diff mbox

[RFC,3/4] disas: add libvixl source code for aarch64

Message ID 52306AB5.3050701@huawei.com
State New
Headers show

Commit Message

Claudio Fontana Sept. 11, 2013, 1:05 p.m. UTC
this just adds the libvixl source code verbatim from ARM,
including the source code itself (src/), license and readmes.
Examples, the scons build system etc, are excluded.
Not yet bound to QEMU's build system.

Signed-off-by: Claudio Fontana <claudio.fontana@linaro.org>
---
 disas/libvixl/LICENCE                        |   30 +
 disas/libvixl/README.md                      |  128 ++
 disas/libvixl/doc/changelog.md               |   12 +
 disas/libvixl/doc/supported-instructions.md  | 1133 ++++++++++++++
 disas/libvixl/src/a64/assembler-a64.cc       | 2172 ++++++++++++++++++++++++++
 disas/libvixl/src/a64/assembler-a64.h        | 1784 +++++++++++++++++++++
 disas/libvixl/src/a64/constants-a64.h        | 1104 +++++++++++++
 disas/libvixl/src/a64/cpu-a64.cc             |  148 ++
 disas/libvixl/src/a64/cpu-a64.h              |   56 +
 disas/libvixl/src/a64/debugger-a64.cc        | 1511 ++++++++++++++++++
 disas/libvixl/src/a64/debugger-a64.h         |  188 +++
 disas/libvixl/src/a64/decoder-a64.cc         |  712 +++++++++
 disas/libvixl/src/a64/decoder-a64.h          |  198 +++
 disas/libvixl/src/a64/disasm-a64.cc          | 1678 ++++++++++++++++++++
 disas/libvixl/src/a64/disasm-a64.h           |  109 ++
 disas/libvixl/src/a64/instructions-a64.cc    |  238 +++
 disas/libvixl/src/a64/instructions-a64.h     |  344 ++++
 disas/libvixl/src/a64/instrument-a64.cc      |  638 ++++++++
 disas/libvixl/src/a64/instrument-a64.h       |  108 ++
 disas/libvixl/src/a64/macro-assembler-a64.cc | 1108 +++++++++++++
 disas/libvixl/src/a64/macro-assembler-a64.h  | 1175 ++++++++++++++
 disas/libvixl/src/a64/simulator-a64.cc       | 2077 ++++++++++++++++++++++++
 disas/libvixl/src/a64/simulator-a64.h        |  576 +++++++
 disas/libvixl/src/globals.h                  |   66 +
 disas/libvixl/src/platform.h                 |   43 +
 disas/libvixl/src/utils.cc                   |  120 ++
 disas/libvixl/src/utils.h                    |  126 ++
 27 files changed, 17582 insertions(+)
 create mode 100644 disas/libvixl/LICENCE
 create mode 100644 disas/libvixl/README.md
 create mode 100644 disas/libvixl/doc/changelog.md
 create mode 100644 disas/libvixl/doc/supported-instructions.md
 create mode 100644 disas/libvixl/src/a64/assembler-a64.cc
 create mode 100644 disas/libvixl/src/a64/assembler-a64.h
 create mode 100644 disas/libvixl/src/a64/constants-a64.h
 create mode 100644 disas/libvixl/src/a64/cpu-a64.cc
 create mode 100644 disas/libvixl/src/a64/cpu-a64.h
 create mode 100644 disas/libvixl/src/a64/debugger-a64.cc
 create mode 100644 disas/libvixl/src/a64/debugger-a64.h
 create mode 100644 disas/libvixl/src/a64/decoder-a64.cc
 create mode 100644 disas/libvixl/src/a64/decoder-a64.h
 create mode 100644 disas/libvixl/src/a64/disasm-a64.cc
 create mode 100644 disas/libvixl/src/a64/disasm-a64.h
 create mode 100644 disas/libvixl/src/a64/instructions-a64.cc
 create mode 100644 disas/libvixl/src/a64/instructions-a64.h
 create mode 100644 disas/libvixl/src/a64/instrument-a64.cc
 create mode 100644 disas/libvixl/src/a64/instrument-a64.h
 create mode 100644 disas/libvixl/src/a64/macro-assembler-a64.cc
 create mode 100644 disas/libvixl/src/a64/macro-assembler-a64.h
 create mode 100644 disas/libvixl/src/a64/simulator-a64.cc
 create mode 100644 disas/libvixl/src/a64/simulator-a64.h
 create mode 100644 disas/libvixl/src/globals.h
 create mode 100644 disas/libvixl/src/platform.h
 create mode 100644 disas/libvixl/src/utils.cc
 create mode 100644 disas/libvixl/src/utils.h
diff mbox

Patch

diff --git a/disas/libvixl/LICENCE b/disas/libvixl/LICENCE
new file mode 100644
index 0000000..b7e160a
--- /dev/null
+++ b/disas/libvixl/LICENCE
@@ -0,0 +1,30 @@ 
+LICENCE
+=======
+
+The software in this repository is covered by the following licence.
+
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/disas/libvixl/README.md b/disas/libvixl/README.md
new file mode 100644
index 0000000..e3c3a0a
--- /dev/null
+++ b/disas/libvixl/README.md
@@ -0,0 +1,128 @@ 
+VIXL: AArch64 Runtime Code Generation Library Version 1.1
+=========================================================
+
+Contents:
+
+ * Requirements
+ * Overview
+ * Known limitations
+ * Usage
+
+
+Requirements
+============
+
+To build VIXL the following software is required:
+
+ 1. Python 2.7
+ 2. SCons 2.0
+ 3. GCC 4.4
+
+A 64-bit host machine is required, implementing an LP64 data model. VIXL has
+only been tested using GCC on Ubuntu systems.
+
+To run the linter stage of the tests, the following software is also required:
+
+ 1. Git
+ 2. [Google's `cpplint.py`][cpplint]
+
+Refer to the 'Usage' section for details.
+
+
+Overview
+========
+
+VIXL is made of three components.
+
+ 1. A programmatic assembler to generate A64 code at runtime. The assembler
+    abstracts some of the constraints of the A64 ISA, for example most
+    instructions support any immediate.
+ 2. A disassembler which can print any instruction emitted by the assembler.
+ 3. A simulator which can simulate any instruction emitted by the assembler.
+    The simulator allows generated code to be run on another architecture
+    without the need for a full ISA model.
+
+The VIXL git repository can be found [on GitHub][vixl]. Changes from previous
+versions of VIXL can be found in the [Changelog](doc/changelog.md).
+
+
+Known Limitations
+=================
+
+VIXL was developed to target JavaScript engines so a number of features from A64
+were deemed unnecessary:
+
+ * No Advanced SIMD support.
+ * Limited rounding mode support for floating point.
+ * No support for synchronisation instructions.
+ * Limited support for system instructions.
+ * A few miscellaneous integer and floating point instructions are missing.
+
+The VIXL simulator supports only those instructions that the VIXL assembler can
+generate. The `doc` directory contains a
+[list of supported instructions](doc/supported-instructions.md).
+
+
+Usage
+=====
+
+
+Running all Tests
+-----------------
+
+The helper script `tools/presubmit.py` will build and run every test that is
+provided with VIXL, in both release and debug mode. It is a useful script for
+verifying that all of VIXL's dependencies are in place and that VIXL is working
+as it should.
+
+By default, the `tools/presubmit.py` script runs a linter to check that the
+source code conforms with the code style guide, and to detect several common
+errors that the compiler may not warn about. This is most useful for VIXL
+developers. The linter has the following dependencies:
+
+ 1. Git must be installed, and the VIXL project must be in a valid Git
+    repository, such as one produced using `git clone`.
+ 2. `cpplint.py`, [as provided by Google][cpplint], must be available (and
+    executable) on the `PATH`. Only revision 104 has been tested with VIXL.
+
+It is possible to tell `tools/presubmit.py` to skip the linter stage by passing
+`--nolint`. This removes the dependency on `cpplint.py` and Git. The `--nolint`
+option is implied if the VIXL project is a snapshot (with no `.git` directory).
+
+
+Building and Running Specific Tests
+-----------------------------------
+
+The helper script `tools/test.py` will build and run all the tests for the
+assembler and disassembler in release mode. Add `--mode=debug` to build and run
+in debug mode. The tests can be built separately using SCons: `scons
+target=cctest`.
+
+
+Building and Running the Benchmarks
+-----------------------------------
+
+There are two very basic benchmarks provided with VIXL:
+
+ 1. bench\_dataop, emitting adds
+ 2. bench\_branch, emitting branches
+
+To build one benchmark: `scons target=bench_xxx`, then run it as
+`./bench_xxx_sim <number of iterations>`. The benchmarks do not report a
+figure; they should be timed using the `time` command.
+
+
+Getting Started
+---------------
+
+A short introduction to using VIXL can be found [here](doc/getting-started.md).
+Example source code is provided in the `examples` directory. Build this using
+`scons target=examples` from the root directory.
+
+
+
+[cpplint]: https://google-styleguide.googlecode.com/svn-history/r104/trunk/cpplint/cpplint.py
+           "Google's cpplint.py script."
+
+[vixl]: https://github.com/armvixl/vixl
+        "The VIXL repository on GitHub."
diff --git a/disas/libvixl/doc/changelog.md b/disas/libvixl/doc/changelog.md
new file mode 100644
index 0000000..8bab932
--- /dev/null
+++ b/disas/libvixl/doc/changelog.md
@@ -0,0 +1,12 @@ 
+VIXL Change Log
+===============
+
+* 1.1
+    + Improved robustness of instruction decoder and disassembler.
+    + Added support for double-to-float conversions using `fcvt`.
+    + Added support for more fixed-point to floating-point conversions (`ucvtf`
+      and `scvtf`).
+    + Added instruction statistics collection class `instrument-a64.cc`.
+
+* 1.0
+    + Initial release.
diff --git a/disas/libvixl/doc/supported-instructions.md b/disas/libvixl/doc/supported-instructions.md
new file mode 100644
index 0000000..90d63ec
--- /dev/null
+++ b/disas/libvixl/doc/supported-instructions.md
@@ -0,0 +1,1133 @@ 
+VIXL Supported Instruction List
+===============================
+
+This is a list of the AArch64 instructions supported by the VIXL assembler,
+disassembler and simulator. The simulator may not support all floating point
+operations to the precision required by AArch64 - please check the simulator
+source code for details.
+
+AArch64 integer instructions
+----------------------------
+
+### adc ###
+
+Add with carry bit.
+
+    void adc(const Register& rd,
+             const Register& rn,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### add ###
+
+Add.
+
+    void add(const Register& rd,
+             const Register& rn,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### adr ###
+
+Calculate the address of a PC offset.
+
+    void adr(const Register& rd, int imm21)
+
+
+### adr ###
+
+Calculate the address of a label.
+
+    void adr(const Register& rd, Label* label)
+
+
+### asr ###
+
+Arithmetic shift right.
+
+    inline void asr(const Register& rd, const Register& rn, unsigned shift)
+
+
+### asrv ###
+
+Arithmetic shift right by variable.
+
+    void asrv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### b ###
+
+Branch to PC offset.
+
+    void b(int imm26, Condition cond = al)
+
+
+### b ###
+
+Branch to label.
+
+    void b(Label* label, Condition cond = al)
+
+
+### bfi ###
+
+Bitfield insert.
+
+    inline void bfi(const Register& rd,
+                    const Register& rn,
+                    unsigned lsb,
+                    unsigned width)
+
+
+### bfm ###
+
+Bitfield move.
+
+    void bfm(const Register& rd,
+             const Register& rn,
+             unsigned immr,
+             unsigned imms)
+
+
+### bfxil ###
+
+Bitfield extract and insert low.
+
+    inline void bfxil(const Register& rd,
+                      const Register& rn,
+                      unsigned lsb,
+                      unsigned width)
+
+
+### bic ###
+
+Bit clear (A & ~B).
+
+    void bic(const Register& rd,
+             const Register& rn,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### bl ###
+
+Branch with link to PC offset.
+
+    void bl(int imm26)
+
+
+### bl ###
+
+Branch with link to label.
+
+    void bl(Label* label)
+
+
+### blr ###
+
+Branch with link to register.
+
+    void blr(const Register& xn)
+
+
+### br ###
+
+Branch to register.
+
+    void br(const Register& xn)
+
+
+### brk ###
+
+Monitor debug-mode breakpoint.
+
+    void brk(int code)
+
+
+### cbnz ###
+
+Compare and branch to PC offset if not zero.
+
+    void cbnz(const Register& rt, int imm19)
+
+
+### cbnz ###
+
+Compare and branch to label if not zero.
+
+    void cbnz(const Register& rt, Label* label)
+
+
+### cbz ###
+
+Compare and branch to PC offset if zero.
+
+    void cbz(const Register& rt, int imm19)
+
+
+### cbz ###
+
+Compare and branch to label if zero.
+
+    void cbz(const Register& rt, Label* label)
+
+
+### ccmn ###
+
+Conditional compare negative.
+
+    void ccmn(const Register& rn,
+              const Operand& operand,
+              StatusFlags nzcv,
+              Condition cond)
+
+
+### ccmp ###
+
+Conditional compare.
+
+    void ccmp(const Register& rn,
+              const Operand& operand,
+              StatusFlags nzcv,
+              Condition cond)
+
+
+### cinc ###
+
+Conditional increment: rd = cond ? rn + 1 : rn.
+
+    void cinc(const Register& rd, const Register& rn, Condition cond)
+
+
+### cinv ###
+
+Conditional invert: rd = cond ? ~rn : rn.
+
+    void cinv(const Register& rd, const Register& rn, Condition cond)
+
+
+### cls ###
+
+Count leading sign bits.
+
+    void cls(const Register& rd, const Register& rn)
+
+
+### clz ###
+
+Count leading zeroes.
+
+    void clz(const Register& rd, const Register& rn)
+
+
+### cmn ###
+
+Compare negative.
+
+    void cmn(const Register& rn, const Operand& operand)
+
+
+### cmp ###
+
+Compare.
+
+    void cmp(const Register& rn, const Operand& operand)
+
+
+### cneg ###
+
+Conditional negate: rd = cond ? -rn : rn.
+
+    void cneg(const Register& rd, const Register& rn, Condition cond)
+
+
+### csel ###
+
+Conditional select: rd = cond ? rn : rm.
+
+    void csel(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              Condition cond)
+
+
+### cset ###
+
+Conditional set: rd = cond ? 1 : 0.
+
+    void cset(const Register& rd, Condition cond)
+
+
+### csetm ###
+
+Conditional set mask: rd = cond ? -1 : 0.
+
+    void csetm(const Register& rd, Condition cond)
+
+
+### csinc ###
+
+Conditional select increment: rd = cond ? rn : rm + 1.
+
+    void csinc(const Register& rd,
+               const Register& rn,
+               const Register& rm,
+               Condition cond)
+
+
+### csinv ###
+
+Conditional select inversion: rd = cond ? rn : ~rm.
+
+    void csinv(const Register& rd,
+               const Register& rn,
+               const Register& rm,
+               Condition cond)
+
+
+### csneg ###
+
+Conditional select negation: rd = cond ? rn : -rm.
+
+    void csneg(const Register& rd,
+               const Register& rn,
+               const Register& rm,
+               Condition cond)
+
+
+### eon ###
+
+Bitwise enor/xnor (A ^ ~B).
+
+    void eon(const Register& rd, const Register& rn, const Operand& operand)
+
+
+### eor ###
+
+Bitwise eor/xor (A ^ B).
+
+    void eor(const Register& rd, const Register& rn, const Operand& operand)
+
+
+### extr ###
+
+Extract.
+
+    void extr(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              unsigned lsb)
+
+
+### hint ###
+
+System hint.
+
+    void hint(SystemHint code)
+
+
+### hlt ###
+
+Halting debug-mode breakpoint.
+
+    void hlt(int code)
+
+
+### ldnp ###
+
+Load integer or FP register pair, non-temporal.
+
+    void ldnp(const CPURegister& rt, const CPURegister& rt2,
+              const MemOperand& src)
+
+
+### ldp ###
+
+Load integer or FP register pair.
+
+    void ldp(const CPURegister& rt, const CPURegister& rt2,
+             const MemOperand& src)
+
+
+### ldpsw ###
+
+Load word pair with sign extension.
+
+    void ldpsw(const Register& rt, const Register& rt2, const MemOperand& src)
+
+
+### ldr ###
+
+Load integer or FP register.
+
+    void ldr(const CPURegister& rt, const MemOperand& src)
+
+
+### ldr ###
+
+Load literal to FP register.
+
+    void ldr(const FPRegister& ft, double imm)
+
+
+### ldr ###
+
+Load literal to register.
+
+    void ldr(const Register& rt, uint64_t imm)
+
+
+### ldrb ###
+
+Load byte.
+
+    void ldrb(const Register& rt, const MemOperand& src)
+
+
+### ldrh ###
+
+Load half-word.
+
+    void ldrh(const Register& rt, const MemOperand& src)
+
+
+### ldrsb ###
+
+Load byte with sign extension.
+
+    void ldrsb(const Register& rt, const MemOperand& src)
+
+
+### ldrsh ###
+
+Load half-word with sign extension.
+
+    void ldrsh(const Register& rt, const MemOperand& src)
+
+
+### ldrsw ###
+
+Load word with sign extension.
+
+    void ldrsw(const Register& rt, const MemOperand& src)
+
+
+### lsl ###
+
+Logical shift left.
+
+    inline void lsl(const Register& rd, const Register& rn, unsigned shift)
+
+
+### lslv ###
+
+Logical shift left by variable.
+
+    void lslv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### lsr ###
+
+Logical shift right.
+
+    inline void lsr(const Register& rd, const Register& rn, unsigned shift)
+
+
+### lsrv ###
+
+Logical shift right by variable.
+
+    void lsrv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### madd ###
+
+Multiply and accumulate.
+
+    void madd(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra)
+
+
+### mneg ###
+
+Negated multiply.
+
+    void mneg(const Register& rd, const Register& rn, const Register& rm)
+
+
+### mov ###
+
+Move register to register.
+
+    void mov(const Register& rd, const Register& rn)
+
+
+### movk ###
+
+Move immediate and keep.
+
+    void movk(const Register& rd, uint64_t imm, int shift = -1)
+
+
+### movn ###
+
+Move inverted immediate.
+
+    void movn(const Register& rd, uint64_t imm, int shift = -1)
+
+
+### movz ###
+
+Move immediate.
+
+    void movz(const Register& rd, uint64_t imm, int shift = -1)
+
+
+### mrs ###
+
+Move to register from system register.
+
+    void mrs(const Register& rt, SystemRegister sysreg)
+
+
+### msr ###
+
+Move from register to system register.
+
+    void msr(SystemRegister sysreg, const Register& rt)
+
+
+### msub ###
+
+Multiply and subtract.
+
+    void msub(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra)
+
+
+### mul ###
+
+Multiply.
+
+    void mul(const Register& rd, const Register& rn, const Register& rm)
+
+
+### mvn ###
+
+Move inverted operand to register.
+
+    void mvn(const Register& rd, const Operand& operand)
+
+
+### neg ###
+
+Negate.
+
+    void neg(const Register& rd,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### ngc ###
+
+Negate with carry bit.
+
+    void ngc(const Register& rd,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### nop ###
+
+No-op.
+
+    void nop()
+
+
+### orn ###
+
+Bitwise nor (A | ~B).
+
+    void orn(const Register& rd, const Register& rn, const Operand& operand)
+
+
+### orr ###
+
+Bitwise or (A | B).
+
+    void orr(const Register& rd, const Register& rn, const Operand& operand)
+
+
+### rbit ###
+
+Bit reverse.
+
+    void rbit(const Register& rd, const Register& rn)
+
+
+### ret ###
+
+Branch to register with return hint.
