #+0xcc  (addr 0xb0002cc)
+  //    0xb010200 |       0x0:  b #+0xcc  (addr 0xcc)
+  void MapCodeAddress(int64_t base_address, const Instruction* instr_address);
+  int64_t CodeRelativeAddress(const void* instr);
+
  private:
   void Format(
       const Instruction* instr, const char* mnemonic, const char* format);
@@ -101,32 +123,40 @@ class Disassembler: public DecoderVisitor {
   int SubstitutePrefetchField(const Instruction* instr, const char* format);
   int SubstituteBarrierField(const Instruction* instr, const char* format);
 
-  inline bool RdIsZROrSP(const Instruction* instr) const {
+  bool RdIsZROrSP(const Instruction* instr) const {
     return (instr->Rd() == kZeroRegCode);
   }
 
-  inline bool RnIsZROrSP(const Instruction* instr) const {
+  bool RnIsZROrSP(const Instruction* instr) const {
     return (instr->Rn() == kZeroRegCode);
   }
 
-  inline bool RmIsZROrSP(const Instruction* instr) const {
+  bool RmIsZROrSP(const Instruction* instr) const {
     return (instr->Rm() == kZeroRegCode);
   }
 
-  inline bool RaIsZROrSP(const Instruction* instr) const {
+  bool RaIsZROrSP(const Instruction* instr) const {
     return (instr->Ra() == kZeroRegCode);
   }
 
   bool IsMovzMovnImm(unsigned reg_size, uint64_t value);
 
+  int64_t code_address_offset() const { return code_address_offset_; }
+
  protected:
   void ResetOutput();
   void AppendToOutput(const char* string, ...) PRINTF_CHECK(2, 3);
 
+  void set_code_address_offset(int64_t code_address_offset) {
+    code_address_offset_ = code_address_offset;
+  }
+
   char* buffer_;
   uint32_t buffer_pos_;
   uint32_t buffer_size_;
   bool own_buffer_;
+
+  int64_t code_address_offset_;
 };
 
 
diff --git a/disas/libvixl/a64/instructions-a64.cc b/disas/libvixl/a64/instructions-a64.cc
index 1f08c78..b091886 100644
--- a/disas/libvixl/a64/instructions-a64.cc
+++ b/disas/libvixl/a64/instructions-a64.cc
@@ -30,6 +30,20 @@
 namespace vixl {
 
 
+// Floating-point infinity values.
+const float kFP32PositiveInfinity = rawbits_to_float(0x7f800000);
+const float kFP32NegativeInfinity = rawbits_to_float(0xff800000);
+const double kFP64PositiveInfinity =
+    rawbits_to_double(UINT64_C(0x7ff0000000000000));
+const double kFP64NegativeInfinity =
+    rawbits_to_double(UINT64_C(0xfff0000000000000));
+
+
+// The default NaN values (for FPCR.DN=1).
+const double kFP64DefaultNaN = rawbits_to_double(UINT64_C(0x7ff8000000000000));
+const float kFP32DefaultNaN = rawbits_to_float(0x7fc00000);
+
+
 static uint64_t RotateRight(uint64_t value,
                             unsigned int rotate,
                             unsigned int width) {
@@ -54,6 +68,55 @@ static uint64_t RepeatBitsAcrossReg(unsigned reg_size,
 }
 
 
+bool Instruction::IsLoad() const {
+  if (Mask(LoadStoreAnyFMask) != LoadStoreAnyFixed) {
+    return false;
+  }
+
+  if (Mask(LoadStorePairAnyFMask) == LoadStorePairAnyFixed) {
+    return Mask(LoadStorePairLBit) != 0;
+  } else {
+    LoadStoreOp op = static_cast<LoadStoreOp>(Mask(LoadStoreOpMask));
+    switch (op) {
+      case LDRB_w:
+      case LDRH_w:
+      case LDR_w:
+      case LDR_x:
+      case LDRSB_w:
+      case LDRSB_x:
+      case LDRSH_w:
+      case LDRSH_x:
+      case LDRSW_x:
+      case LDR_s:
+      case LDR_d: return true;
+      default: return false;
+    }
+  }
+}
+
+
+bool Instruction::IsStore() const {
+  if (Mask(LoadStoreAnyFMask) != LoadStoreAnyFixed) {
+    return false;
+  }
+
+  if (Mask(LoadStorePairAnyFMask) == LoadStorePairAnyFixed) {
+    return Mask(LoadStorePairLBit) == 0;
+  } else {
+    LoadStoreOp op = static_cast<LoadStoreOp>(Mask(LoadStoreOpMask));
+    switch (op) {
+      case STRB_w:
+      case STRH_w:
+      case STR_w:
+      case STR_x:
+      case STR_s:
+      case STR_d: return true;
+      default: return false;
+    }
+  }
+}
+
+
 // Logical immediates can't encode zero, so a return value of zero is used to
 // indicate a failure case. Specifically, where the constraints on imm_s are
 // not met.
diff --git a/disas/libvixl/a64/instructions-a64.h b/disas/libvixl/a64/instructions-a64.h
index 29f9722..f1d883c 100644
--- a/disas/libvixl/a64/instructions-a64.h
+++ b/disas/libvixl/a64/instructions-a64.h
@@ -96,6 +96,17 @@ const unsigned kDoubleExponentBits = 11;
 const unsigned kFloatMantissaBits = 23;
 const unsigned kFloatExponentBits = 8;
 