+
+    void ret(const Register& xn = lr)
+
+
+### rev ###
+
+Reverse bytes.
+
+    void rev(const Register& rd, const Register& rn)
+
+
+### rev16 ###
+
+Reverse bytes in 16-bit half words.
+
+    void rev16(const Register& rd, const Register& rn)
+
+
+### rev32 ###
+
+Reverse bytes in 32-bit words.
+
+    void rev32(const Register& rd, const Register& rn)
+
+
+### ror ###
+
+Rotate right.
+
+    inline void ror(const Register& rd, const Register& rs, unsigned shift)
+
+
+### rorv ###
+
+Rotate right by variable.
+
+    void rorv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### sbc ###
+
+Subtract with carry bit.
+
+    void sbc(const Register& rd,
+             const Register& rn,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### sbfiz ###
+
+Signed bitfield insert with zero at right.
+
+    inline void sbfiz(const Register& rd,
+                      const Register& rn,
+                      unsigned lsb,
+                      unsigned width)
+
+
+### sbfm ###
+
+Signed bitfield move.
+
+    void sbfm(const Register& rd,
+              const Register& rn,
+              unsigned immr,
+              unsigned imms)
+
+
+### sbfx ###
+
+Signed bitfield extract.
+
+    inline void sbfx(const Register& rd,
+                     const Register& rn,
+                     unsigned lsb,
+                     unsigned width)
+
+
+### scvtf ###
+
+Convert signed integer or fixed point to FP.
+
+    void scvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0)
+
+
+### sdiv ###
+
+Signed integer divide.
+
+    void sdiv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### smaddl ###
+
+Signed long multiply and accumulate: 32 x 32 + 64 -> 64-bit.
+
+    void smaddl(const Register& rd,
+                const Register& rn,
+                const Register& rm,
+                const Register& ra)
+
+
+### smsubl ###
+
+Signed long multiply and subtract: 64 - (32 x 32) -> 64-bit.
+
+    void smsubl(const Register& rd,
+                const Register& rn,
+                const Register& rm,
+                const Register& ra)
+
+
+### smulh ###
+
+Signed multiply high: 64 x 64 -> 64-bit <127:64>.
+
+    void smulh(const Register& xd, const Register& xn, const Register& xm)
+
+
+### smull ###
+
+Signed long multiply: 32 x 32 -> 64-bit.
+
+    void smull(const Register& rd, const Register& rn, const Register& rm)
+
+
+### stnp ###
+
+Store integer or FP register pair, non-temporal.
+
+    void stnp(const CPURegister& rt, const CPURegister& rt2,
+              const MemOperand& dst)
+
+
+### stp ###
+
+Store integer or FP register pair.
+
+    void stp(const CPURegister& rt, const CPURegister& rt2,
+             const MemOperand& dst)
+
+
+### str ###
+
+Store integer or FP register.
+
+    void str(const CPURegister& rt, const MemOperand& dst)
+
+
+### strb ###
+
+Store byte.
+
+    void strb(const Register& rt, const MemOperand& dst)
+
+
+### strh ###
+
+Store half-word.
+
+    void strh(const Register& rt, const MemOperand& dst)
+
+
+### sub ###
+
+Subtract.
+
+    void sub(const Register& rd,
+             const Register& rn,
+             const Operand& operand,
+             FlagsUpdate S = LeaveFlags)
+
+
+### sxtb ###
+
+Signed extend byte.
+
+    inline void sxtb(const Register& rd, const Register& rn)
+
+
+### sxth ###
+
+Signed extend halfword.
+
+    inline void sxth(const Register& rd, const Register& rn)
+
+
+### sxtw ###
+
+Signed extend word.
+
+    inline void sxtw(const Register& rd, const Register& rn)
+
+
+### tbnz ###
+
+Test bit and branch to PC offset if not zero.
+
+    void tbnz(const Register& rt, unsigned bit_pos, int imm14)
+
+
+### tbnz ###
+
+Test bit and branch to label if not zero.
+
+    void tbnz(const Register& rt, unsigned bit_pos, Label* label)
+
+
+### tbz ###
+
+Test bit and branch to PC offset if zero.
+
+    void tbz(const Register& rt, unsigned bit_pos, int imm14)
+
+
+### tbz ###
+
+Test bit and branch to label if zero.
+
+    void tbz(const Register& rt, unsigned bit_pos, Label* label)
+
+
+### tst ###
+
+Bit test and set flags.
+
+    void tst(const Register& rn, const Operand& operand)
+
+
+### ubfiz ###
+
+Unsigned bitfield insert with zero at right.
+
+    inline void ubfiz(const Register& rd,
+                      const Register& rn,
+                      unsigned lsb,
+                      unsigned width)
+
+
+### ubfm ###
+
+Unsigned bitfield move.
+
+    void ubfm(const Register& rd,
+              const Register& rn,
+              unsigned immr,
+              unsigned imms)
+
+
+### ubfx ###
+
+Unsigned bitfield extract.
+
+    inline void ubfx(const Register& rd,
+                     const Register& rn,
+                     unsigned lsb,
+                     unsigned width)
+
+
+### ucvtf ###
+
+Convert unsigned integer or fixed point to FP.
+
+    void ucvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0)
+
+
+### udiv ###
+
+Unsigned integer divide.
+
+    void udiv(const Register& rd, const Register& rn, const Register& rm)
+
+
+### umaddl ###
+
+Unsigned long multiply and accumulate: 32 x 32 + 64 -> 64-bit.
+
+    void umaddl(const Register& rd,
+                const Register& rn,
+                const Register& rm,
+                const Register& ra)
+
+
+### umsubl ###
+
+Unsigned long multiply and subtract: 64 - (32 x 32) -> 64-bit.
+
+    void umsubl(const Register& rd,
+                const Register& rn,
+                const Register& rm,
+                const Register& ra)
+
+
+### uxtb ###
+
+Unsigned extend byte.
+
+    inline void uxtb(const Register& rd, const Register& rn)
+
+
+### uxth ###
+
+Unsigned extend halfword.
+
+    inline void uxth(const Register& rd, const Register& rn)
+
+
+### uxtw ###
+
+Unsigned extend word.
+
+    inline void uxtw(const Register& rd, const Register& rn)
+
+
+
+AArch64 floating point instructions
+-----------------------------------
+
+### fabs ###
+
+FP absolute.
+
+    void fabs(const FPRegister& fd, const FPRegister& fn)
+
+
+### fadd ###
+
+FP add.
+
+    void fadd(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+### fccmp ###
+
+FP conditional compare.
+
+    void fccmp(const FPRegister& fn,
+               const FPRegister& fm,
+               StatusFlags nzcv,
+               Condition cond)
+
+
+### fcmp ###
+
+FP compare immediate.
+
+    void fcmp(const FPRegister& fn, double value)
+
+
+### fcmp ###
+
+FP compare registers.
+
+    void fcmp(const FPRegister& fn, const FPRegister& fm)
+
+
+### fcsel ###
+
+FP conditional select.
+
+    void fcsel(const FPRegister& fd,
+               const FPRegister& fn,
+               const FPRegister& fm,
+               Condition cond)
+
+
+### fcvt ###
+
+FP convert single to double precision.
+
+    void fcvt(const FPRegister& fd, const FPRegister& fn)
+
+
+### fcvtms ###
+
+Convert FP to signed integer (round towards -infinity).
+
+    void fcvtms(const Register& rd, const FPRegister& fn)
+
+
+### fcvtmu ###
+
+Convert FP to unsigned integer (round towards -infinity).
+
+    void fcvtmu(const Register& rd, const FPRegister& fn)
+
+
+### fcvtns ###
+
+Convert FP to signed integer (nearest with ties to even).
+
+    void fcvtns(const Register& rd, const FPRegister& fn)
+
+
+### fcvtnu ###
+
+Convert FP to unsigned integer (nearest with ties to even).
+
+    void fcvtnu(const Register& rd, const FPRegister& fn)
+
+
+### fcvtzs ###
+
+Convert FP to signed integer (round towards zero).
+
+    void fcvtzs(const Register& rd, const FPRegister& fn)
+
+
+### fcvtzu ###
+
+Convert FP to unsigned integer (round towards zero).
+
+    void fcvtzu(const Register& rd, const FPRegister& fn)
+
+
+### fdiv ###
+
+FP divide.
+
+    void fdiv(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+### fmax ###
+
+FP maximum.
+
+    void fmax(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+### fmin ###
+
+FP minimum.
+
+    void fmin(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+### fmov ###
+
+Move FP register to FP register.
+
+    void fmov(FPRegister fd, FPRegister fn)
+
+
+### fmov ###
+
+Move FP register to register.
+
+    void fmov(Register rd, FPRegister fn)
+
+
+### fmov ###
+
+Move immediate to FP register.
+
+    void fmov(FPRegister fd, double imm)
+
+
+### fmov ###
+
+Move register to FP register.
+
+    void fmov(FPRegister fd, Register rn)
+
+
+### fmsub ###
+
+FP multiply and subtract.
+
+    void fmsub(const FPRegister& fd,
+               const FPRegister& fn,
+               const FPRegister& fm,
+               const FPRegister& fa)
+
+
+### fmul ###
+
+FP multiply.
+
+    void fmul(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+### fneg ###
+
+FP negate.
+
+    void fneg(const FPRegister& fd, const FPRegister& fn)
+
+
+### frintn ###
+
+FP round to integer (nearest with ties to even).
+
+    void frintn(const FPRegister& fd, const FPRegister& fn)
+
+
+### frintz ###
+
+FP round to integer (towards zero).
+
+    void frintz(const FPRegister& fd, const FPRegister& fn)
+
+
+### fsqrt ###
+
+FP square root.
+
+    void fsqrt(const FPRegister& fd, const FPRegister& fn)
+
+
+### fsub ###
+
+FP subtract.
+
+    void fsub(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm)
+
+
+
+Additional or pseudo instructions
+---------------------------------
+
+### bind ###
+
+Bind a label to the current PC.
+
+    void bind(Label* label)
+
+
+### dc32 ###
+
+Emit 32 bits of data into the instruction stream.
+
+    inline void dc32(uint32_t data)
+
+
+### dc64 ###
+
+Emit 64 bits of data into the instruction stream.
+
+    inline void dc64(uint64_t data)
+
+
+### dci ###
+
+Emit raw instructions into the instruction stream.
+
+    inline void dci(Instr raw_inst)
+
+
+### debug ###
+
+Debug control pseudo instruction, only supported by the debugger.
+
+    void debug(const char* message, uint32_t code, Instr params = BREAK)
+
+
+
diff --git a/disas/libvixl/src/a64/assembler-a64.cc b/disas/libvixl/src/a64/assembler-a64.cc
new file mode 100644
index 0000000..89f7406
--- /dev/null
+++ b/disas/libvixl/src/a64/assembler-a64.cc
@@ -0,0 +1,2172 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+#include <cmath>
+#include "a64/assembler-a64.h"
+
+namespace vixl {
+
+// CPURegList utilities.
+CPURegister CPURegList::PopLowestIndex() {
+  if (IsEmpty()) {
+    return NoCPUReg;
+  }
+  int index = CountTrailingZeros(list_, kRegListSizeInBits);
+  ASSERT((1 << index) & list_);
+  Remove(index);
+  return CPURegister(index, size_, type_);
+}
+
+
+CPURegister CPURegList::PopHighestIndex() {
+  ASSERT(IsValid());
+  if (IsEmpty()) {
+    return NoCPUReg;
+  }
+  int index = CountLeadingZeros(list_, kRegListSizeInBits);
+  index = kRegListSizeInBits - 1 - index;
+  ASSERT((1 << index) & list_);
+  Remove(index);
+  return CPURegister(index, size_, type_);
+}
+
+
+bool CPURegList::IsValid() const {
+  if ((type_ == CPURegister::kRegister) ||
+      (type_ == CPURegister::kFPRegister)) {
+    bool is_valid = true;
+    // Try to create a CPURegister for each element in the list.
+    for (int i = 0; i < kRegListSizeInBits; i++) {
+      if (((list_ >> i) & 1) != 0) {
+        is_valid &= CPURegister(i, size_, type_).IsValid();
+      }
+    }
+    return is_valid;
+  } else if (type_ == CPURegister::kNoRegister) {
+    // We can't use IsEmpty here because that asserts IsValid().
+    return list_ == 0;
+  } else {
+    return false;
+  }
+}
+
+
+void CPURegList::RemoveCalleeSaved() {
+  if (type() == CPURegister::kRegister) {
+    Remove(GetCalleeSaved(RegisterSizeInBits()));
+  } else if (type() == CPURegister::kFPRegister) {
+    Remove(GetCalleeSavedFP(RegisterSizeInBits()));
+  } else {
+    ASSERT(type() == CPURegister::kNoRegister);
+    ASSERT(IsEmpty());
+    // The list must already be empty, so do nothing.
+  }
+}
+
+
+CPURegList CPURegList::GetCalleeSaved(unsigned size) {
+  return CPURegList(CPURegister::kRegister, size, 19, 29);
+}
+
+
+CPURegList CPURegList::GetCalleeSavedFP(unsigned size) {
+  return CPURegList(CPURegister::kFPRegister, size, 8, 15);
+}
+
+
+CPURegList CPURegList::GetCallerSaved(unsigned size) {
+  // Registers x0-x18 and lr (x30) are caller-saved.
+  CPURegList list = CPURegList(CPURegister::kRegister, size, 0, 18);
+  list.Combine(lr);
+  return list;
+}
+
+
+CPURegList CPURegList::GetCallerSavedFP(unsigned size) {
+  // Registers d0-d7 and d16-d31 are caller-saved.
+  CPURegList list = CPURegList(CPURegister::kFPRegister, size, 0, 7);
+  list.Combine(CPURegList(CPURegister::kFPRegister, size, 16, 31));
+  return list;
+}
+
+
+const CPURegList kCalleeSaved = CPURegList::GetCalleeSaved();
+const CPURegList kCalleeSavedFP = CPURegList::GetCalleeSavedFP();
+const CPURegList kCallerSaved = CPURegList::GetCallerSaved();
+const CPURegList kCallerSavedFP = CPURegList::GetCallerSavedFP();
+
+
+// Registers.
+#define WREG(n) w##n,
+const Register Register::wregisters[] = {
+REGISTER_CODE_LIST(WREG)
+};
+#undef WREG
+
+#define XREG(n) x##n,
+const Register Register::xregisters[] = {
+REGISTER_CODE_LIST(XREG)
+};
+#undef XREG
+
+#define SREG(n) s##n,
+const FPRegister FPRegister::sregisters[] = {
+REGISTER_CODE_LIST(SREG)
+};
+#undef SREG
+
+#define DREG(n) d##n,
+const FPRegister FPRegister::dregisters[] = {
+REGISTER_CODE_LIST(DREG)
+};
+#undef DREG
+
+
+const Register& Register::WRegFromCode(unsigned code) {
+  // This function returns the zero register when code = 31. The stack pointer
+  // can not be returned.
+  ASSERT(code < kNumberOfRegisters);
+  return wregisters[code];
+}
+
+
+const Register& Register::XRegFromCode(unsigned code) {
+  // This function returns the zero register when code = 31. The stack pointer
+  // can not be returned.
+  ASSERT(code < kNumberOfRegisters);
+  return xregisters[code];
+}
+
+
+const FPRegister& FPRegister::SRegFromCode(unsigned code) {
+  ASSERT(code < kNumberOfFPRegisters);
+  return sregisters[code];
+}
+
+
+const FPRegister& FPRegister::DRegFromCode(unsigned code) {
+  ASSERT(code < kNumberOfFPRegisters);
+  return dregisters[code];
+}
+
+
+const Register& CPURegister::W() const {
+  ASSERT(IsValidRegister());
+  ASSERT(Is64Bits());
+  return Register::WRegFromCode(code_);
+}
+
+
+const Register& CPURegister::X() const {
+  ASSERT(IsValidRegister());
+  ASSERT(Is32Bits());
+  return Register::XRegFromCode(code_);
+}
+
+
+const FPRegister& CPURegister::S() const {
+  ASSERT(IsValidFPRegister());
+  ASSERT(Is64Bits());
+  return FPRegister::SRegFromCode(code_);
+}
+
+
+const FPRegister& CPURegister::D() const {
+  ASSERT(IsValidFPRegister());
+  ASSERT(Is32Bits());
+  return FPRegister::DRegFromCode(code_);
+}
+
+
+// Operand.
+Operand::Operand(int64_t immediate)
+    : immediate_(immediate),
+      reg_(NoReg),
+      shift_(NO_SHIFT),
+      extend_(NO_EXTEND),
+      shift_amount_(0) {}
+
+
+Operand::Operand(Register reg, Shift shift, unsigned shift_amount)
+    : reg_(reg),
+      shift_(shift),
+      extend_(NO_EXTEND),
+      shift_amount_(shift_amount) {
+  ASSERT(reg.Is64Bits() || (shift_amount < kWRegSize));
+  ASSERT(reg.Is32Bits() || (shift_amount < kXRegSize));
+  ASSERT(!reg.IsSP());
+}
+
+
+Operand::Operand(Register reg, Extend extend, unsigned shift_amount)
+    : reg_(reg),
+      shift_(NO_SHIFT),
+      extend_(extend),
+      shift_amount_(shift_amount) {
+  ASSERT(reg.IsValid());
+  ASSERT(shift_amount <= 4);
+  ASSERT(!reg.IsSP());
+}
+
+
+bool Operand::IsImmediate() const {
+  return reg_.Is(NoReg);
+}
+
+
+bool Operand::IsShiftedRegister() const {
+  return reg_.IsValid() && (shift_ != NO_SHIFT);
+}
+
+
+bool Operand::IsExtendedRegister() const {
+  return reg_.IsValid() && (extend_ != NO_EXTEND);
+}
+
+
+Operand Operand::ToExtendedRegister() const {
+  ASSERT(IsShiftedRegister());
+  ASSERT((shift_ == LSL) && (shift_amount_ <= 4));
+  return Operand(reg_, reg_.Is64Bits() ? UXTX : UXTW, shift_amount_);
+}
+
+
+// MemOperand
+MemOperand::MemOperand(Register base, ptrdiff_t offset, AddrMode addrmode)
+  : base_(base), regoffset_(NoReg), offset_(offset), addrmode_(addrmode) {
+  ASSERT(base.Is64Bits() && !base.IsZero());
+}
+
+
+MemOperand::MemOperand(Register base,
+                       Register regoffset,
+                       Extend extend,
+                       unsigned shift_amount)
+  : base_(base), regoffset_(regoffset), offset_(0), addrmode_(Offset),
+    shift_(NO_SHIFT), extend_(extend), shift_amount_(shift_amount) {
+  ASSERT(base.Is64Bits() && !base.IsZero());
+  ASSERT(!regoffset.IsSP());
+  ASSERT((extend == UXTW) || (extend == SXTW) || (extend == SXTX));
+}
+
+
+MemOperand::MemOperand(Register base,
+                       Register regoffset,
+                       Shift shift,
+                       unsigned shift_amount)
+  : base_(base), regoffset_(regoffset), offset_(0), addrmode_(Offset),
+    shift_(shift), extend_(NO_EXTEND), shift_amount_(shift_amount) {
+  ASSERT(base.Is64Bits() && !base.IsZero());
+  ASSERT(!regoffset.IsSP());
+  ASSERT(shift == LSL);
+}
+
+
+MemOperand::MemOperand(Register base, const Operand& offset, AddrMode addrmode)
+  : base_(base), regoffset_(NoReg), addrmode_(addrmode) {
+  ASSERT(base.Is64Bits() && !base.IsZero());
+
+  if (offset.IsImmediate()) {
+    offset_ = offset.immediate();
+  } else if (offset.IsShiftedRegister()) {
+    ASSERT(addrmode == Offset);
+
+    regoffset_ = offset.reg();
+    shift_= offset.shift();
+    shift_amount_ = offset.shift_amount();
+
+    extend_ = NO_EXTEND;
+    offset_ = 0;
+
+    // These assertions match those in the shifted-register constructor.
+    ASSERT(!regoffset_.IsSP());
+    ASSERT(shift_ == LSL);
+  } else {
+    ASSERT(offset.IsExtendedRegister());
+    ASSERT(addrmode == Offset);
+
+    regoffset_ = offset.reg();
+    extend_ = offset.extend();
+    shift_amount_ = offset.shift_amount();
+
+    shift_= NO_SHIFT;
+    offset_ = 0;
+
+    // These assertions match those in the extended-register constructor.
+    ASSERT(!regoffset_.IsSP());
+    ASSERT((extend_ == UXTW) || (extend_ == SXTW) || (extend_ == SXTX));
+  }
+}
+
+
+bool MemOperand::IsImmediateOffset() const {
+  return (addrmode_ == Offset) && regoffset_.Is(NoReg);
+}
+
+
+bool MemOperand::IsRegisterOffset() const {
+  return (addrmode_ == Offset) && !regoffset_.Is(NoReg);
+}
+
+
+bool MemOperand::IsPreIndex() const {
+  return addrmode_ == PreIndex;
+}
+
+
+bool MemOperand::IsPostIndex() const {
+  return addrmode_ == PostIndex;
+}
+
+
+// Assembler
+Assembler::Assembler(byte* buffer, unsigned buffer_size)
+    : buffer_size_(buffer_size), literal_pool_monitor_(0) {
+  // Assert that this is an LP64 system.
+  ASSERT(sizeof(int) == sizeof(int32_t));     // NOLINT(runtime/sizeof)
+  ASSERT(sizeof(long) == sizeof(int64_t));    // NOLINT(runtime/int)
+  ASSERT(sizeof(void *) == sizeof(int64_t));  // NOLINT(runtime/sizeof)
+  ASSERT(sizeof(1) == sizeof(int32_t));       // NOLINT(runtime/sizeof)
+  ASSERT(sizeof(1L) == sizeof(int64_t));      // NOLINT(runtime/sizeof)
+
+  buffer_ = reinterpret_cast<Instruction*>(buffer);
+  pc_ = buffer_;
+  Reset();
+}
+
+
+Assembler::~Assembler() {
+  ASSERT(finalized_ || (pc_ == buffer_));
+  ASSERT(literals_.empty());
+}
+
+
+void Assembler::Reset() {
+#ifdef DEBUG
+  ASSERT((pc_ >= buffer_) && (pc_ < buffer_ + buffer_size_));
+  ASSERT(literal_pool_monitor_ == 0);
+  memset(buffer_, 0, pc_ - buffer_);
+  finalized_ = false;
+#endif
+  pc_ = buffer_;
+  literals_.clear();
+  next_literal_pool_check_ = pc_ + kLiteralPoolCheckInterval;
+}
+
+
+void Assembler::FinalizeCode() {
+  EmitLiteralPool();
+#ifdef DEBUG
+  finalized_ = true;
+#endif
+}
+
+
+void Assembler::bind(Label* label) {
+  label->is_bound_ = true;
+  label->target_ = pc_;
+  while (label->IsLinked()) {
+    // Get the address of the following instruction in the chain.
+    Instruction* next_link = label->link_->ImmPCOffsetTarget();
+    // Update the instruction target.
+    label->link_->SetImmPCOffsetTarget(label->target_);
+    // Update the label's link.
+    // If the offset of the branch we just updated was 0 (kEndOfChain) we are
+    // done.
+    label->link_ = (label->link_ != next_link) ? next_link : NULL;
+  }
+}
+
+
+int Assembler::UpdateAndGetByteOffsetTo(Label* label) {
+  int offset;
+  ASSERT(sizeof(*pc_) == 1);
+  if (label->IsBound()) {
+    offset = label->target() - pc_;
+  } else if (label->IsLinked()) {
+    offset = label->link() - pc_;
+  } else {
+    offset = Label::kEndOfChain;
+  }
+  label->set_link(pc_);
+  return offset;
+}
+
+
+// Code generation.
+void Assembler::br(const Register& xn) {
+  ASSERT(xn.Is64Bits());
+  Emit(BR | Rn(xn));
+}
+
+
+void Assembler::blr(const Register& xn) {
+  ASSERT(xn.Is64Bits());
+  Emit(BLR | Rn(xn));
+}
+
+
+void Assembler::ret(const Register& xn) {
+  ASSERT(xn.Is64Bits());
+  Emit(RET | Rn(xn));
+}
+
+
+void Assembler::b(int imm26) {
+  Emit(B | ImmUncondBranch(imm26));
+}
+
+
+void Assembler::b(int imm19, Condition cond) {
+  Emit(B_cond | ImmCondBranch(imm19) | cond);
+}
+
+
+void Assembler::b(Label* label) {
+  b(UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::b(Label* label, Condition cond) {
+  b(UpdateAndGetInstructionOffsetTo(label), cond);
+}
+
+
+void Assembler::bl(int imm26) {
+  Emit(BL | ImmUncondBranch(imm26));
+}
+
+
+void Assembler::bl(Label* label) {
+  bl(UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::cbz(const Register& rt,
+                    int imm19) {
+  Emit(SF(rt) | CBZ | ImmCmpBranch(imm19) | Rt(rt));
+}
+
+
+void Assembler::cbz(const Register& rt,
+                    Label* label) {
+  cbz(rt, UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::cbnz(const Register& rt,
+                     int imm19) {
+  Emit(SF(rt) | CBNZ | ImmCmpBranch(imm19) | Rt(rt));
+}
+
+
+void Assembler::cbnz(const Register& rt,
+                     Label* label) {
+  cbnz(rt, UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::tbz(const Register& rt,
+                    unsigned bit_pos,
+                    int imm14) {
+  ASSERT(rt.Is64Bits());
+  Emit(TBZ | ImmTestBranchBit(bit_pos) | ImmTestBranch(imm14) | Rt(rt));
+}
+
+
+void Assembler::tbz(const Register& rt,
+                    unsigned bit_pos,
+                    Label* label) {
+  tbz(rt, bit_pos, UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::tbnz(const Register& rt,
+                     unsigned bit_pos,
+                     int imm14) {
+  ASSERT(rt.Is64Bits());
+  Emit(TBNZ | ImmTestBranchBit(bit_pos) | ImmTestBranch(imm14) | Rt(rt));
+}
+
+
+void Assembler::tbnz(const Register& rt,
+                     unsigned bit_pos,
+                     Label* label) {
+  tbnz(rt, bit_pos, UpdateAndGetInstructionOffsetTo(label));
+}
+
+
+void Assembler::adr(const Register& rd, int imm21) {
+  ASSERT(rd.Is64Bits());
+  Emit(ADR | ImmPCRelAddress(imm21) | Rd(rd));
+}
+
+
+void Assembler::adr(const Register& rd, Label* label) {
+  adr(rd, UpdateAndGetByteOffsetTo(label));
+}
+
+
+void Assembler::add(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    FlagsUpdate S) {
+  AddSub(rd, rn, operand, S, ADD);
+}
+
+
+void Assembler::cmn(const Register& rn,
+                    const Operand& operand) {
+  Register zr = AppropriateZeroRegFor(rn);
+  add(zr, rn, operand, SetFlags);
+}
+
+
+void Assembler::sub(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    FlagsUpdate S) {
+  AddSub(rd, rn, operand, S, SUB);
+}
+
+
+void Assembler::cmp(const Register& rn, const Operand& operand) {
+  Register zr = AppropriateZeroRegFor(rn);
+  sub(zr, rn, operand, SetFlags);
+}
+
+
+void Assembler::neg(const Register& rd, const Operand& operand, FlagsUpdate S) {
+  Register zr = AppropriateZeroRegFor(rd);
+  sub(rd, zr, operand, S);
+}
+
+
+void Assembler::adc(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    FlagsUpdate S) {
+  AddSubWithCarry(rd, rn, operand, S, ADC);
+}
+
+
+void Assembler::sbc(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    FlagsUpdate S) {
+  AddSubWithCarry(rd, rn, operand, S, SBC);
+}
+
+
+void Assembler::ngc(const Register& rd, const Operand& operand, FlagsUpdate S) {
+  Register zr = AppropriateZeroRegFor(rd);
+  sbc(rd, zr, operand, S);
+}
+
+
+// Logical instructions.
+void Assembler::and_(const Register& rd,
+                     const Register& rn,
+                     const Operand& operand,
+                     FlagsUpdate S) {
+  Logical(rd, rn, operand, (S == SetFlags) ? ANDS : AND);
+}
+
+
+void Assembler::tst(const Register& rn,
+                    const Operand& operand) {
+  and_(AppropriateZeroRegFor(rn), rn, operand, SetFlags);
+}
+
+
+void Assembler::bic(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    FlagsUpdate S) {
+  Logical(rd, rn, operand, (S == SetFlags) ? BICS : BIC);
+}
+
+
+void Assembler::orr(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand) {
+  Logical(rd, rn, operand, ORR);
+}
+
+
+void Assembler::orn(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand) {
+  Logical(rd, rn, operand, ORN);
+}
+
+
+void Assembler::eor(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand) {
+  Logical(rd, rn, operand, EOR);
+}
+
+
+void Assembler::eon(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand) {
+  Logical(rd, rn, operand, EON);
+}
+
+
+void Assembler::lslv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | LSLV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::lsrv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | LSRV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::asrv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | ASRV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::rorv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | RORV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+// Bitfield operations.
+void Assembler::bfm(const Register& rd,
+                     const Register& rn,
+                     unsigned immr,
+                     unsigned imms) {
+  ASSERT(rd.size() == rn.size());
+  Instr N = SF(rd) >> (kSFOffset - kBitfieldNOffset);
+  Emit(SF(rd) | BFM | N |
+       ImmR(immr, rd.size()) | ImmS(imms, rd.size()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::sbfm(const Register& rd,
+                     const Register& rn,
+                     unsigned immr,
+                     unsigned imms) {
+  ASSERT(rd.size() == rn.size());
+  Instr N = SF(rd) >> (kSFOffset - kBitfieldNOffset);
+  Emit(SF(rd) | SBFM | N |
+       ImmR(immr, rd.size()) | ImmS(imms, rd.size()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::ubfm(const Register& rd,
+                     const Register& rn,
+                     unsigned immr,
+                     unsigned imms) {
+  ASSERT(rd.size() == rn.size());
+  Instr N = SF(rd) >> (kSFOffset - kBitfieldNOffset);
+  Emit(SF(rd) | UBFM | N |
+       ImmR(immr, rd.size()) | ImmS(imms, rd.size()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::extr(const Register& rd,
+                     const Register& rn,
+                     const Register& rm,
+                     unsigned lsb) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Instr N = SF(rd) >> (kSFOffset - kBitfieldNOffset);
+  Emit(SF(rd) | EXTR | N | Rm(rm) | ImmS(lsb, rd.size()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::csel(const Register& rd,
+                     const Register& rn,
+                     const Register& rm,
+                     Condition cond) {
+  ConditionalSelect(rd, rn, rm, cond, CSEL);
+}
+
+
+void Assembler::csinc(const Register& rd,
+                      const Register& rn,
+                      const Register& rm,
+                      Condition cond) {
+  ConditionalSelect(rd, rn, rm, cond, CSINC);
+}
+
+
+void Assembler::csinv(const Register& rd,
+                      const Register& rn,
+                      const Register& rm,
+                      Condition cond) {
+  ConditionalSelect(rd, rn, rm, cond, CSINV);
+}
+
+
+void Assembler::csneg(const Register& rd,
+                      const Register& rn,
+                      const Register& rm,
+                      Condition cond) {
+  ConditionalSelect(rd, rn, rm, cond, CSNEG);
+}
+
+
+void Assembler::cset(const Register &rd, Condition cond) {
+  ASSERT((cond != al) && (cond != nv));
+  Register zr = AppropriateZeroRegFor(rd);
+  csinc(rd, zr, zr, InvertCondition(cond));
+}
+
+
+void Assembler::csetm(const Register &rd, Condition cond) {
+  ASSERT((cond != al) && (cond != nv));
+  Register zr = AppropriateZeroRegFor(rd);
+  csinv(rd, zr, zr, InvertCondition(cond));
+}
+
+
+void Assembler::cinc(const Register &rd, const Register &rn, Condition cond) {
+  ASSERT((cond != al) && (cond != nv));
+  csinc(rd, rn, rn, InvertCondition(cond));
+}
+
+
+void Assembler::cinv(const Register &rd, const Register &rn, Condition cond) {
+  ASSERT((cond != al) && (cond != nv));
+  csinv(rd, rn, rn, InvertCondition(cond));
+}
+
+
+void Assembler::cneg(const Register &rd, const Register &rn, Condition cond) {
+  ASSERT((cond != al) && (cond != nv));
+  csneg(rd, rn, rn, InvertCondition(cond));
+}
+
+
+void Assembler::ConditionalSelect(const Register& rd,
+                                  const Register& rn,
+                                  const Register& rm,
+                                  Condition cond,
+                                  ConditionalSelectOp op) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | op | Rm(rm) | Cond(cond) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::ccmn(const Register& rn,
+                     const Operand& operand,
+                     StatusFlags nzcv,
+                     Condition cond) {
+  ConditionalCompare(rn, operand, nzcv, cond, CCMN);
+}
+
+
+void Assembler::ccmp(const Register& rn,
+                     const Operand& operand,
+                     StatusFlags nzcv,
+                     Condition cond) {
+  ConditionalCompare(rn, operand, nzcv, cond, CCMP);
+}
+
+
+void Assembler::DataProcessing3Source(const Register& rd,
+                     const Register& rn,
+                     const Register& rm,
+                     const Register& ra,
+                     DataProcessing3SourceOp op) {
+  Emit(SF(rd) | op | Rm(rm) | Ra(ra) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::mul(const Register& rd,
+                    const Register& rn,
+                    const Register& rm) {
+  ASSERT(AreSameSizeAndType(rd, rn, rm));
+  DataProcessing3Source(rd, rn, rm, AppropriateZeroRegFor(rd), MADD);
+}
+
+
+void Assembler::madd(const Register& rd,
+                     const Register& rn,
+                     const Register& rm,
+                     const Register& ra) {
+  DataProcessing3Source(rd, rn, rm, ra, MADD);
+}
+
+
+void Assembler::mneg(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(AreSameSizeAndType(rd, rn, rm));
+  DataProcessing3Source(rd, rn, rm, AppropriateZeroRegFor(rd), MSUB);
+}
+
+
+void Assembler::msub(const Register& rd,
+                     const Register& rn,
+                     const Register& rm,
+                     const Register& ra) {
+  DataProcessing3Source(rd, rn, rm, ra, MSUB);
+}
+
+
+void Assembler::umaddl(const Register& rd,
+                       const Register& rn,
+                       const Register& rm,
+                       const Register& ra) {
+  ASSERT(rd.Is64Bits() && ra.Is64Bits());
+  ASSERT(rn.Is32Bits() && rm.Is32Bits());
+  DataProcessing3Source(rd, rn, rm, ra, UMADDL_x);
+}
+
+
+void Assembler::smaddl(const Register& rd,
+                       const Register& rn,
+                       const Register& rm,
+                       const Register& ra) {
+  ASSERT(rd.Is64Bits() && ra.Is64Bits());
+  ASSERT(rn.Is32Bits() && rm.Is32Bits());
+  DataProcessing3Source(rd, rn, rm, ra, SMADDL_x);
+}
+
+
+void Assembler::umsubl(const Register& rd,
+                       const Register& rn,
+                       const Register& rm,
+                       const Register& ra) {
+  ASSERT(rd.Is64Bits() && ra.Is64Bits());
+  ASSERT(rn.Is32Bits() && rm.Is32Bits());
+  DataProcessing3Source(rd, rn, rm, ra, UMSUBL_x);
+}
+
+
+void Assembler::smsubl(const Register& rd,
+                       const Register& rn,
+                       const Register& rm,
+                       const Register& ra) {
+  ASSERT(rd.Is64Bits() && ra.Is64Bits());
+  ASSERT(rn.Is32Bits() && rm.Is32Bits());
+  DataProcessing3Source(rd, rn, rm, ra, SMSUBL_x);
+}
+
+
+void Assembler::smull(const Register& rd,
+                      const Register& rn,
+                      const Register& rm) {
+  ASSERT(rd.Is64Bits());
+  ASSERT(rn.Is32Bits() && rm.Is32Bits());
+  DataProcessing3Source(rd, rn, rm, xzr, SMADDL_x);
+}
+
+
+void Assembler::sdiv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | SDIV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::smulh(const Register& xd,
+                      const Register& xn,
+                      const Register& xm) {
+  ASSERT(xd.Is64Bits() && xn.Is64Bits() && xm.Is64Bits());
+  DataProcessing3Source(xd, xn, xm, xzr, SMULH_x);
+}
+
+void Assembler::udiv(const Register& rd,
+                     const Register& rn,
+                     const Register& rm) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == rm.size());
+  Emit(SF(rd) | UDIV | Rm(rm) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::rbit(const Register& rd,
+                     const Register& rn) {
+  DataProcessing1Source(rd, rn, RBIT);
+}
+
+
+void Assembler::rev16(const Register& rd,
+                      const Register& rn) {
+  DataProcessing1Source(rd, rn, REV16);
+}
+
+
+void Assembler::rev32(const Register& rd,
+                      const Register& rn) {
+  ASSERT(rd.Is64Bits());
+  DataProcessing1Source(rd, rn, REV);
+}
+
+
+void Assembler::rev(const Register& rd,
+                    const Register& rn) {
+  DataProcessing1Source(rd, rn, rd.Is64Bits() ? REV_x : REV_w);
+}
+
+
+void Assembler::clz(const Register& rd,
+                    const Register& rn) {
+  DataProcessing1Source(rd, rn, CLZ);
+}
+
+
+void Assembler::cls(const Register& rd,
+                    const Register& rn) {
+  DataProcessing1Source(rd, rn, CLS);
+}
+
+
+void Assembler::ldp(const CPURegister& rt,
+                    const CPURegister& rt2,
+                    const MemOperand& src) {
+  LoadStorePair(rt, rt2, src, LoadPairOpFor(rt, rt2));
+}
+
+
+void Assembler::stp(const CPURegister& rt,
+                    const CPURegister& rt2,
+                    const MemOperand& dst) {
+  LoadStorePair(rt, rt2, dst, StorePairOpFor(rt, rt2));
+}
+
+
+void Assembler::ldpsw(const Register& rt,
+                      const Register& rt2,
+                      const MemOperand& src) {
+  ASSERT(rt.Is64Bits());
+  LoadStorePair(rt, rt2, src, LDPSW_x);
+}
+
+
+void Assembler::LoadStorePair(const CPURegister& rt,
+                              const CPURegister& rt2,
+                              const MemOperand& addr,
+                              LoadStorePairOp op) {
+  // 'rt' and 'rt2' can only be aliased for stores.