+// Floating-point infinity values.
+extern const float kFP32PositiveInfinity;
+extern const float kFP32NegativeInfinity;
+extern const double kFP64PositiveInfinity;
+extern const double kFP64NegativeInfinity;
+
+// The default NaN values (for FPCR.DN=1).
+extern const double kFP64DefaultNaN;
+extern const float kFP32DefaultNaN;
+
+
 enum LSDataSize {
   LSByte        = 0,
   LSHalfword    = 1,
@@ -140,33 +151,33 @@ enum Reg31Mode {
 
 class Instruction {
  public:
-  inline Instr InstructionBits() const {
+  Instr InstructionBits() const {
     return *(reinterpret_cast<const Instr*>(this));
   }
 
-  inline void SetInstructionBits(Instr new_instr) {
+  void SetInstructionBits(Instr new_instr) {
     *(reinterpret_cast<Instr*>(this)) = new_instr;
   }
 
-  inline int Bit(int pos) const {
+  int Bit(int pos) const {
     return (InstructionBits() >> pos) & 1;
   }
 
-  inline uint32_t Bits(int msb, int lsb) const {
+  uint32_t Bits(int msb, int lsb) const {
     return unsigned_bitextract_32(msb, lsb, InstructionBits());
   }
 
-  inline int32_t SignedBits(int msb, int lsb) const {
+  int32_t SignedBits(int msb, int lsb) const {
     int32_t bits = *(reinterpret_cast<const int32_t*>(this));
     return signed_bitextract_32(msb, lsb, bits);
   }
 
-  inline Instr Mask(uint32_t mask) const {
+  Instr Mask(uint32_t mask) const {
     return InstructionBits() & mask;
   }
 
   #define DEFINE_GETTER(Name, HighBit, LowBit, Func)             \
-  inline int64_t Name() const { return Func(HighBit, LowBit); }
+  int64_t Name() const { return Func(HighBit, LowBit); }
   INSTRUCTION_FIELDS_LIST(DEFINE_GETTER)
   #undef DEFINE_GETTER
 
@@ -182,56 +193,64 @@ class Instruction {
   float ImmFP32() const;
   double ImmFP64() const;
 
-  inline LSDataSize SizeLSPair() const {
+  LSDataSize SizeLSPair() const {
     return CalcLSPairDataSize(
              static_cast<LoadStorePairOp>(Mask(LoadStorePairMask)));
   }
 
   // Helpers.
-  inline bool IsCondBranchImm() const {
+  bool IsCondBranchImm() const {
     return Mask(ConditionalBranchFMask) == ConditionalBranchFixed;
   }
 
-  inline bool IsUncondBranchImm() const {
+  bool IsUncondBranchImm() const {
     return Mask(UnconditionalBranchFMask) == UnconditionalBranchFixed;
   }
 
-  inline bool IsCompareBranch() const {
+  bool IsCompareBranch() const {
     return Mask(CompareBranchFMask) == CompareBranchFixed;
   }
 
-  inline bool IsTestBranch() const {
+  bool IsTestBranch() const {
     return Mask(TestBranchFMask) == TestBranchFixed;
   }
 
-  inline bool IsPCRelAddressing() const {
+  bool IsPCRelAddressing() const {
     return Mask(PCRelAddressingFMask) == PCRelAddressingFixed;
   }
 
-  inline bool IsLogicalImmediate() const {
+  bool IsLogicalImmediate() const {
     return Mask(LogicalImmediateFMask) == LogicalImmediateFixed;
   }
 
-  inline bool IsAddSubImmediate() const {
+  bool IsAddSubImmediate() const {
     return Mask(AddSubImmediateFMask) == AddSubImmediateFixed;
   }
 