+  ASSERT(((op & LoadStorePairLBit) == 0) || !rt.Is(rt2));
+  ASSERT(AreSameSizeAndType(rt, rt2));
+
+  Instr memop = op | Rt(rt) | Rt2(rt2) | RnSP(addr.base()) |
+                ImmLSPair(addr.offset(), CalcLSPairDataSize(op));
+
+  Instr addrmodeop;
+  if (addr.IsImmediateOffset()) {
+    addrmodeop = LoadStorePairOffsetFixed;
+  } else {
+    ASSERT(addr.offset() != 0);
+    if (addr.IsPreIndex()) {
+      addrmodeop = LoadStorePairPreIndexFixed;
+    } else {
+      ASSERT(addr.IsPostIndex());
+      addrmodeop = LoadStorePairPostIndexFixed;
+    }
+  }
+  Emit(addrmodeop | memop);
+}
+
+
+void Assembler::ldnp(const CPURegister& rt,
+                     const CPURegister& rt2,
+                     const MemOperand& src) {
+  LoadStorePairNonTemporal(rt, rt2, src,
+                           LoadPairNonTemporalOpFor(rt, rt2));
+}
+
+
+void Assembler::stnp(const CPURegister& rt,
+                     const CPURegister& rt2,
+                     const MemOperand& dst) {
+  LoadStorePairNonTemporal(rt, rt2, dst,
+                           StorePairNonTemporalOpFor(rt, rt2));
+}
+
+
+void Assembler::LoadStorePairNonTemporal(const CPURegister& rt,
+                                         const CPURegister& rt2,
+                                         const MemOperand& addr,
+                                         LoadStorePairNonTemporalOp op) {
+  ASSERT(!rt.Is(rt2));
+  ASSERT(AreSameSizeAndType(rt, rt2));
+  ASSERT(addr.IsImmediateOffset());
+
+  LSDataSize size = CalcLSPairDataSize(
+    static_cast<LoadStorePairOp>(op & LoadStorePairMask));
+  Emit(op | Rt(rt) | Rt2(rt2) | RnSP(addr.base()) |
+       ImmLSPair(addr.offset(), size));
+}
+
+
+// Memory instructions.
+void Assembler::ldrb(const Register& rt, const MemOperand& src) {
+  LoadStore(rt, src, LDRB_w);
+}
+
+
+void Assembler::strb(const Register& rt, const MemOperand& dst) {
+  LoadStore(rt, dst, STRB_w);
+}
+
+
+void Assembler::ldrsb(const Register& rt, const MemOperand& src) {
+  LoadStore(rt, src, rt.Is64Bits() ? LDRSB_x : LDRSB_w);
+}
+
+
+void Assembler::ldrh(const Register& rt, const MemOperand& src) {
+  LoadStore(rt, src, LDRH_w);
+}
+
+
+void Assembler::strh(const Register& rt, const MemOperand& dst) {
+  LoadStore(rt, dst, STRH_w);
+}
+
+
+void Assembler::ldrsh(const Register& rt, const MemOperand& src) {
+  LoadStore(rt, src, rt.Is64Bits() ? LDRSH_x : LDRSH_w);
+}
+
+
+void Assembler::ldr(const CPURegister& rt, const MemOperand& src) {
+  LoadStore(rt, src, LoadOpFor(rt));
+}
+
+
+void Assembler::str(const CPURegister& rt, const MemOperand& src) {
+  LoadStore(rt, src, StoreOpFor(rt));
+}
+
+
+void Assembler::ldrsw(const Register& rt, const MemOperand& src) {
+  ASSERT(rt.Is64Bits());
+  LoadStore(rt, src, LDRSW_x);
+}
+
+
+void Assembler::ldr(const Register& rt, uint64_t imm) {
+  LoadLiteral(rt, imm, rt.Is64Bits() ? LDR_x_lit : LDR_w_lit);
+}
+
+
+void Assembler::ldr(const FPRegister& ft, double imm) {
+  uint64_t rawbits = 0;
+  LoadLiteralOp op;
+
+  if (ft.Is64Bits()) {
+    rawbits = double_to_rawbits(imm);
+    op = LDR_d_lit;
+  } else {
+    ASSERT(ft.Is32Bits());
+    float float_imm = static_cast<float>(imm);
+    rawbits = float_to_rawbits(float_imm);
+    op = LDR_s_lit;
+  }
+
+  LoadLiteral(ft, rawbits, op);
+}
+
+
+void Assembler::mov(const Register& rd, const Register& rm) {
+  // Moves involving the stack pointer are encoded as add immediate with
+  // second operand of zero. Otherwise, orr with first operand zr is
+  // used.
+  if (rd.IsSP() || rm.IsSP()) {
+    add(rd, rm, 0);
+  } else {
+    orr(rd, AppropriateZeroRegFor(rd), rm);
+  }
+}
+
+
+void Assembler::mvn(const Register& rd, const Operand& operand) {
+  orn(rd, AppropriateZeroRegFor(rd), operand);
+}
+
+
+void Assembler::mrs(const Register& rt, SystemRegister sysreg) {
+  ASSERT(rt.Is64Bits());
+  Emit(MRS | ImmSystemRegister(sysreg) | Rt(rt));
+}
+
+
+void Assembler::msr(SystemRegister sysreg, const Register& rt) {
+  ASSERT(rt.Is64Bits());
+  Emit(MSR | Rt(rt) | ImmSystemRegister(sysreg));
+}
+
+
+void Assembler::hint(SystemHint code) {
+  Emit(HINT | ImmHint(code) | Rt(xzr));
+}
+
+
+void Assembler::fmov(FPRegister fd, double imm) {
+  if (fd.Is64Bits() && IsImmFP64(imm)) {
+    Emit(FMOV_d_imm | Rd(fd) | ImmFP64(imm));
+  } else if (fd.Is32Bits() && IsImmFP32(imm)) {
+    Emit(FMOV_s_imm | Rd(fd) | ImmFP32(static_cast<float>(imm)));
+  } else if ((imm == 0.0) && (copysign(1.0, imm) == 1.0)) {
+    Register zr = AppropriateZeroRegFor(fd);
+    fmov(fd, zr);
+  } else {
+    ldr(fd, imm);
+  }
+}
+
+
+void Assembler::fmov(Register rd, FPRegister fn) {
+  ASSERT(rd.size() == fn.size());
+  FPIntegerConvertOp op = rd.Is32Bits() ? FMOV_ws : FMOV_xd;
+  Emit(op | Rd(rd) | Rn(fn));
+}
+
+
+void Assembler::fmov(FPRegister fd, Register rn) {
+  ASSERT(fd.size() == rn.size());
+  FPIntegerConvertOp op = fd.Is32Bits() ? FMOV_sw : FMOV_dx;
+  Emit(op | Rd(fd) | Rn(rn));
+}
+
+
+void Assembler::fmov(FPRegister fd, FPRegister fn) {
+  ASSERT(fd.size() == fn.size());
+  Emit(FPType(fd) | FMOV | Rd(fd) | Rn(fn));
+}
+
+
+void Assembler::fadd(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FADD);
+}
+
+
+void Assembler::fsub(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FSUB);
+}
+
+
+void Assembler::fmul(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FMUL);
+}
+
+
+void Assembler::fmsub(const FPRegister& fd,
+                      const FPRegister& fn,
+                      const FPRegister& fm,
+                      const FPRegister& fa) {
+  FPDataProcessing3Source(fd, fn, fm, fa, fd.Is32Bits() ? FMSUB_s : FMSUB_d);
+}
+
+
+void Assembler::fdiv(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FDIV);
+}
+
+
+void Assembler::fmax(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FMAX);
+}
+
+
+void Assembler::fmin(const FPRegister& fd,
+                     const FPRegister& fn,
+                     const FPRegister& fm) {
+  FPDataProcessing2Source(fd, fn, fm, FMIN);
+}
+
+
+void Assembler::fabs(const FPRegister& fd,
+                     const FPRegister& fn) {
+  ASSERT(fd.SizeInBits() == fn.SizeInBits());
+  FPDataProcessing1Source(fd, fn, FABS);
+}
+
+
+void Assembler::fneg(const FPRegister& fd,
+                     const FPRegister& fn) {
+  ASSERT(fd.SizeInBits() == fn.SizeInBits());
+  FPDataProcessing1Source(fd, fn, FNEG);
+}
+
+
+void Assembler::fsqrt(const FPRegister& fd,
+                      const FPRegister& fn) {
+  ASSERT(fd.SizeInBits() == fn.SizeInBits());
+  FPDataProcessing1Source(fd, fn, FSQRT);
+}
+
+
+void Assembler::frintn(const FPRegister& fd,
+                       const FPRegister& fn) {
+  ASSERT(fd.SizeInBits() == fn.SizeInBits());
+  FPDataProcessing1Source(fd, fn, FRINTN);
+}
+
+
+void Assembler::frintz(const FPRegister& fd,
+                       const FPRegister& fn) {
+  ASSERT(fd.SizeInBits() == fn.SizeInBits());
+  FPDataProcessing1Source(fd, fn, FRINTZ);
+}
+
+
+void Assembler::fcmp(const FPRegister& fn,
+                     const FPRegister& fm) {
+  ASSERT(fn.size() == fm.size());
+  Emit(FPType(fn) | FCMP | Rm(fm) | Rn(fn));
+}
+
+
+void Assembler::fcmp(const FPRegister& fn,
+                     double value) {
+  USE(value);
+  // Although the fcmp instruction can strictly only take an immediate value of
+  // +0.0, we don't need to check for -0.0 because the sign of 0.0 doesn't
+  // affect the result of the comparison.
+  ASSERT(value == 0.0);
+  Emit(FPType(fn) | FCMP_zero | Rn(fn));
+}
+
+
+void Assembler::fccmp(const FPRegister& fn,
+                      const FPRegister& fm,
+                      StatusFlags nzcv,
+                      Condition cond) {
+  ASSERT(fn.size() == fm.size());
+  Emit(FPType(fn) | FCCMP | Rm(fm) | Cond(cond) | Rn(fn) | Nzcv(nzcv));
+}
+
+
+void Assembler::fcsel(const FPRegister& fd,
+                      const FPRegister& fn,
+                      const FPRegister& fm,
+                      Condition cond) {
+  ASSERT(fd.size() == fn.size());
+  ASSERT(fd.size() == fm.size());
+  Emit(FPType(fd) | FCSEL | Rm(fm) | Cond(cond) | Rn(fn) | Rd(fd));
+}
+
+
+void Assembler::FPConvertToInt(const Register& rd,
+                               const FPRegister& fn,
+                               FPIntegerConvertOp op) {
+  Emit(SF(rd) | FPType(fn) | op | Rn(fn) | Rd(rd));
+}
+
+
+void Assembler::fcvt(const FPRegister& fd,
+                     const FPRegister& fn) {
+  if (fd.Is64Bits()) {
+    // Convert float to double.
+    ASSERT(fn.Is32Bits());
+    FPDataProcessing1Source(fd, fn, FCVT_ds);
+  } else {
+    // Convert double to float.
+    ASSERT(fn.Is64Bits());
+    FPDataProcessing1Source(fd, fn, FCVT_sd);
+  }
+}
+
+
+void Assembler::fcvtmu(const Register& rd, const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTMU);
+}
+
+
+void Assembler::fcvtms(const Register& rd, const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTMS);
+}
+
+
+void Assembler::fcvtnu(const Register& rd,
+                       const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTNU);
+}
+
+
+void Assembler::fcvtns(const Register& rd,
+                       const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTNS);
+}
+
+
+void Assembler::fcvtzu(const Register& rd,
+                       const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTZU);
+}
+
+
+void Assembler::fcvtzs(const Register& rd,
+                       const FPRegister& fn) {
+  FPConvertToInt(rd, fn, FCVTZS);
+}
+
+
+void Assembler::scvtf(const FPRegister& fd,
+                      const Register& rn,
+                      unsigned fbits) {
+  if (fbits == 0) {
+    Emit(SF(rn) | FPType(fd) | SCVTF | Rn(rn) | Rd(fd));
+  } else {
+    Emit(SF(rn) | FPType(fd) | SCVTF_fixed | FPScale(64 - fbits) | Rn(rn) |
+         Rd(fd));
+  }
+}
+
+
+void Assembler::ucvtf(const FPRegister& fd,
+                      const Register& rn,
+                      unsigned fbits) {
+  if (fbits == 0) {
+    Emit(SF(rn) | FPType(fd) | UCVTF | Rn(rn) | Rd(fd));
+  } else {
+    Emit(SF(rn) | FPType(fd) | UCVTF_fixed | FPScale(64 - fbits) | Rn(rn) |
+         Rd(fd));
+  }
+}
+
+
+// Note:
+// Below, a difference in case for the same letter indicates a
+// negated bit.
+// If b is 1, then B is 0.
+Instr Assembler::ImmFP32(float imm) {
+  ASSERT(IsImmFP32(imm));
+  // bits: aBbb.bbbc.defg.h000.0000.0000.0000.0000
+  uint32_t bits = float_to_rawbits(imm);
+  // bit7: a000.0000
+  uint32_t bit7 = ((bits >> 31) & 0x1) << 7;
+  // bit6: 0b00.0000
+  uint32_t bit6 = ((bits >> 29) & 0x1) << 6;
+  // bit5_to_0: 00cd.efgh
+  uint32_t bit5_to_0 = (bits >> 19) & 0x3f;
+
+  return (bit7 | bit6 | bit5_to_0) << ImmFP_offset;
+}
+
+
+Instr Assembler::ImmFP64(double imm) {
+  ASSERT(IsImmFP64(imm));
+  // bits: aBbb.bbbb.bbcd.efgh.0000.0000.0000.0000
+  //       0000.0000.0000.0000.0000.0000.0000.0000
+  uint64_t bits = double_to_rawbits(imm);
+  // bit7: a000.0000
+  uint32_t bit7 = ((bits >> 63) & 0x1) << 7;
+  // bit6: 0b00.0000
+  uint32_t bit6 = ((bits >> 61) & 0x1) << 6;
+  // bit5_to_0: 00cd.efgh
+  uint32_t bit5_to_0 = (bits >> 48) & 0x3f;
+
+  return (bit7 | bit6 | bit5_to_0) << ImmFP_offset;
+}
+
+
+// Code generation helpers.
+void Assembler::MoveWide(const Register& rd,
+                         uint64_t imm,
+                         int shift,
+                         MoveWideImmediateOp mov_op) {
+  if (shift >= 0) {
+    // Explicit shift specified.
+    ASSERT((shift == 0) || (shift == 16) || (shift == 32) || (shift == 48));
+    ASSERT(rd.Is64Bits() || (shift == 0) || (shift == 16));
+    shift /= 16;
+  } else {
+    // Calculate a new immediate and shift combination to encode the immediate
+    // argument.
+    shift = 0;
+    if ((imm & ~0xffffUL) == 0) {
+      // Nothing to do.
+    } else if ((imm & ~(0xffffUL << 16)) == 0) {
+      imm >>= 16;
+      shift = 1;
+    } else if ((imm & ~(0xffffUL << 32)) == 0) {
+      ASSERT(rd.Is64Bits());
+      imm >>= 32;
+      shift = 2;
+    } else if ((imm & ~(0xffffUL << 48)) == 0) {
+      ASSERT(rd.Is64Bits());
+      imm >>= 48;
+      shift = 3;
+    }
+  }
+
+  ASSERT(is_uint16(imm));
+
+  Emit(SF(rd) | MoveWideImmediateFixed | mov_op |
+       Rd(rd) | ImmMoveWide(imm) | ShiftMoveWide(shift));
+}
+
+
+void Assembler::AddSub(const Register& rd,
+                       const Register& rn,
+                       const Operand& operand,
+                       FlagsUpdate S,
+                       AddSubOp op) {
+  ASSERT(rd.size() == rn.size());
+  if (operand.IsImmediate()) {
+    int64_t immediate = operand.immediate();
+    ASSERT(IsImmAddSub(immediate));
+    Instr dest_reg = (S == SetFlags) ? Rd(rd) : RdSP(rd);
+    Emit(SF(rd) | AddSubImmediateFixed | op | Flags(S) |
+         ImmAddSub(immediate) | dest_reg | RnSP(rn));
+  } else if (operand.IsShiftedRegister()) {
+    ASSERT(operand.reg().size() == rd.size());
+    ASSERT(operand.shift() != ROR);
+
+    // For instructions of the form:
+    //   add/sub   wsp, <Wn>, <Wm> [, LSL #0-3 ]
+    //   add/sub   <Wd>, wsp, <Wm> [, LSL #0-3 ]
+    //   add/sub   wsp, wsp, <Wm> [, LSL #0-3 ]
+    //   adds/subs <Wd>, wsp, <Wm> [, LSL #0-3 ]
+    // or their 64-bit register equivalents, convert the operand from shifted to
+    // extended register mode, and emit an add/sub extended instruction.
+    if (rn.IsSP() || rd.IsSP()) {
+      ASSERT(!(rd.IsSP() && (S == SetFlags)));
+      DataProcExtendedRegister(rd, rn, operand.ToExtendedRegister(), S,
+                               AddSubExtendedFixed | op);
+    } else {
+      DataProcShiftedRegister(rd, rn, operand, S, AddSubShiftedFixed | op);
+    }
+  } else {
+    ASSERT(operand.IsExtendedRegister());
+    DataProcExtendedRegister(rd, rn, operand, S, AddSubExtendedFixed | op);
+  }
+}
+
+
+void Assembler::AddSubWithCarry(const Register& rd,
+                                const Register& rn,
+                                const Operand& operand,
+                                FlagsUpdate S,
+                                AddSubWithCarryOp op) {
+  ASSERT(rd.size() == rn.size());
+  ASSERT(rd.size() == operand.reg().size());
+  ASSERT(operand.IsShiftedRegister() && (operand.shift_amount() == 0));
+  Emit(SF(rd) | op | Flags(S) | Rm(operand.reg()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::hlt(int code) {
+  ASSERT(is_uint16(code));
+  Emit(HLT | ImmException(code));
+}
+
+
+void Assembler::brk(int code) {
+  ASSERT(is_uint16(code));
+  Emit(BRK | ImmException(code));
+}
+
+
+void Assembler::Logical(const Register& rd,
+                        const Register& rn,
+                        const Operand& operand,
+                        LogicalOp op) {
+  ASSERT(rd.size() == rn.size());
+  if (operand.IsImmediate()) {
+    int64_t immediate = operand.immediate();
+    unsigned reg_size = rd.size();
+
+    ASSERT(immediate != 0);
+    ASSERT(immediate != -1);
+    ASSERT(rd.Is64Bits() || is_uint32(immediate));
+
+    // If the operation is NOT, invert the operation and immediate.
+    if ((op & NOT) == NOT) {
+      op = static_cast<LogicalOp>(op & ~NOT);
+      immediate = rd.Is64Bits() ? ~immediate : (~immediate & kWRegMask);
+    }
+
+    unsigned n, imm_s, imm_r;
+    if (IsImmLogical(immediate, reg_size, &n, &imm_s, &imm_r)) {
+      // Immediate can be encoded in the instruction.
+      LogicalImmediate(rd, rn, n, imm_s, imm_r, op);
+    } else {
+      // This case is handled in the macro assembler.
+      UNREACHABLE();
+    }
+  } else {
+    ASSERT(operand.IsShiftedRegister());
+    ASSERT(operand.reg().size() == rd.size());
+    Instr dp_op = static_cast<Instr>(op | LogicalShiftedFixed);
+    DataProcShiftedRegister(rd, rn, operand, LeaveFlags, dp_op);
+  }
+}
+
+
+void Assembler::LogicalImmediate(const Register& rd,
+                                 const Register& rn,
+                                 unsigned n,
+                                 unsigned imm_s,
+                                 unsigned imm_r,
+                                 LogicalOp op) {
+  unsigned reg_size = rd.size();
+  Instr dest_reg = (op == ANDS) ? Rd(rd) : RdSP(rd);
+  Emit(SF(rd) | LogicalImmediateFixed | op | BitN(n, reg_size) |
+       ImmSetBits(imm_s, reg_size) | ImmRotate(imm_r, reg_size) | dest_reg |
+       Rn(rn));
+}
+
+
+void Assembler::ConditionalCompare(const Register& rn,
+                                   const Operand& operand,
+                                   StatusFlags nzcv,
+                                   Condition cond,
+                                   ConditionalCompareOp op) {
+  Instr ccmpop;
+  if (operand.IsImmediate()) {
+    int64_t immediate = operand.immediate();
+    ASSERT(IsImmConditionalCompare(immediate));
+    ccmpop = ConditionalCompareImmediateFixed | op | ImmCondCmp(immediate);
+  } else {
+    ASSERT(operand.IsShiftedRegister() && (operand.shift_amount() == 0));
+    ccmpop = ConditionalCompareRegisterFixed | op | Rm(operand.reg());
+  }
+  Emit(SF(rn) | ccmpop | Cond(cond) | Rn(rn) | Nzcv(nzcv));
+}
+
+
+void Assembler::DataProcessing1Source(const Register& rd,
+                                      const Register& rn,
+                                      DataProcessing1SourceOp op) {
+  ASSERT(rd.size() == rn.size());
+  Emit(SF(rn) | op | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::FPDataProcessing1Source(const FPRegister& fd,
+                                        const FPRegister& fn,
+                                        FPDataProcessing1SourceOp op) {
+  Emit(FPType(fn) | op | Rn(fn) | Rd(fd));
+}
+
+
+void Assembler::FPDataProcessing2Source(const FPRegister& fd,
+                                        const FPRegister& fn,
+                                        const FPRegister& fm,
+                                        FPDataProcessing2SourceOp op) {
+  ASSERT(fd.size() == fn.size());
+  ASSERT(fd.size() == fm.size());
+  Emit(FPType(fd) | op | Rm(fm) | Rn(fn) | Rd(fd));
+}
+
+
+void Assembler::FPDataProcessing3Source(const FPRegister& fd,
+                                        const FPRegister& fn,
+                                        const FPRegister& fm,
+                                        const FPRegister& fa,
+                                        FPDataProcessing3SourceOp op) {
+  ASSERT(AreSameSizeAndType(fd, fn, fm, fa));
+  Emit(FPType(fd) | op | Rm(fm) | Rn(fn) | Rd(fd) | Ra(fa));
+}
+
+
+void Assembler::EmitShift(const Register& rd,
+                          const Register& rn,
+                          Shift shift,
+                          unsigned shift_amount) {
+  switch (shift) {
+    case LSL:
+      lsl(rd, rn, shift_amount);
+      break;
+    case LSR:
+      lsr(rd, rn, shift_amount);
+      break;
+    case ASR:
+      asr(rd, rn, shift_amount);
+      break;
+    case ROR:
+      ror(rd, rn, shift_amount);
+      break;
+    default:
+      UNREACHABLE();
+  }
+}
+
+
+void Assembler::EmitExtendShift(const Register& rd,
+                                const Register& rn,
+                                Extend extend,
+                                unsigned left_shift) {
+  ASSERT(rd.size() >= rn.size());
+  unsigned reg_size = rd.size();
+  // Use the correct size of register.
+  Register rn_ = Register(rn.code(), rd.size());
+  // Bits extracted are high_bit:0.
+  unsigned high_bit = (8 << (extend & 0x3)) - 1;
+  // Number of bits left in the result that are not introduced by the shift.
+  unsigned non_shift_bits = (reg_size - left_shift) & (reg_size - 1);
+
+  if ((non_shift_bits > high_bit) || (non_shift_bits == 0)) {
+    switch (extend) {
+      case UXTB:
+      case UXTH:
+      case UXTW: ubfm(rd, rn_, non_shift_bits, high_bit); break;
+      case SXTB:
+      case SXTH:
+      case SXTW: sbfm(rd, rn_, non_shift_bits, high_bit); break;
+      case UXTX:
+      case SXTX: {
+        ASSERT(rn.size() == kXRegSize);
+        // Nothing to extend. Just shift.
+        lsl(rd, rn_, left_shift);
+        break;
+      }
+      default: UNREACHABLE();
+    }
+  } else {
+    // No need to extend as the extended bits would be shifted away.
+    lsl(rd, rn_, left_shift);
+  }
+}
+
+
+void Assembler::DataProcShiftedRegister(const Register& rd,
+                                        const Register& rn,
+                                        const Operand& operand,
+                                        FlagsUpdate S,
+                                        Instr op) {
+  ASSERT(operand.IsShiftedRegister());
+  ASSERT(rn.Is64Bits() || (rn.Is32Bits() && is_uint5(operand.shift_amount())));
+  Emit(SF(rd) | op | Flags(S) |
+       ShiftDP(operand.shift()) | ImmDPShift(operand.shift_amount()) |
+       Rm(operand.reg()) | Rn(rn) | Rd(rd));
+}
+
+
+void Assembler::DataProcExtendedRegister(const Register& rd,
+                                         const Register& rn,
+                                         const Operand& operand,
+                                         FlagsUpdate S,
+                                         Instr op) {
+  Instr dest_reg = (S == SetFlags) ? Rd(rd) : RdSP(rd);
+  Emit(SF(rd) | op | Flags(S) | Rm(operand.reg()) |
+       ExtendMode(operand.extend()) | ImmExtendShift(operand.shift_amount()) |
+       dest_reg | RnSP(rn));
+}
+
+
+bool Assembler::IsImmAddSub(int64_t immediate) {
+  return is_uint12(immediate) ||
+         (is_uint12(immediate >> 12) && ((immediate & 0xfff) == 0));
+}
+
+void Assembler::LoadStore(const CPURegister& rt,
+                          const MemOperand& addr,
+                          LoadStoreOp op) {
+  Instr memop = op | Rt(rt) | RnSP(addr.base());
+  ptrdiff_t offset = addr.offset();
+
+  if (addr.IsImmediateOffset()) {
+    LSDataSize size = CalcLSDataSize(op);
+    if (IsImmLSScaled(offset, size)) {
+      // Use the scaled addressing mode.
+      Emit(LoadStoreUnsignedOffsetFixed | memop |
+           ImmLSUnsigned(offset >> size));
+    } else if (IsImmLSUnscaled(offset)) {
+      // Use the unscaled addressing mode.
+      Emit(LoadStoreUnscaledOffsetFixed | memop | ImmLS(offset));
+    } else {
+      // This case is handled in the macro assembler.
+      UNREACHABLE();
+    }
+  } else if (addr.IsRegisterOffset()) {
+    Extend ext = addr.extend();
+    Shift shift = addr.shift();
+    unsigned shift_amount = addr.shift_amount();
+
+    // LSL is encoded in the option field as UXTX.
+    if (shift == LSL) {
+      ext = UXTX;
+    }
+
+    // Shifts are encoded in one bit, indicating a left shift by the memory
+    // access size.
+    ASSERT((shift_amount == 0) ||
+           (shift_amount == static_cast<unsigned>(CalcLSDataSize(op))));
+    Emit(LoadStoreRegisterOffsetFixed | memop | Rm(addr.regoffset()) |
+         ExtendMode(ext) | ImmShiftLS((shift_amount > 0) ? 1 : 0));
+  } else {
+    if (IsImmLSUnscaled(offset)) {
+      if (addr.IsPreIndex()) {
+        Emit(LoadStorePreIndexFixed | memop | ImmLS(offset));
+      } else {
+        ASSERT(addr.IsPostIndex());
+        Emit(LoadStorePostIndexFixed | memop | ImmLS(offset));
+      }
+    } else {
+      // This case is handled in the macro assembler.
+      UNREACHABLE();
+    }
+  }
+}
+
+
+bool Assembler::IsImmLSUnscaled(ptrdiff_t offset) {
+  return is_int9(offset);
+}
+
+
+bool Assembler::IsImmLSScaled(ptrdiff_t offset, LSDataSize size) {
+  bool offset_is_size_multiple = (((offset >> size) << size) == offset);
+  return offset_is_size_multiple && is_uint12(offset >> size);
+}
+
+
+void Assembler::LoadLiteral(const CPURegister& rt,
+                            uint64_t imm,
+                            LoadLiteralOp op) {
+  ASSERT(is_int32(imm) || is_uint32(imm) || (rt.Is64Bits()));
+
+  BlockLiteralPoolScope scope(this);
+  RecordLiteral(imm, rt.SizeInBytes());
+  Emit(op | ImmLLiteral(0) | Rt(rt));
+}
+
+
+// Test if a given value can be encoded in the immediate field of a logical
+// instruction.
+// If it can be encoded, the function returns true, and values pointed to by n,
+// imm_s and imm_r are updated with immediates encoded in the format required
+// by the corresponding fields in the logical instruction.
+// If it can not be encoded, the function returns false, and the values pointed
+// to by n, imm_s and imm_r are undefined.
+bool Assembler::IsImmLogical(uint64_t value,
+                             unsigned width,
+                             unsigned* n,
+                             unsigned* imm_s,
+                             unsigned* imm_r) {
+  ASSERT((n != NULL) && (imm_s != NULL) && (imm_r != NULL));
+  ASSERT((width == kWRegSize) || (width == kXRegSize));
+
+  // Logical immediates are encoded using parameters n, imm_s and imm_r using
+  // the following table:
+  //
+  //  N   imms    immr    size        S             R
+  //  1  ssssss  rrrrrr    64    UInt(ssssss)  UInt(rrrrrr)
+  //  0  0sssss  xrrrrr    32    UInt(sssss)   UInt(rrrrr)
+  //  0  10ssss  xxrrrr    16    UInt(ssss)    UInt(rrrr)
+  //  0  110sss  xxxrrr     8    UInt(sss)     UInt(rrr)
+  //  0  1110ss  xxxxrr     4    UInt(ss)      UInt(rr)
+  //  0  11110s  xxxxxr     2    UInt(s)       UInt(r)
+  // (s bits must not be all set)
+  //
+  // A pattern is constructed of size bits, where the least significant S+1
+  // bits are set. The pattern is rotated right by R, and repeated across a
+  // 32 or 64-bit value, depending on destination register width.
+  //
+  // To test if an arbitrary immediate can be encoded using this scheme, an
+  // iterative algorithm is used.
+  //
+  // TODO: This code does not consider using X/W register overlap to support
+  // 64-bit immediates where the top 32-bits are zero, and the bottom 32-bits
+  // are an encodable logical immediate.
+
+  // 1. If the value has all set or all clear bits, it can't be encoded.
+  if ((value == 0) || (value == 0xffffffffffffffffUL) ||
+      ((width == kWRegSize) && (value == 0xffffffff))) {
+    return false;
+  }
+
+  unsigned lead_zero = CountLeadingZeros(value, width);
+  unsigned lead_one = CountLeadingZeros(~value, width);
+  unsigned trail_zero = CountTrailingZeros(value, width);
+  unsigned trail_one = CountTrailingZeros(~value, width);
+  unsigned set_bits = CountSetBits(value, width);
+
+  // The fixed bits in the immediate s field.
+  // If width == 64 (X reg), start at 0xFFFFFF80.
+  // If width == 32 (W reg), start at 0xFFFFFFC0, as the iteration for 64-bit
+  // widths won't be executed.
+  int imm_s_fixed = (width == kXRegSize) ? -128 : -64;
+  int imm_s_mask = 0x3F;
+
+  for (;;) {
+    // 2. If the value is two bits wide, it can be encoded.
+    if (width == 2) {
+      *n = 0;
+      *imm_s = 0x3C;
+      *imm_r = (value & 3) - 1;
+      return true;
+    }
+
+    *n = (width == 64) ? 1 : 0;
+    *imm_s = ((imm_s_fixed | (set_bits - 1)) & imm_s_mask);
+    if ((lead_zero + set_bits) == width) {
+      *imm_r = 0;
+    } else {
+      *imm_r = (lead_zero > 0) ? (width - trail_zero) : lead_one;
+    }
+
+    // 3. If the sum of leading zeros, trailing zeros and set bits is equal to
+    //    the bit width of the value, it can be encoded.
+    if (lead_zero + trail_zero + set_bits == width) {
+      return true;
+    }
+
+    // 4. If the sum of leading ones, trailing ones and unset bits in the
+    //    value is equal to the bit width of the value, it can be encoded.
+    if (lead_one + trail_one + (width - set_bits) == width) {
+      return true;
+    }
+
+    // 5. If the most-significant half of the bitwise value is equal to the
+    //    least-significant half, return to step 2 using the least-significant
+    //    half of the value.
+    uint64_t mask = (1UL << (width >> 1)) - 1;
+    if ((value & mask) == ((value >> (width >> 1)) & mask)) {
+      width >>= 1;
+      set_bits >>= 1;
+      imm_s_fixed >>= 1;
+      continue;
+    }
+
+    // 6. Otherwise, the value can't be encoded.
+    return false;
+  }
+}
+
+bool Assembler::IsImmConditionalCompare(int64_t immediate) {
+  return is_uint5(immediate);
+}
+
+
+bool Assembler::IsImmFP32(float imm) {
+  // Valid values will have the form:
+  // aBbb.bbbc.defg.h000.0000.0000.0000.0000
+  uint32_t bits = float_to_rawbits(imm);
+  // bits[19..0] are cleared.
+  if ((bits & 0x7ffff) != 0) {
+    return false;
+  }
+
+  // bits[29..25] are all set or all cleared.
+  uint32_t b_pattern = (bits >> 16) & 0x3e00;
+  if (b_pattern != 0 && b_pattern != 0x3e00) {
+    return false;
+  }
+
+  // bit[30] and bit[29] are opposite.
+  if (((bits ^ (bits << 1)) & 0x40000000) == 0) {
+    return false;
+  }
+
+  return true;
+}
+
+
+bool Assembler::IsImmFP64(double imm) {
+  // Valid values will have the form:
+  // aBbb.bbbb.bbcd.efgh.0000.0000.0000.0000
+  // 0000.0000.0000.0000.0000.0000.0000.0000
+  uint64_t bits = double_to_rawbits(imm);
+  // bits[47..0] are cleared.
+  if ((bits & 0xffffffffffffL) != 0) {
+    return false;
+  }
+
+  // bits[61..54] are all set or all cleared.
+  uint32_t b_pattern = (bits >> 48) & 0x3fc0;
+  if (b_pattern != 0 && b_pattern != 0x3fc0) {
+    return false;
+  }
+
+  // bit[62] and bit[61] are opposite.
+  if (((bits ^ (bits << 1)) & 0x4000000000000000L) == 0) {
+    return false;
+  }
+
+  return true;
+}
+
+
+LoadStoreOp Assembler::LoadOpFor(const CPURegister& rt) {
+  ASSERT(rt.IsValid());
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? LDR_x : LDR_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? LDR_d : LDR_s;
+  }
+}
+
+
+LoadStorePairOp Assembler::LoadPairOpFor(const CPURegister& rt,
+    const CPURegister& rt2) {
+  ASSERT(AreSameSizeAndType(rt, rt2));
+  USE(rt2);
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? LDP_x : LDP_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? LDP_d : LDP_s;
+  }
+}
+
+
+LoadStoreOp Assembler::StoreOpFor(const CPURegister& rt) {
+  ASSERT(rt.IsValid());
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? STR_x : STR_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? STR_d : STR_s;
+  }
+}
+
+
+LoadStorePairOp Assembler::StorePairOpFor(const CPURegister& rt,
+    const CPURegister& rt2) {
+  ASSERT(AreSameSizeAndType(rt, rt2));
+  USE(rt2);
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? STP_x : STP_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? STP_d : STP_s;
+  }
+}
+
+
+LoadStorePairNonTemporalOp Assembler::LoadPairNonTemporalOpFor(
+    const CPURegister& rt, const CPURegister& rt2) {
+  ASSERT(AreSameSizeAndType(rt, rt2));
+  USE(rt2);
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? LDNP_x : LDNP_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? LDNP_d : LDNP_s;
+  }
+}
+
+
+LoadStorePairNonTemporalOp Assembler::StorePairNonTemporalOpFor(
+    const CPURegister& rt, const CPURegister& rt2) {
+  ASSERT(AreSameSizeAndType(rt, rt2));
+  USE(rt2);
+  if (rt.IsRegister()) {
+    return rt.Is64Bits() ? STNP_x : STNP_w;
+  } else {
+    ASSERT(rt.IsFPRegister());
+    return rt.Is64Bits() ? STNP_d : STNP_s;
+  }
+}
+
+
+void Assembler::RecordLiteral(int64_t imm, unsigned size) {
+  literals_.push_front(new Literal(pc_, imm, size));
+}
+
+
+// Check if a literal pool should be emitted. Currently a literal is emitted
+// when:
+//  * the distance to the first literal load handled by this pool is greater
+//    than the recommended distance and the literal pool can be emitted without
+//    generating a jump over it.
+//  * the distance to the first literal load handled by this pool is greater
+//    than twice the recommended distance.
+// TODO: refine this heuristic using real world data.
+void Assembler::CheckLiteralPool(LiteralPoolEmitOption option) {
+  if (IsLiteralPoolBlocked()) {
+    // Literal pool emission is forbidden, no point in doing further checks.
+    return;
+  }
+
+  if (literals_.empty()) {
+    // No literal pool to emit.
+    next_literal_pool_check_ += kLiteralPoolCheckInterval;
+    return;
+  }
+
+  intptr_t distance = pc_ - literals_.back()->pc_;
+  if ((distance < kRecommendedLiteralPoolRange) ||
+      ((option == JumpRequired) &&
+       (distance < (2 * kRecommendedLiteralPoolRange)))) {
+    // We prefer not to have to jump over the literal pool.
+    next_literal_pool_check_ += kLiteralPoolCheckInterval;
+    return;
+  }
+
+  EmitLiteralPool(option);
+}
+
+
+void Assembler::EmitLiteralPool(LiteralPoolEmitOption option) {
+  // Prevent recursive calls while emitting the literal pool.
+  BlockLiteralPoolScope scope(this);
+
+  Label marker;
+  Label start_of_pool;
+  Label end_of_pool;
+
+  if (option == JumpRequired) {
+    b(&end_of_pool);
+  }
+
+  // Leave space for a literal pool marker. This is populated later, once the
+  // size of the pool is known.
+  bind(&marker);
+  nop();
+
+  // Now populate the literal pool.
+  bind(&start_of_pool);
+  std::list<Literal*>::iterator it;
+  for (it = literals_.begin(); it != literals_.end(); it++) {
+    // Update the load-literal instruction to point to this pool entry.
+    Instruction* load_literal = (*it)->pc_;
+    load_literal->SetImmLLiteral(pc_);
+    // Copy the data into the pool.
+    uint64_t value= (*it)->value_;
+    unsigned size = (*it)->size_;
+    ASSERT((size == kXRegSizeInBytes) || (size == kWRegSizeInBytes));
+    ASSERT((pc_ + size) <= (buffer_ + buffer_size_));
+    memcpy(pc_, &value, size);
+    pc_ += size;
+    delete *it;
+  }
+  literals_.clear();
+  bind(&end_of_pool);
+
+  // The pool size should always be a multiple of four bytes because that is the
+  // scaling applied by the LDR(literal) instruction, even for X-register loads.
+  ASSERT((SizeOfCodeGeneratedSince(&start_of_pool) % 4) == 0);
+  uint64_t pool_size = SizeOfCodeGeneratedSince(&start_of_pool) / 4;
+
+  // Literal pool marker indicating the size in words of the literal pool.
+  // We use a literal load to the zero register, the offset indicating the
+  // size in words. This instruction can encode a large enough offset to span
+  // the entire pool at its maximum size.
+  Instr marker_instruction = LDR_x_lit | ImmLLiteral(pool_size) | Rt(xzr);
+  memcpy(marker.target(), &marker_instruction, kInstructionSize);
+
+  next_literal_pool_check_ = pc_ + kLiteralPoolCheckInterval;
+}
+
+
+// Return the size in bytes, required by the literal pool entries. This does
+// not include any marker or branch over the literal pool itself.