-  inline bool IsAddSubExtended() const {
+  bool IsAddSubExtended() const {
     return Mask(AddSubExtendedFMask) == AddSubExtendedFixed;
   }
 
-  inline bool IsLoadOrStore() const {
+  bool IsLoadOrStore() const {
     return Mask(LoadStoreAnyFMask) == LoadStoreAnyFixed;
   }
 
-  inline bool IsMovn() const {
+  bool IsLoad() const;
+  bool IsStore() const;
+
+  bool IsLoadLiteral() const {
+    // This includes PRFM_lit.
+    return Mask(LoadLiteralFMask) == LoadLiteralFixed;
+  }
+
+  bool IsMovn() const {
     return (Mask(MoveWideImmediateMask) == MOVN_x) ||
            (Mask(MoveWideImmediateMask) == MOVN_w);
   }
 
   // Indicate whether Rd can be the stack pointer or the zero register. This
   // does not check that the instruction actually has an Rd field.
-  inline Reg31Mode RdMode() const {
+  Reg31Mode RdMode() const {
     // The following instructions use sp or wsp as Rd:
     //  Add/sub (immediate) when not setting the flags.
     //  Add/sub (extended) when not setting the flags.
@@ -260,7 +279,7 @@ class Instruction {
 
   // Indicate whether Rn can be the stack pointer or the zero register. This
   // does not check that the instruction actually has an Rn field.
-  inline Reg31Mode RnMode() const {
+  Reg31Mode RnMode() const {
     // The following instructions use sp or wsp as Rn:
     //  All loads and stores.
     //  Add/sub (immediate).
@@ -272,7 +291,7 @@ class Instruction {
     return Reg31IsZeroRegister;
   }
 
-  inline ImmBranchType BranchType() const {
+  ImmBranchType BranchType() const {
     if (IsCondBranchImm()) {
       return CondBranchType;
     } else if (IsUncondBranchImm()) {
@@ -296,55 +315,66 @@ class Instruction {
   // Patch a literal load instruction to load from 'source'.
   void SetImmLLiteral(const Instruction* source);
 
-  inline uint8_t* LiteralAddress() const {
-    int offset = ImmLLiteral() << kLiteralEntrySizeLog2;
-    const uint8_t* address = reinterpret_cast<const uint8_t*>(this) + offset;
-    // Note that the result is safely mutable only if the backing buffer is
-    // safely mutable.
-    return const_cast<uint8_t*>(address);
+  // Calculate the address of a literal referred to by a load-literal
+  // instruction, and return it as the specified type.
+  //
+  // The literal itself is safely mutable only if the backing buffer is safely
+  // mutable.
+  template <typename T>
+  T LiteralAddress() const {
+    uint64_t base_raw = reinterpret_cast<uintptr_t>(this);
+    ptrdiff_t offset = ImmLLiteral() << kLiteralEntrySizeLog2;
+    uint64_t address_raw = base_raw + offset;
+
+    // Cast the address using a C-style cast. A reinterpret_cast would be
+    // appropriate, but it can't cast one integral type to another.
+    T address = (T)(address_raw);
+
+    // Assert that the address can be represented by the specified type.
+    VIXL_ASSERT((uint64_t)(address) == address_raw);
+
+    return address;
   }
 
-  inline uint32_t Literal32() const {
+  uint32_t Literal32() const {
     uint32_t literal;
-    memcpy(&literal, LiteralAddress(), sizeof(literal));
-
+    memcpy(&literal, LiteralAddress<const void*>(), sizeof(literal));
     return literal;
   }
 
-  inline uint64_t Literal64() const {
+  uint64_t Literal64() const {
     uint64_t literal;
-    memcpy(&literal, LiteralAddress(), sizeof(literal));
-
+    memcpy(&literal, LiteralAddress<const void*>(), sizeof(literal));
     return literal;
   }
 
-  inline float LiteralFP32() const {
+  float LiteralFP32() const {
     return rawbits_to_float(Literal32());
   }
 
-  inline double LiteralFP64() const {
+  double LiteralFP64() const {
     return rawbits_to_double(Literal64());
   }
 
-  inline const Instruction* NextInstruction() const {
+  const Instruction* NextInstruction() const {
     return this + kInstructionSize;
   }
 
-  inline const Instruction* InstructionAtOffset(int64_t offset) const {
+  const Instruction* InstructionAtOffset(int64_t offset) const {
     VIXL_ASSERT(IsWordAligned(this + offset));
     return this + offset;
   }
 
-  template<typename T> static inline Instruction* Cast(T src) {
+  template<typename T> static Instruction* Cast(T src) {
     return reinterpret_cast<Instruction*>(src);
   }
 