+size_t Assembler::LiteralPoolSize() {
+  size_t size = 0;
+
+  std::list<Literal*>::iterator it;
+  for (it = literals_.begin(); it != literals_.end(); it++) {
+    size += (*it)->size_;
+  }
+
+  return size;
+}
+
+
+bool AreAliased(const CPURegister& reg1, const CPURegister& reg2,
+                const CPURegister& reg3, const CPURegister& reg4,
+                const CPURegister& reg5, const CPURegister& reg6,
+                const CPURegister& reg7, const CPURegister& reg8) {
+  int number_of_valid_regs = 0;
+  int number_of_valid_fpregs = 0;
+
+  RegList unique_regs = 0;
+  RegList unique_fpregs = 0;
+
+  const CPURegister regs[] = {reg1, reg2, reg3, reg4, reg5, reg6, reg7, reg8};
+
+  for (unsigned i = 0; i < sizeof(regs) / sizeof(regs[0]); i++) {
+    if (regs[i].IsRegister()) {
+      number_of_valid_regs++;
+      unique_regs |= regs[i].Bit();
+    } else if (regs[i].IsFPRegister()) {
+      number_of_valid_fpregs++;
+      unique_fpregs |= regs[i].Bit();
+    } else {
+      ASSERT(!regs[i].IsValid());
+    }
+  }
+
+  int number_of_unique_regs =
+    CountSetBits(unique_regs, sizeof(unique_regs) * 8);
+  int number_of_unique_fpregs =
+    CountSetBits(unique_fpregs, sizeof(unique_fpregs) * 8);
+
+  ASSERT(number_of_valid_regs >= number_of_unique_regs);
+  ASSERT(number_of_valid_fpregs >= number_of_unique_fpregs);
+
+  return (number_of_valid_regs != number_of_unique_regs) ||
+         (number_of_valid_fpregs != number_of_unique_fpregs);
+}
+
+
+bool AreSameSizeAndType(const CPURegister& reg1, const CPURegister& reg2,
+                        const CPURegister& reg3, const CPURegister& reg4,
+                        const CPURegister& reg5, const CPURegister& reg6,
+                        const CPURegister& reg7, const CPURegister& reg8) {
+  ASSERT(reg1.IsValid());
+  bool match = true;
+  match &= !reg2.IsValid() || reg2.IsSameSizeAndType(reg1);
+  match &= !reg3.IsValid() || reg3.IsSameSizeAndType(reg1);
+  match &= !reg4.IsValid() || reg4.IsSameSizeAndType(reg1);
+  match &= !reg5.IsValid() || reg5.IsSameSizeAndType(reg1);
+  match &= !reg6.IsValid() || reg6.IsSameSizeAndType(reg1);
+  match &= !reg7.IsValid() || reg7.IsSameSizeAndType(reg1);
+  match &= !reg8.IsValid() || reg8.IsSameSizeAndType(reg1);
+  return match;
+}
+
+
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/assembler-a64.h b/disas/libvixl/src/a64/assembler-a64.h
new file mode 100644
index 0000000..93b3011
--- /dev/null
+++ b/disas/libvixl/src/a64/assembler-a64.h
@@ -0,0 +1,1784 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_ASSEMBLER_A64_H_
+#define VIXL_A64_ASSEMBLER_A64_H_
+
+#include <list>
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/instructions-a64.h"
+
+namespace vixl {
+
+typedef uint64_t RegList;
+static const int kRegListSizeInBits = sizeof(RegList) * 8;
+
+// Registers.
+
+// Some CPURegister methods can return Register and FPRegister types, so we
+// need to declare them in advance.
+class Register;
+class FPRegister;
+
+
+class CPURegister {
+ public:
+  enum RegisterType {
+    // The kInvalid value is used to detect uninitialized static instances,
+    // which are always zero-initialized before any constructors are called.
+    kInvalid = 0,
+    kRegister,
+    kFPRegister,
+    kNoRegister
+  };
+
+  CPURegister() : code_(0), size_(0), type_(kNoRegister) {
+    ASSERT(!IsValid());
+    ASSERT(IsNone());
+  }
+
+  CPURegister(unsigned code, unsigned size, RegisterType type)
+      : code_(code), size_(size), type_(type) {
+    ASSERT(IsValidOrNone());
+  }
+
+  unsigned code() const {
+    ASSERT(IsValid());
+    return code_;
+  }
+
+  RegisterType type() const {
+    ASSERT(IsValidOrNone());
+    return type_;
+  }
+
+  RegList Bit() const {
+    ASSERT(code_ < (sizeof(RegList) * 8));
+    return IsValid() ? (static_cast<RegList>(1) << code_) : 0;
+  }
+
+  unsigned size() const {
+    ASSERT(IsValid());
+    return size_;
+  }
+
+  int SizeInBytes() const {
+    ASSERT(IsValid());
+    ASSERT(size() % 8 == 0);
+    return size_ / 8;
+  }
+
+  int SizeInBits() const {
+    ASSERT(IsValid());
+    return size_;
+  }
+
+  bool Is32Bits() const {
+    ASSERT(IsValid());
+    return size_ == 32;
+  }
+
+  bool Is64Bits() const {
+    ASSERT(IsValid());
+    return size_ == 64;
+  }
+
+  bool IsValid() const {
+    if (IsValidRegister() || IsValidFPRegister()) {
+      ASSERT(!IsNone());
+      return true;
+    } else {
+      ASSERT(IsNone());
+      return false;
+    }
+  }
+
+  bool IsValidRegister() const {
+    return IsRegister() &&
+           ((size_ == kWRegSize) || (size_ == kXRegSize)) &&
+           ((code_ < kNumberOfRegisters) || (code_ == kSPRegInternalCode));
+  }
+
+  bool IsValidFPRegister() const {
+    return IsFPRegister() &&
+           ((size_ == kSRegSize) || (size_ == kDRegSize)) &&
+           (code_ < kNumberOfFPRegisters);
+  }
+
+  bool IsNone() const {
+    // kNoRegister types should always have size 0 and code 0.
+    ASSERT((type_ != kNoRegister) || (code_ == 0));
+    ASSERT((type_ != kNoRegister) || (size_ == 0));
+
+    return type_ == kNoRegister;
+  }
+
+  bool Is(const CPURegister& other) const {
+    ASSERT(IsValidOrNone() && other.IsValidOrNone());
+    return (code_ == other.code_) && (size_ == other.size_) &&
+           (type_ == other.type_);
+  }
+
+  inline bool IsZero() const {
+    ASSERT(IsValid());
+    return IsRegister() && (code_ == kZeroRegCode);
+  }
+
+  inline bool IsSP() const {
+    ASSERT(IsValid());
+    return IsRegister() && (code_ == kSPRegInternalCode);
+  }
+
+  inline bool IsRegister() const {
+    return type_ == kRegister;
+  }
+
+  inline bool IsFPRegister() const {
+    return type_ == kFPRegister;
+  }
+
+  const Register& W() const;
+  const Register& X() const;
+  const FPRegister& S() const;
+  const FPRegister& D() const;
+
+  inline bool IsSameSizeAndType(const CPURegister& other) const {
+    return (size_ == other.size_) && (type_ == other.type_);
+  }
+
+ protected:
+  unsigned code_;
+  unsigned size_;
+  RegisterType type_;
+
+ private:
+  bool IsValidOrNone() const {
+    return IsValid() || IsNone();
+  }
+};
+
+
+class Register : public CPURegister {
+ public:
+  explicit Register() : CPURegister() {}
+  inline explicit Register(const CPURegister& other)
+      : CPURegister(other.code(), other.size(), other.type()) {
+    ASSERT(IsValidRegister());
+  }
+  explicit Register(unsigned code, unsigned size)
+      : CPURegister(code, size, kRegister) {}
+
+  bool IsValid() const {
+    ASSERT(IsRegister() || IsNone());
+    return IsValidRegister();
+  }
+
+  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[];
+};
+
+
+class FPRegister : public CPURegister {
+ public:
+  inline FPRegister() : CPURegister() {}
+  inline explicit FPRegister(const CPURegister& other)
+      : CPURegister(other.code(), other.size(), other.type()) {
+    ASSERT(IsValidFPRegister());
+  }
+  inline FPRegister(unsigned code, unsigned size)
+      : CPURegister(code, size, kFPRegister) {}
+
+  bool IsValid() const {
+    ASSERT(IsFPRegister() || IsNone());
+    return IsValidFPRegister();
+  }
+
+  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[];
+};
+
+
+// No*Reg is used to indicate an unused argument, or an error case. Note that
+// these all compare equal (using the Is() method). The Register and FPRegister
+// variants are provided for convenience.
+const Register NoReg;
+const FPRegister NoFPReg;
+const CPURegister NoCPUReg;
+
+
+#define DEFINE_REGISTERS(N)  \
+const Register w##N(N, kWRegSize);  \
+const Register x##N(N, kXRegSize);
+REGISTER_CODE_LIST(DEFINE_REGISTERS)
+#undef DEFINE_REGISTERS
+const Register wsp(kSPRegInternalCode, kWRegSize);
+const Register sp(kSPRegInternalCode, kXRegSize);
+
+
+#define DEFINE_FPREGISTERS(N)  \
+const FPRegister s##N(N, kSRegSize);  \
+const FPRegister d##N(N, kDRegSize);
+REGISTER_CODE_LIST(DEFINE_FPREGISTERS)
+#undef DEFINE_FPREGISTERS
+
+
+// Registers aliases.
+const Register ip0 = x16;
+const Register ip1 = x17;
+const Register lr = x30;
+const Register xzr = x31;
+const Register wzr = w31;
+
+
+// AreAliased returns true if any of the named registers overlap. Arguments
+// set to NoReg are ignored. The system stack pointer may be specified.
+bool AreAliased(const CPURegister& reg1,
+                const CPURegister& reg2,
+                const CPURegister& reg3 = NoReg,
+                const CPURegister& reg4 = NoReg,
+                const CPURegister& reg5 = NoReg,
+                const CPURegister& reg6 = NoReg,
+                const CPURegister& reg7 = NoReg,
+                const CPURegister& reg8 = NoReg);
+
+
+// AreSameSizeAndType returns true if all of the specified registers have the
+// same size, and are of the same type. The system stack pointer may be
+// specified. Arguments set to NoReg are ignored, as are any subsequent
+// arguments. At least one argument (reg1) must be valid (not NoCPUReg).
+bool AreSameSizeAndType(const CPURegister& reg1,
+                        const CPURegister& reg2,
+                        const CPURegister& reg3 = NoCPUReg,
+                        const CPURegister& reg4 = NoCPUReg,
+                        const CPURegister& reg5 = NoCPUReg,
+                        const CPURegister& reg6 = NoCPUReg,
+                        const CPURegister& reg7 = NoCPUReg,
+                        const CPURegister& reg8 = NoCPUReg);
+
+
+// Lists of registers.
+class CPURegList {
+ public:
+  inline 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()) {
+    ASSERT(AreSameSizeAndType(reg1, reg2, reg3, reg4));
+    ASSERT(IsValid());
+  }
+
+  inline CPURegList(CPURegister::RegisterType type, unsigned size, RegList list)
+      : list_(list), size_(size), type_(type) {
+    ASSERT(IsValid());
+  }
+
+  inline CPURegList(CPURegister::RegisterType type, unsigned size,
+                    unsigned first_reg, unsigned last_reg)
+      : size_(size), type_(type) {
+    ASSERT(((type == CPURegister::kRegister) &&
+            (last_reg < kNumberOfRegisters)) ||
+           ((type == CPURegister::kFPRegister) &&
+            (last_reg < kNumberOfFPRegisters)));
+    ASSERT(last_reg >= first_reg);
+    list_ = (1UL << (last_reg + 1)) - 1;
+    list_ &= ~((1UL << first_reg) - 1);
+    ASSERT(IsValid());
+  }
+
+  inline CPURegister::RegisterType type() const {
+    ASSERT(IsValid());
+    return type_;
+  }
+
+  // Combine another CPURegList into this one. Registers that already exist in
+  // this list are left unchanged. The type and size of the registers in the
+  // 'other' list must match those in this list.
+  void Combine(const CPURegList& other) {
+    ASSERT(IsValid());
+    ASSERT(other.type() == type_);
+    ASSERT(other.RegisterSizeInBits() == size_);
+    list_ |= other.list();
+  }
+
+  // Remove every register in the other CPURegList from this one. Registers that
+  // do not exist in this list are ignored. The type and size of the registers
+  // in the 'other' list must match those in this list.
+  void Remove(const CPURegList& other) {
+    ASSERT(IsValid());
+    ASSERT(other.type() == type_);
+    ASSERT(other.RegisterSizeInBits() == size_);
+    list_ &= ~other.list();
+  }
+
+  // Variants of Combine and Remove which take a single register.
+  inline void Combine(const CPURegister& other) {
+    ASSERT(other.type() == type_);
+    ASSERT(other.size() == size_);
+    Combine(other.code());
+  }
+
+  inline void Remove(const CPURegister& other) {
+    ASSERT(other.type() == type_);
+    ASSERT(other.size() == size_);
+    Remove(other.code());
+  }
+
+  // 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) {
+    ASSERT(IsValid());
+    ASSERT(CPURegister(code, size_, type_).IsValid());
+    list_ |= (1UL << code);
+  }
+
+  inline void Remove(int code) {
+    ASSERT(IsValid());
+    ASSERT(CPURegister(code, size_, type_).IsValid());
+    list_ &= ~(1UL << code);
+  }
+
+  inline RegList list() const {
+    ASSERT(IsValid());
+    return list_;
+  }
+
+  // Remove all callee-saved registers from the list. This can be useful when
+  // preparing registers for an AAPCS64 function call, for example.
+  void RemoveCalleeSaved();
+
+  CPURegister PopLowestIndex();
+  CPURegister PopHighestIndex();
+
+  // AAPCS64 callee-saved registers.
+  static CPURegList GetCalleeSaved(unsigned size = kXRegSize);
+  static CPURegList GetCalleeSavedFP(unsigned size = kDRegSize);
+
+  // AAPCS64 caller-saved registers. Note that this includes lr.
+  static CPURegList GetCallerSaved(unsigned size = kXRegSize);
+  static CPURegList GetCallerSavedFP(unsigned size = kDRegSize);
+
+  inline bool IsEmpty() const {
+    ASSERT(IsValid());
+    return list_ == 0;
+  }
+
+  inline bool IncludesAliasOf(const CPURegister& other) const {
+    ASSERT(IsValid());
+    return (type_ == other.type()) && (other.Bit() & list_);
+  }
+
+  inline int Count() const {
+    ASSERT(IsValid());
+    return CountSetBits(list_, kRegListSizeInBits);
+  }
+
+  inline unsigned RegisterSizeInBits() const {
+    ASSERT(IsValid());
+    return size_;
+  }
+
+  inline unsigned RegisterSizeInBytes() const {
+    int size_in_bits = RegisterSizeInBits();
+    ASSERT((size_in_bits % 8) == 0);
+    return size_in_bits / 8;
+  }
+
+ private:
+  RegList list_;
+  unsigned size_;
+  CPURegister::RegisterType type_;
+
+  bool IsValid() const;
+};
+
+
+// AAPCS64 callee-saved registers.
+extern const CPURegList kCalleeSaved;
+extern const CPURegList kCalleeSavedFP;
+
+
+// AAPCS64 caller-saved registers. Note that this includes lr.
+extern const CPURegList kCallerSaved;
+extern const CPURegList kCallerSavedFP;
+
+
+// Operand.
+class Operand {
+ public:
+  // #<immediate>
+  // where <immediate> is int64_t.
+  // This is allowed to be an implicit constructor because Operand is
+  // a wrapper class that doesn't normally perform any type conversion.
+  Operand(int64_t immediate);           // NOLINT(runtime/explicit)
+
+  // rm, {<shift> #<shift_amount>}
+  // where <shift> is one of {LSL, LSR, ASR, ROR}.
+  //       <shift_amount> is uint6_t.
+  // This is allowed to be an implicit constructor because Operand is
+  // a wrapper class that doesn't normally perform any type conversion.
+  Operand(Register reg,
+          Shift shift = LSL,
+          unsigned shift_amount = 0);   // NOLINT(runtime/explicit)
+
+  // rm, {<extend> {#<shift_amount>}}
+  // where <extend> is one of {UXTB, UXTH, UXTW, UXTX, SXTB, SXTH, SXTW, SXTX}.
+  //       <shift_amount> is uint2_t.
+  explicit Operand(Register reg, Extend extend, unsigned shift_amount = 0);
+
+  bool IsImmediate() const;
+  bool IsShiftedRegister() const;
+  bool IsExtendedRegister() const;
+
+  // This returns an LSL shift (<= 4) operand as an equivalent extend operand,
+  // which helps in the encoding of instructions that use the stack pointer.
+  Operand ToExtendedRegister() const;
+
+  int64_t immediate() const {
+    ASSERT(IsImmediate());
+    return immediate_;
+  }
+
+  Register reg() const {
+    ASSERT(IsShiftedRegister() || IsExtendedRegister());
+    return reg_;
+  }
+
+  Shift shift() const {
+    ASSERT(IsShiftedRegister());
+    return shift_;
+  }
+
+  Extend extend() const {
+    ASSERT(IsExtendedRegister());
+    return extend_;
+  }
+
+  unsigned shift_amount() const {
+    ASSERT(IsShiftedRegister() || IsExtendedRegister());
+    return shift_amount_;
+  }
+
+ private:
+  int64_t immediate_;
+  Register reg_;
+  Shift shift_;
+  Extend extend_;
+  unsigned shift_amount_;
+};
+
+
+// MemOperand represents the addressing mode of a load or store instruction.
+class MemOperand {
+ public:
+  explicit MemOperand(Register base,
+                      ptrdiff_t offset = 0,
+                      AddrMode addrmode = Offset);
+  explicit MemOperand(Register base,
+                      Register regoffset,
+                      Shift shift = LSL,
+                      unsigned shift_amount = 0);
+  explicit MemOperand(Register base,
+                      Register regoffset,
+                      Extend extend,
+                      unsigned shift_amount = 0);
+  explicit MemOperand(Register base,
+                      const Operand& offset,
+                      AddrMode addrmode = Offset);
+
+  const Register& base() const { return base_; }
+  const Register& regoffset() const { return regoffset_; }
+  ptrdiff_t offset() const { return offset_; }
+  AddrMode addrmode() const { return addrmode_; }
+  Shift shift() const { return shift_; }
+  Extend extend() const { return extend_; }
+  unsigned shift_amount() const { return shift_amount_; }
+  bool IsImmediateOffset() const;
+  bool IsRegisterOffset() const;
+  bool IsPreIndex() const;
+  bool IsPostIndex() const;
+
+ private:
+  Register base_;
+  Register regoffset_;
+  ptrdiff_t offset_;
+  AddrMode addrmode_;
+  Shift shift_;
+  Extend extend_;
+  unsigned shift_amount_;
+};
+
+
+class Label {
+ public:
+  Label() : is_bound_(false), link_(NULL), target_(NULL) {}
+  ~Label() {
+    // If the label has been linked to, it needs to be bound to a target.
+    ASSERT(!IsLinked() || IsBound());
+  }
+
+  inline Instruction* link() const { return link_; }
+  inline Instruction* target() const { return target_; }
+
+  inline bool IsBound() const { return is_bound_; }
+  inline bool IsLinked() const { return link_ != NULL; }
+
+  inline void set_link(Instruction* new_link) { link_ = new_link; }
+
+  static const int kEndOfChain = 0;
+
+ private:
+  // Indicates if the label has been bound, ie its location is fixed.
+  bool is_bound_;
+  // Branches instructions branching to this label form a chained list, with
+  // their offset indicating where the next instruction is located.
+  // link_ points to the latest branch instruction generated branching to this
+  // branch.
+  // If link_ is not NULL, the label has been linked to.
+  Instruction* link_;
+  // The label location.