-  template<typename T> static inline const Instruction* CastConst(T src) {
+  template<typename T> static const Instruction* CastConst(T src) {
     return reinterpret_cast<const Instruction*>(src);
   }
 
  private:
-  inline int ImmBranch() const;
+  int ImmBranch() const;
 
   void SetPCRelImmTarget(const Instruction* target);
   void SetBranchImmTarget(const Instruction* target);
diff --git a/disas/libvixl/globals.h b/disas/libvixl/globals.h
index e28dc66..0c24931 100644
--- a/disas/libvixl/globals.h
+++ b/disas/libvixl/globals.h
@@ -58,7 +58,7 @@ const int KBytes = 1024;
 const int MBytes = 1024 * KBytes;
 
 #define VIXL_ABORT() printf("in %s, line %i", __FILE__, __LINE__); abort()
-#ifdef DEBUG
+#ifdef VIXL_DEBUG
   #define VIXL_ASSERT(condition) assert(condition)
   #define VIXL_CHECK(condition) VIXL_ASSERT(condition)
   #define VIXL_UNIMPLEMENTED() printf("UNIMPLEMENTED\t"); VIXL_ABORT()
diff --git a/disas/libvixl/utils.cc b/disas/libvixl/utils.cc
index 21965d7..80b132a 100644
--- a/disas/libvixl/utils.cc
+++ b/disas/libvixl/utils.cc
@@ -135,4 +135,17 @@ bool IsPowerOf2(int64_t value) {
   return (value != 0) && ((value & (value - 1)) == 0);
 }
 
+
+unsigned CountClearHalfWords(uint64_t imm, unsigned reg_size) {
+  VIXL_ASSERT((reg_size % 8) == 0);
+  int count = 0;
+  for (unsigned i = 0; i < (reg_size / 16); i++) {
+    if ((imm & 0xffff) == 0) {
+      count++;
+    }
+    imm >>= 16;
+  }
+  return count;
+}
+
 }  // namespace vixl
diff --git a/disas/libvixl/utils.h b/disas/libvixl/utils.h
index 1540c30..b440626 100644
--- a/disas/libvixl/utils.h
+++ b/disas/libvixl/utils.h
@@ -166,6 +166,8 @@ int CountSetBits(uint64_t value, int width);
 uint64_t LowestSetBit(uint64_t value);
 bool IsPowerOf2(int64_t value);
 
+unsigned CountClearHalfWords(uint64_t imm, unsigned reg_size);
+
 // Pointer alignment
 // TODO: rename/refactor to make it specific to instructions.
 template<typename T>
@@ -174,14 +176,14 @@ bool IsWordAligned(T pointer) {
   return ((intptr_t)(pointer) & 3) == 0;
 }
 
-// Increment a pointer until it has the specified alignment.
+// Increment a pointer (up to 64 bits) until it has the specified alignment.
 template<class T>
 T AlignUp(T pointer, size_t alignment) {
   // Use C-style casts to get static_cast behaviour for integral types (T), and
   // reinterpret_cast behaviour for other types.
 
-  uintptr_t pointer_raw = (uintptr_t)pointer;
-  VIXL_STATIC_ASSERT(sizeof(pointer) == sizeof(pointer_raw));
+  uint64_t pointer_raw = (uint64_t)pointer;
+  VIXL_STATIC_ASSERT(sizeof(pointer) <= sizeof(pointer_raw));
 
   size_t align_step = (alignment - pointer_raw) % alignment;
   VIXL_ASSERT((pointer_raw + align_step) % alignment == 0);
@@ -189,14 +191,14 @@ T AlignUp(T pointer, size_t alignment) {
   return (T)(pointer_raw + align_step);
 }
 
-// Decrement a pointer until it has the specified alignment.
+// Decrement a pointer (up to 64 bits) until it has the specified alignment.
 template<class T>
 T AlignDown(T pointer, size_t alignment) {
   // Use C-style casts to get static_cast behaviour for integral types (T), and
   // reinterpret_cast behaviour for other types.
 
-  uintptr_t pointer_raw = (uintptr_t)pointer;
-  VIXL_STATIC_ASSERT(sizeof(pointer) == sizeof(pointer_raw));
+  uint64_t pointer_raw = (uint64_t)pointer;
+  VIXL_STATIC_ASSERT(sizeof(pointer) <= sizeof(pointer_raw));
 
   size_t align_step = pointer_raw % alignment;
   VIXL_ASSERT((pointer_raw - align_step) % alignment == 0);