+  Instruction* target_;
+
+  friend class Assembler;
+};
+
+
+// TODO: Obtain better values for these, based on real-world data.
+const int kLiteralPoolCheckInterval = 4 * KBytes;
+const int kRecommendedLiteralPoolRange = 2 * kLiteralPoolCheckInterval;
+
+
+// Control whether a branch over the literal pool should also be emitted. This
+// is needed if the literal pool has to be emitted in the middle of the JITted
+// code.
+enum LiteralPoolEmitOption {
+  JumpRequired,
+  NoJumpRequired
+};
+
+
+// Literal pool entry.
+class Literal {
+ public:
+  Literal(Instruction* pc, uint64_t imm, unsigned size)
+      : pc_(pc), value_(imm), size_(size) {}
+
+ private:
+  Instruction* pc_;
+  int64_t value_;
+  unsigned size_;
+
+  friend class Assembler;
+};
+
+
+// Assembler.
+class Assembler {
+ public:
+  Assembler(byte* buffer, unsigned buffer_size);
+
+  // The destructor asserts that one of the following is true:
+  //  * The Assembler object has not been used.
+  //  * Nothing has been emitted since the last Reset() call.
+  //  * Nothing has been emitted since the last FinalizeCode() call.
+  ~Assembler();
+
+  // System functions.
+
+  // Start generating code from the beginning of the buffer, discarding any code
+  // and data that has already been emitted into the buffer.
+  //
+  // In order to avoid any accidental transfer of state, Reset ASSERTs that the
+  // constant pool is not blocked.
+  void Reset();
+
+  // Finalize a code buffer of generated instructions. This function must be
+  // called before executing or copying code from the buffer.
+  void FinalizeCode();
+
+  // Label.
+  // Bind a label to the current PC.
+  void bind(Label* label);
+  int UpdateAndGetByteOffsetTo(Label* label);
+  inline int UpdateAndGetInstructionOffsetTo(Label* label) {
+    ASSERT(Label::kEndOfChain == 0);
+    return UpdateAndGetByteOffsetTo(label) >> kInstructionSizeLog2;
+  }
+
+
+  // Instruction set functions.
+
+  // Branch / Jump instructions.
+  // Branch to register.
+  void br(const Register& xn);
+
+  // Branch with link to register.
+  void blr(const Register& xn);
+
+  // Branch to register with return hint.
+  void ret(const Register& xn = lr);
+
+  // Unconditional branch to label.
+  void b(Label* label);
+
+  // Conditional branch to label.
+  void b(Label* label, Condition cond);
+
+  // Unconditional branch to PC offset.
+  void b(int imm26);
+
+  // Conditional branch to PC offset.
+  void b(int imm19, Condition cond);
+
+  // Branch with link to label.
+  void bl(Label* label);
+
+  // Branch with link to PC offset.
+  void bl(int imm26);
+
+  // Compare and branch to label if zero.
+  void cbz(const Register& rt, Label* label);
+
+  // Compare and branch to PC offset if zero.
+  void cbz(const Register& rt, int imm19);
+
+  // Compare and branch to label if not zero.
+  void cbnz(const Register& rt, Label* label);
+
+  // Compare and branch to PC offset if not zero.
+  void cbnz(const Register& rt, int imm19);
+
+  // Test bit and branch to label if zero.
+  void tbz(const Register& rt, unsigned bit_pos, Label* label);
+
+  // Test bit and branch to PC offset if zero.
+  void tbz(const Register& rt, unsigned bit_pos, int imm14);
+
+  // Test bit and branch to label if not zero.
+  void tbnz(const Register& rt, unsigned bit_pos, Label* label);
+
+  // Test bit and branch to PC offset if not zero.
+  void tbnz(const Register& rt, unsigned bit_pos, int imm14);
+
+  // Address calculation instructions.
+  // Calculate a PC-relative address. Unlike for branches the offset in adr is
+  // unscaled (i.e. the result can be unaligned).
+
+  // Calculate the address of a label.
+  void adr(const Register& rd, Label* label);
+
+  // Calculate the address of a PC offset.
+  void adr(const Register& rd, int imm21);
+
+  // Data Processing instructions.
+  // Add.
+  void add(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Compare negative.
+  void cmn(const Register& rn, const Operand& operand);
+
+  // Subtract.
+  void sub(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Compare.
+  void cmp(const Register& rn, const Operand& operand);
+
+  // Negate.
+  void neg(const Register& rd,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Add with carry bit.
+  void adc(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Subtract with carry bit.
+  void sbc(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Negate with carry bit.
+  void ngc(const Register& rd,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Logical instructions.
+  // Bitwise and (A & B).
+  void and_(const Register& rd,
+            const Register& rn,
+            const Operand& operand,
+            FlagsUpdate S = LeaveFlags);
+
+  // Bit test and set flags.
+  void tst(const Register& rn, const Operand& operand);
+
+  // Bit clear (A & ~B).
+  void bic(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+
+  // Bitwise or (A | B).
+  void orr(const Register& rd, const Register& rn, const Operand& operand);
+
+  // Bitwise nor (A | ~B).
+  void orn(const Register& rd, const Register& rn, const Operand& operand);
+
+  // Bitwise eor/xor (A ^ B).
+  void eor(const Register& rd, const Register& rn, const Operand& operand);
+
+  // Bitwise enor/xnor (A ^ ~B).
+  void eon(const Register& rd, const Register& rn, const Operand& operand);
+
+  // Logical shift left by variable.
+  void lslv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Logical shift right by variable.
+  void lsrv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Arithmetic shift right by variable.
+  void asrv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Rotate right by variable.
+  void rorv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Bitfield instructions.
+  // Bitfield move.
+  void bfm(const Register& rd,
+           const Register& rn,
+           unsigned immr,
+           unsigned imms);
+
+  // Signed bitfield move.
+  void sbfm(const Register& rd,
+            const Register& rn,
+            unsigned immr,
+            unsigned imms);
+
+  // Unsigned bitfield move.
+  void ubfm(const Register& rd,
+            const Register& rn,
+            unsigned immr,
+            unsigned imms);
+
+  // Bfm aliases.
+  // Bitfield insert.
+  inline void bfi(const Register& rd,
+                  const Register& rn,
+                  unsigned lsb,
+                  unsigned width) {
+    ASSERT(width >= 1);
+    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) {
+    ASSERT(width >= 1);
+    ASSERT(lsb + width <= rn.size());
+    bfm(rd, rn, lsb, lsb + width - 1);
+  }
+
+  // Sbfm aliases.
+  // Arithmetic shift right.
+  inline void asr(const Register& rd, const Register& rn, unsigned shift) {
+    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) {
+    ASSERT(width >= 1);
+    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) {
+    ASSERT(width >= 1);
+    ASSERT(lsb + width <= rn.size());
+    sbfm(rd, rn, lsb, lsb + width - 1);
+  }
+
+  // Signed extend byte.
+  inline 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) {
+    sbfm(rd, rn, 0, 15);
+  }
+
+  // Signed extend word.
+  inline 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) {
+    unsigned reg_size = rd.size();
+    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) {
+    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) {
+    ASSERT(width >= 1);
+    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) {
+    ASSERT(width >= 1);
+    ASSERT(lsb + width <= rn.size());
+    ubfm(rd, rn, lsb, lsb + width - 1);
+  }
+
+  // Unsigned extend byte.
+  inline 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) {
+    ubfm(rd, rn, 0, 15);
+  }
+
+  // Unsigned extend word.
+  inline void uxtw(const Register& rd, const Register& rn) {
+    ubfm(rd, rn, 0, 31);
+  }
+
+  // Extract.
+  void extr(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            unsigned lsb);
+
+  // Conditional select: rd = cond ? rn : rm.
+  void csel(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            Condition cond);
+
+  // Conditional select increment: rd = cond ? rn : rm + 1.
+  void csinc(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond);
+
+  // Conditional select inversion: rd = cond ? rn : ~rm.
+  void csinv(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond);
+
+  // Conditional select negation: rd = cond ? rn : -rm.
+  void csneg(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond);
+
+  // Conditional set: rd = cond ? 1 : 0.
+  void cset(const Register& rd, Condition cond);
+
+  // Conditional set mask: rd = cond ? -1 : 0.
+  void csetm(const Register& rd, Condition cond);
+
+  // Conditional increment: rd = cond ? rn + 1 : rn.
+  void cinc(const Register& rd, const Register& rn, Condition cond);
+
+  // Conditional invert: rd = cond ? ~rn : rn.
+  void cinv(const Register& rd, const Register& rn, Condition cond);
+
+  // Conditional negate: rd = cond ? -rn : rn.
+  void cneg(const Register& rd, const Register& rn, Condition cond);
+
+  // Rotate right.
+  inline void ror(const Register& rd, const Register& rs, unsigned shift) {
+    extr(rd, rs, rs, shift);
+  }
+
+  // Conditional comparison.
+  // Conditional compare negative.
+  void ccmn(const Register& rn,
+            const Operand& operand,
+            StatusFlags nzcv,
+            Condition cond);
+
+  // Conditional compare.
+  void ccmp(const Register& rn,
+            const Operand& operand,
+            StatusFlags nzcv,
+            Condition cond);
+
+  // Multiply.
+  void mul(const Register& rd, const Register& rn, const Register& rm);
+
+  // Negated multiply.
+  void mneg(const Register& rd, const Register& rn, const Register& rm);
+
+  // Signed long multiply: 32 x 32 -> 64-bit.
+  void smull(const Register& rd, const Register& rn, const Register& rm);
+
+  // Signed multiply high: 64 x 64 -> 64-bit <127:64>.
+  void smulh(const Register& xd, const Register& xn, const Register& xm);
+
+  // Multiply and accumulate.
+  void madd(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            const Register& ra);
+
+  // Multiply and subtract.
+  void msub(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            const Register& ra);
+
+  // Signed long multiply and accumulate: 32 x 32 + 64 -> 64-bit.
+  void smaddl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra);
+
+  // Unsigned long multiply and accumulate: 32 x 32 + 64 -> 64-bit.
+  void umaddl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra);
+
+  // Signed long multiply and subtract: 64 - (32 x 32) -> 64-bit.
+  void smsubl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra);
+
+  // Unsigned long multiply and subtract: 64 - (32 x 32) -> 64-bit.
+  void umsubl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra);
+
+  // Signed integer divide.
+  void sdiv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Unsigned integer divide.
+  void udiv(const Register& rd, const Register& rn, const Register& rm);
+
+  // Bit reverse.
+  void rbit(const Register& rd, const Register& rn);
+
+  // Reverse bytes in 16-bit half words.
+  void rev16(const Register& rd, const Register& rn);
+
+  // Reverse bytes in 32-bit words.
+  void rev32(const Register& rd, const Register& rn);
+
+  // Reverse bytes.
+  void rev(const Register& rd, const Register& rn);
+
+  // Count leading zeroes.
+  void clz(const Register& rd, const Register& rn);
+
+  // Count leading sign bits.
+  void cls(const Register& rd, const Register& rn);
+
+  // Memory instructions.
+  // Load integer or FP register.
+  void ldr(const CPURegister& rt, const MemOperand& src);
+
+  // Store integer or FP register.
+  void str(const CPURegister& rt, const MemOperand& dst);
+
+  // Load word with sign extension.
+  void ldrsw(const Register& rt, const MemOperand& src);
+
+  // Load byte.
+  void ldrb(const Register& rt, const MemOperand& src);
+
+  // Store byte.
+  void strb(const Register& rt, const MemOperand& dst);
+
+  // Load byte with sign extension.
+  void ldrsb(const Register& rt, const MemOperand& src);
+
+  // Load half-word.
+  void ldrh(const Register& rt, const MemOperand& src);
+
+  // Store half-word.
+  void strh(const Register& rt, const MemOperand& dst);
+
+  // Load half-word with sign extension.
+  void ldrsh(const Register& rt, const MemOperand& src);
+
+  // Load integer or FP register pair.
+  void ldp(const CPURegister& rt, const CPURegister& rt2,
+           const MemOperand& src);
+
+  // Store integer or FP register pair.
+  void stp(const CPURegister& rt, const CPURegister& rt2,
+           const MemOperand& dst);
+
+  // Load word pair with sign extension.
+  void ldpsw(const Register& rt, const Register& rt2, const MemOperand& src);
+
+  // Load integer or FP register pair, non-temporal.
+  void ldnp(const CPURegister& rt, const CPURegister& rt2,
+            const MemOperand& src);
+
+  // Store integer or FP register pair, non-temporal.
+  void stnp(const CPURegister& rt, const CPURegister& rt2,
+            const MemOperand& dst);
+
+  // Load literal to register.
+  void ldr(const Register& rt, uint64_t imm);
+
+  // Load literal to FP register.
+  void ldr(const FPRegister& ft, double imm);
+
+  // Move instructions. The default shift of -1 indicates that the move
+  // instruction will calculate an appropriate 16-bit immediate and left shift
+  // that is equal to the 64-bit immediate argument. If an explicit left shift
+  // is specified (0, 16, 32 or 48), the immediate must be a 16-bit value.
+  //
+  // For movk, an explicit shift can be used to indicate which half word should
+  // be overwritten, eg. movk(x0, 0, 0) will overwrite the least-significant
+  // half word with zero, whereas movk(x0, 0, 48) will overwrite the
+  // most-significant.
+
+  // Move immediate and keep.
+  void movk(const Register& rd, uint64_t imm, int shift = -1) {
+    MoveWide(rd, imm, shift, MOVK);
+  }
+
+  // Move inverted immediate.
+  void movn(const Register& rd, uint64_t imm, int shift = -1) {
+    MoveWide(rd, imm, shift, MOVN);
+  }
+
+  // Move immediate.
+  void movz(const Register& rd, uint64_t imm, int shift = -1) {
+    MoveWide(rd, imm, shift, MOVZ);
+  }
+
+  // Misc instructions.
+  // Monitor debug-mode breakpoint.
+  void brk(int code);
+
+  // Halting debug-mode breakpoint.
+  void hlt(int code);
+
+  // Move register to register.
+  void mov(const Register& rd, const Register& rn);
+
+  // Move inverted operand to register.
+  void mvn(const Register& rd, const Operand& operand);
+
+  // System instructions.
+  // Move to register from system register.
+  void mrs(const Register& rt, SystemRegister sysreg);
+
+  // Move from register to system register.
+  void msr(SystemRegister sysreg, const Register& rt);
+
+  // System hint.
+  void hint(SystemHint code);
+
+  // Alias for system instructions.
+  // No-op.
+  void nop() {
+    hint(NOP);
+  }
+
+  // FP instructions.
+  // Move immediate to FP register.
+  void fmov(FPRegister fd, double imm);
+
+  // Move FP register to register.
+  void fmov(Register rd, FPRegister fn);
+
+  // Move register to FP register.
+  void fmov(FPRegister fd, Register rn);
+
+  // Move FP register to FP register.
+  void fmov(FPRegister fd, FPRegister fn);
+
+  // FP add.
+  void fadd(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP subtract.
+  void fsub(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP multiply.
+  void fmul(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP multiply and subtract.
+  void fmsub(const FPRegister& fd,
+             const FPRegister& fn,
+             const FPRegister& fm,
+             const FPRegister& fa);
+
+  // FP divide.
+  void fdiv(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP maximum.
+  void fmax(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP minimum.
+  void fmin(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm);
+
+  // FP absolute.
+  void fabs(const FPRegister& fd, const FPRegister& fn);
+
+  // FP negate.
+  void fneg(const FPRegister& fd, const FPRegister& fn);
+
+  // FP square root.
+  void fsqrt(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 (towards zero).
+  void frintz(const FPRegister& fd, const FPRegister& fn);
+
+  // FP compare registers.
+  void fcmp(const FPRegister& fn, const FPRegister& fm);
+
+  // FP compare immediate.
+  void fcmp(const FPRegister& fn, double value);
+
+  // FP conditional compare.
+  void fccmp(const FPRegister& fn,
+             const FPRegister& fm,
+             StatusFlags nzcv,
+             Condition cond);
+
+  // FP conditional select.
+  void fcsel(const FPRegister& fd,
+             const FPRegister& fn,
+             const FPRegister& fm,
+             Condition cond);
+
+  // Common FP Convert function.
+  void FPConvertToInt(const Register& rd,
+                      const FPRegister& fn,
+                      FPIntegerConvertOp op);
+
+  // FP convert between single and double precision.
+  void fcvt(const FPRegister& fd, const FPRegister& fn);
+
+  // Convert FP to unsigned integer (round towards -infinity).
+  void fcvtmu(const Register& rd, const FPRegister& fn);
+
+  // Convert FP to signed integer (round towards -infinity).
+  void fcvtms(const Register& rd, const FPRegister& fn);
+
+  // Convert FP to unsigned integer (nearest with ties to even).
+  void fcvtnu(const Register& rd, const FPRegister& fn);
+
+  // Convert FP to signed integer (nearest with ties to even).
+  void fcvtns(const Register& rd, const FPRegister& fn);
+
+  // Convert FP to unsigned integer (round towards zero).
+  void fcvtzu(const Register& rd, const FPRegister& fn);
+
+  // Convert FP to signed integer (round towards zero).
+  void fcvtzs(const Register& rd, const FPRegister& fn);
+
+  // Convert signed integer or fixed point to FP.
+  void scvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0);
+
+  // Convert unsigned integer or fixed point to FP.
+  void ucvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0);
+
+  // Emit generic instructions.
+  // Emit raw instructions into the instruction stream.
+  inline void dci(Instr raw_inst) { Emit(raw_inst); }
+
+  // Emit 32 bits of data into the instruction stream.
+  inline void dc32(uint32_t data) { EmitData(&data, sizeof(data)); }
+
+  // Emit 64 bits of data into the instruction stream.
+  inline void dc64(uint64_t data) { EmitData(&data, sizeof(data)); }
+
+  // Copy a string into the instruction stream, including the terminating NULL
+  // character. The instruction pointer (pc_) is then aligned correctly for
+  // subsequent instructions.
+  void EmitStringData(const char * string) {
+    ASSERT(string != NULL);
+
+    size_t len = strlen(string) + 1;
+    EmitData(string, len);
+
+    // Pad with NULL characters until pc_ is aligned.
+    const char pad[] = {'\0', '\0', '\0', '\0'};
+    ASSERT(sizeof(pad) == kInstructionSize);
+    Instruction* next_pc = AlignUp(pc_, kInstructionSize);
+    EmitData(&pad, next_pc - pc_);
+  }
+
+  // Code generation helpers.
+
+  // Register encoding.
+  static Instr Rd(CPURegister rd) {
+    ASSERT(rd.code() != kSPRegInternalCode);
+    return rd.code() << Rd_offset;
+  }
+
+  static Instr Rn(CPURegister rn) {
+    ASSERT(rn.code() != kSPRegInternalCode);
+    return rn.code() << Rn_offset;
+  }
+
+  static Instr Rm(CPURegister rm) {
+    ASSERT(rm.code() != kSPRegInternalCode);
+    return rm.code() << Rm_offset;
+  }
+
+  static Instr Ra(CPURegister ra) {
+    ASSERT(ra.code() != kSPRegInternalCode);
+    return ra.code() << Ra_offset;
+  }
+
+  static Instr Rt(CPURegister rt) {
+    ASSERT(rt.code() != kSPRegInternalCode);
+    return rt.code() << Rt_offset;
+  }
+
+  static Instr Rt2(CPURegister rt2) {
+    ASSERT(rt2.code() != kSPRegInternalCode);
+    return rt2.code() << Rt2_offset;
+  }
+
+  // These encoding functions allow the stack pointer to be encoded, and
+  // disallow the zero register.
+  static Instr RdSP(Register rd) {
+    ASSERT(!rd.IsZero());
+    return (rd.code() & kRegCodeMask) << Rd_offset;
+  }
+
+  static Instr RnSP(Register rn) {
+    ASSERT(!rn.IsZero());
+    return (rn.code() & kRegCodeMask) << Rn_offset;
+  }
+
+  // Flags encoding.
+  static Instr Flags(FlagsUpdate S) {
+    if (S == SetFlags) {
+      return 1 << FlagsUpdate_offset;
+    } else if (S == LeaveFlags) {
+      return 0 << FlagsUpdate_offset;
+    }
+    UNREACHABLE();
+    return 0;
+  }
+
+  static Instr Cond(Condition cond) {
+    return cond << Condition_offset;
+  }
+
+  // PC-relative address encoding.
+  static Instr ImmPCRelAddress(int imm21) {
+    ASSERT(is_int21(imm21));
+    Instr imm = static_cast<Instr>(truncate_to_int21(imm21));
+    Instr immhi = (imm >> ImmPCRelLo_width) << ImmPCRelHi_offset;
+    Instr immlo = imm << ImmPCRelLo_offset;
+    return (immhi & ImmPCRelHi_mask) | (immlo & ImmPCRelLo_mask);
+  }
+
+  // Branch encoding.
+  static Instr ImmUncondBranch(int imm26) {
+    ASSERT(is_int26(imm26));
+    return truncate_to_int26(imm26) << ImmUncondBranch_offset;
+  }
+
+  static Instr ImmCondBranch(int imm19) {
+    ASSERT(is_int19(imm19));
+    return truncate_to_int19(imm19) << ImmCondBranch_offset;
+  }
+
+  static Instr ImmCmpBranch(int imm19) {
+    ASSERT(is_int19(imm19));
+    return truncate_to_int19(imm19) << ImmCmpBranch_offset;
+  }
+
+  static Instr ImmTestBranch(int imm14) {
+    ASSERT(is_int14(imm14));
+    return truncate_to_int14(imm14) << ImmTestBranch_offset;
+  }
+
+  static Instr ImmTestBranchBit(unsigned bit_pos) {
+    ASSERT(is_uint6(bit_pos));
+    // Subtract five from the shift offset, as we need bit 5 from bit_pos.
+    unsigned b5 = bit_pos << (ImmTestBranchBit5_offset - 5);
+    unsigned b40 = bit_pos << ImmTestBranchBit40_offset;
+    b5 &= ImmTestBranchBit5_mask;
+    b40 &= ImmTestBranchBit40_mask;
+    return b5 | b40;
+  }
+
+  // Data Processing encoding.
+  static Instr SF(Register rd) {
+      return rd.Is64Bits() ? SixtyFourBits : ThirtyTwoBits;
+  }
+
+  static Instr ImmAddSub(int64_t imm) {
+    ASSERT(IsImmAddSub(imm));
+    if (is_uint12(imm)) {  // No shift required.
+      return imm << ImmAddSub_offset;
+    } else {
+      return ((imm >> 12) << ImmAddSub_offset) | (1 << ShiftAddSub_offset);
+    }
+  }
+
+  static inline Instr ImmS(unsigned imms, unsigned reg_size) {
+    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) {
+    ASSERT(((reg_size == kXRegSize) && is_uint6(immr)) ||
+           ((reg_size == kWRegSize) && is_uint5(immr)));
+    USE(reg_size);
+    ASSERT(is_uint6(immr));
+    return immr << ImmR_offset;
+  }
+
+  static inline Instr ImmSetBits(unsigned imms, unsigned reg_size) {
+    ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
+    ASSERT(is_uint6(imms));
+    ASSERT((reg_size == kXRegSize) || is_uint6(imms + 3));
+    USE(reg_size);
+    return imms << ImmSetBits_offset;
+  }
+
+  static inline Instr ImmRotate(unsigned immr, unsigned reg_size) {
+    ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
+    ASSERT(((reg_size == kXRegSize) && is_uint6(immr)) ||
+           ((reg_size == kWRegSize) && is_uint5(immr)));
+    USE(reg_size);
+    return immr << ImmRotate_offset;
+  }
+
+  static inline Instr ImmLLiteral(int imm19) {
+    ASSERT(is_int19(imm19));
+    return truncate_to_int19(imm19) << ImmLLiteral_offset;
+  }
+
+  static inline Instr BitN(unsigned bitn, unsigned reg_size) {
+    ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
+    ASSERT((reg_size == kXRegSize) || (bitn == 0));
+    USE(reg_size);
+    return bitn << BitN_offset;
+  }
+
+  static Instr ShiftDP(Shift shift) {
+    ASSERT(shift == LSL || shift == LSR || shift == ASR || shift == ROR);
+    return shift << ShiftDP_offset;
+  }
+
+  static Instr ImmDPShift(unsigned amount) {
+    ASSERT(is_uint6(amount));
+    return amount << ImmDPShift_offset;
+  }
+
+  static Instr ExtendMode(Extend extend) {
+    return extend << ExtendMode_offset;
+  }
+
+  static Instr ImmExtendShift(unsigned left_shift) {
+    ASSERT(left_shift <= 4);
+    return left_shift << ImmExtendShift_offset;
+  }
+
+  static Instr ImmCondCmp(unsigned imm) {
+    ASSERT(is_uint5(imm));
+    return imm << ImmCondCmp_offset;
+  }
+
+  static Instr Nzcv(StatusFlags nzcv) {
+    return ((nzcv >> Flags_offset) & 0xf) << Nzcv_offset;
+  }
+
+  // MemOperand offset encoding.
+  static Instr ImmLSUnsigned(int imm12) {
+    ASSERT(is_uint12(imm12));
+    return imm12 << ImmLSUnsigned_offset;
+  }
+
+  static Instr ImmLS(int imm9) {
+    ASSERT(is_int9(imm9));
+    return truncate_to_int9(imm9) << ImmLS_offset;
+  }
+
+  static Instr ImmLSPair(int imm7, LSDataSize size) {
+    ASSERT(((imm7 >> size) << size) == imm7);
+    int scaled_imm7 = imm7 >> size;
+    ASSERT(is_int7(scaled_imm7));
+    return truncate_to_int7(scaled_imm7) << ImmLSPair_offset;
+  }
+
+  static Instr ImmShiftLS(unsigned shift_amount) {
+    ASSERT(is_uint1(shift_amount));
+    return shift_amount << ImmShiftLS_offset;
+  }
+
+  static Instr ImmException(int imm16) {
+    ASSERT(is_uint16(imm16));
+    return imm16 << ImmException_offset;
+  }
+
+  static Instr ImmSystemRegister(int imm15) {
+    ASSERT(is_uint15(imm15));
+    return imm15 << ImmSystemRegister_offset;
+  }
+
+  static Instr ImmHint(int imm7) {
+    ASSERT(is_uint7(imm7));
+    return imm7 << ImmHint_offset;
+  }
+
+  static LSDataSize CalcLSDataSize(LoadStoreOp op) {
+    ASSERT((SizeLS_offset + SizeLS_width) == (kInstructionSize * 8));
+    return static_cast<LSDataSize>(op >> SizeLS_offset);
+  }
+
+  // Move immediates encoding.
+  static Instr ImmMoveWide(uint64_t imm) {
+    ASSERT(is_uint16(imm));
+    return imm << ImmMoveWide_offset;
+  }
+
+  static Instr ShiftMoveWide(int64_t shift) {
+    ASSERT(is_uint2(shift));
+    return shift << ShiftMoveWide_offset;
+  }
+
+  // FP Immediates.
+  static Instr ImmFP32(float imm);
+  static Instr ImmFP64(double imm);
+
+  // FP register type.
+  static Instr FPType(FPRegister fd) {
+    return fd.Is64Bits() ? FP64 : FP32;
+  }
+
+  static Instr FPScale(unsigned scale) {
+    ASSERT(is_uint6(scale));
+    return scale << FPScale_offset;
+  }
+
+  // Size of the code generated in bytes
+  uint64_t SizeOfCodeGenerated() const {
+    ASSERT((pc_ >= buffer_) && (pc_ < (buffer_ + buffer_size_)));
+    return pc_ - buffer_;
+  }
+
+  // Size of the code generated since label to the current position.
+  uint64_t SizeOfCodeGeneratedSince(Label* label) const {
+    ASSERT(label->IsBound());
+    ASSERT((pc_ >= label->target()) && (pc_ < (buffer_ + buffer_size_)));
+    return pc_ - label->target();
+  }
+
+
+  inline void BlockLiteralPool() {
+    literal_pool_monitor_++;
+  }
+
+  inline void ReleaseLiteralPool() {
+    if (--literal_pool_monitor_ == 0) {
+      // Has the literal pool been blocked for too long?
+      ASSERT(literals_.empty() ||
+             (pc_ < (literals_.back()->pc_ + kMaxLoadLiteralRange)));
+    }
+  }
+
+  inline bool IsLiteralPoolBlocked() {
+    return literal_pool_monitor_ != 0;
+  }
+
+  void CheckLiteralPool(LiteralPoolEmitOption option = JumpRequired);
+  void EmitLiteralPool(LiteralPoolEmitOption option = NoJumpRequired);
+  size_t LiteralPoolSize();
+
+ protected:
+  inline const Register& AppropriateZeroRegFor(const CPURegister& reg) const {
+    return reg.Is64Bits() ? xzr : wzr;
+  }
+
+
+  void LoadStore(const CPURegister& rt,
+                 const MemOperand& addr,
+                 LoadStoreOp op);
+  static bool IsImmLSUnscaled(ptrdiff_t offset);
+  static bool IsImmLSScaled(ptrdiff_t offset, LSDataSize size);
+
+  void Logical(const Register& rd,
+               const Register& rn,
+               const Operand& operand,
+               LogicalOp op);
+  void LogicalImmediate(const Register& rd,
+                        const Register& rn,
+                        unsigned n,
+                        unsigned imm_s,
+                        unsigned imm_r,
+                        LogicalOp op);
+  static bool IsImmLogical(uint64_t value,
+                           unsigned width,
+                           unsigned* n,
+                           unsigned* imm_s,
+                           unsigned* imm_r);
+
+  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,
+                       const Operand& operand,
+                       FlagsUpdate S,
+                       AddSubWithCarryOp op);
+
+  // Functions for emulating operands not directly supported by the instruction
+  // set.
+  void EmitShift(const Register& rd,
+                 const Register& rn,
+                 Shift shift,
+                 unsigned amount);
+  void EmitExtendShift(const Register& rd,
+                       const Register& rn,
+                       Extend extend,
+                       unsigned left_shift);
+
+  void AddSub(const Register& rd,
+              const Register& rn,
+              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
+  // as in LDPSW_x or LDRB_w) are not supported.
+  static LoadStoreOp LoadOpFor(const CPURegister& rt);
+  static LoadStorePairOp LoadPairOpFor(const CPURegister& rt,
+                                       const CPURegister& rt2);
+  static LoadStoreOp StoreOpFor(const CPURegister& rt);
+  static LoadStorePairOp StorePairOpFor(const CPURegister& rt,
+                                        const CPURegister& rt2);
+  static LoadStorePairNonTemporalOp LoadPairNonTemporalOpFor(
+    const CPURegister& rt, const CPURegister& rt2);
+  static LoadStorePairNonTemporalOp StorePairNonTemporalOpFor(
+    const CPURegister& rt, const CPURegister& rt2);
+
+
+ private:
+  // Instruction helpers.
+  void MoveWide(const Register& rd,
+                uint64_t imm,
+                int shift,
+                MoveWideImmediateOp mov_op);
+  void DataProcShiftedRegister(const Register& rd,
+                               const Register& rn,
+                               const Operand& operand,
+                               FlagsUpdate S,
+                               Instr op);
+  void DataProcExtendedRegister(const Register& rd,
+                                const Register& rn,
+                                const Operand& operand,
+                                FlagsUpdate S,
+                                Instr op);
+  void LoadStorePair(const CPURegister& rt,
+                     const CPURegister& rt2,
+                     const MemOperand& addr,
+                     LoadStorePairOp op);
+  void LoadStorePairNonTemporal(const CPURegister& rt,
+                                const CPURegister& rt2,
+                                const MemOperand& addr,
+                                LoadStorePairNonTemporalOp op);
+  void LoadLiteral(const CPURegister& rt, uint64_t imm, LoadLiteralOp op);
+  void ConditionalSelect(const Register& rd,
+                         const Register& rn,
+                         const Register& rm,
+                         Condition cond,
+                         ConditionalSelectOp op);
+  void DataProcessing1Source(const Register& rd,
+                             const Register& rn,
+                             DataProcessing1SourceOp op);
+  void DataProcessing3Source(const Register& rd,
+                             const Register& rn,
+                             const Register& rm,
+                             const Register& ra,
+                             DataProcessing3SourceOp op);
+  void FPDataProcessing1Source(const FPRegister& fd,
+                               const FPRegister& fn,
+                               FPDataProcessing1SourceOp op);
+  void FPDataProcessing2Source(const FPRegister& fd,
+                               const FPRegister& fn,
+                               const FPRegister& fm,
+                               FPDataProcessing2SourceOp op);
+  void FPDataProcessing3Source(const FPRegister& fd,
+                               const FPRegister& fn,
+                               const FPRegister& fm,
+                               const FPRegister& fa,
+                               FPDataProcessing3SourceOp op);
+
+  // Encoding helpers.
+  static bool IsImmFP32(float imm);
+  static bool IsImmFP64(double imm);
+
+  void RecordLiteral(int64_t imm, unsigned size);
+
+  // Emit the instruction at pc_.
+  void Emit(Instr instruction) {
+    ASSERT(sizeof(*pc_) == 1);
+    ASSERT(sizeof(instruction) == kInstructionSize);
+    ASSERT((pc_ + sizeof(instruction)) <= (buffer_ + buffer_size_));
+
+#ifdef DEBUG
+    finalized_ = false;
+#endif
+
+    memcpy(pc_, &instruction, sizeof(instruction));
+    pc_ += sizeof(instruction);
+    CheckBufferSpace();
+  }
+
+  // Emit data inline in the instruction stream.
+  void EmitData(void const * data, unsigned size) {
+    ASSERT(sizeof(*pc_) == 1);
+    ASSERT((pc_ + size) <= (buffer_ + buffer_size_));
+
+#ifdef DEBUG
+    finalized_ = false;
+#endif
+
+    // TODO: Record this 'instruction' as data, so that it can be disassembled
+    // correctly.
+    memcpy(pc_, data, size);
+    pc_ += size;
+    CheckBufferSpace();
+  }
+
+  inline void CheckBufferSpace() {
+    ASSERT(pc_ < (buffer_ + buffer_size_));
+    if (pc_ > next_literal_pool_check_) {
+      CheckLiteralPool();
+    }
+  }
+
+  // The buffer into which code and relocation info are generated.
+  Instruction* buffer_;
+  // Buffer size, in bytes.
+  unsigned buffer_size_;
+  Instruction* pc_;
+  std::list<Literal*> literals_;
+  Instruction* next_literal_pool_check_;
+  unsigned literal_pool_monitor_;
+
+  friend class BlockLiteralPoolScope;
+
+#ifdef DEBUG
+  bool finalized_;
+#endif
+};
+
+class BlockLiteralPoolScope {
+ public:
+  explicit BlockLiteralPoolScope(Assembler* assm) : assm_(assm) {
+    assm_->BlockLiteralPool();
+  }
+
+  ~BlockLiteralPoolScope() {
+    assm_->ReleaseLiteralPool();
+  }
+
+ private:
+  Assembler* assm_;
+};
+}  // namespace vixl
+
+#endif  // VIXL_A64_ASSEMBLER_A64_H_
diff --git a/disas/libvixl/src/a64/constants-a64.h b/disas/libvixl/src/a64/constants-a64.h
new file mode 100644
index 0000000..2e0336d
--- /dev/null
+++ b/disas/libvixl/src/a64/constants-a64.h
@@ -0,0 +1,1104 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_CONSTANTS_A64_H_
+#define VIXL_A64_CONSTANTS_A64_H_
+
+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)                                 \
+R(8)  R(9)  R(10) R(11) R(12) R(13) R(14) R(15)                                \
+R(16) R(17) R(18) R(19) R(20) R(21) R(22) R(23)                                \
+R(24) R(25) R(26) R(27) R(28) R(29) R(30) R(31)
+
+#define INSTRUCTION_FIELDS_LIST(V_)                                            \
+/* Register fields */                                                          \
+V_(Rd, 4, 0, Bits)                        /* Destination register.     */      \
+V_(Rn, 9, 5, Bits)                        /* First source register.    */      \
+V_(Rm, 20, 16, Bits)                      /* Second source register.   */      \
+V_(Ra, 14, 10, Bits)                      /* Third source register.    */      \
+V_(Rt, 4, 0, Bits)                        /* Load dest / store source. */      \
+V_(Rt2, 14, 10, Bits)                     /* Load second dest /        */      \
+                                         /* store second source.      */       \
+V_(PrefetchMode, 4, 0, Bits)                                                   \
+                                                                               \
+/* Common bits */                                                              \
+V_(SixtyFourBits, 31, 31, Bits)                                                \
+V_(FlagsUpdate, 29, 29, Bits)                                                  \
+                                                                               \
+/* PC relative addressing */                                                   \
+V_(ImmPCRelHi, 23, 5, SignedBits)                                              \
+V_(ImmPCRelLo, 30, 29, Bits)                                                   \
+                                                                               \
+/* Add/subtract/logical shift register */                                      \
+V_(ShiftDP, 23, 22, Bits)                                                      \
+V_(ImmDPShift, 15, 10, Bits)                                                   \
+                                                                               \
+/* Add/subtract immediate */                                                   \
+V_(ImmAddSub, 21, 10, Bits)                                                    \
+V_(ShiftAddSub, 23, 22, Bits)                                                  \
+                                                                               \
+/* Add/substract extend */                                                     \
+V_(ImmExtendShift, 12, 10, Bits)                                               \
+V_(ExtendMode, 15, 13, Bits)                                                   \
+                                                                               \
+/* Move wide */                                                                \
+V_(ImmMoveWide, 20, 5, Bits)                                                   \
+V_(ShiftMoveWide, 22, 21, Bits)                                                \
+                                                                               \
+/* Logical immediate, bitfield and extract */                                  \
+V_(BitN, 22, 22, Bits)                                                         \
+V_(ImmRotate, 21, 16, Bits)                                                    \
+V_(ImmSetBits, 15, 10, Bits)                                                   \
+V_(ImmR, 21, 16, Bits)                                                         \
+V_(ImmS, 15, 10, Bits)                                                         \
+                                                                               \
+/* Test and branch immediate */                                                \
+V_(ImmTestBranch, 18, 5, SignedBits)                                           \
+V_(ImmTestBranchBit40, 23, 19, Bits)                                           \
+V_(ImmTestBranchBit5, 31, 31, Bits)                                            \
+                                                                               \
+/* Conditionals */                                                             \
+V_(Condition, 15, 12, Bits)                                                    \
+V_(ConditionBranch, 3, 0, Bits)                                                \
+V_(Nzcv, 3, 0, Bits)                                                           \
+V_(ImmCondCmp, 20, 16, Bits)                                                   \
+V_(ImmCondBranch, 23, 5, SignedBits)                                           \
+                                                                               \
+/* Floating point */                                                           \
+V_(FPType, 23, 22, Bits)                                                       \
+V_(ImmFP, 20, 13, Bits)                                                        \
+V_(FPScale, 15, 10, Bits)                                                      \
+                                                                               \
+/* Load Store */                                                               \
+V_(ImmLS, 20, 12, SignedBits)                                                  \
+V_(ImmLSUnsigned, 21, 10, Bits)                                                \
+V_(ImmLSPair, 21, 15, SignedBits)                                              \
+V_(SizeLS, 31, 30, Bits)                                                       \
+V_(ImmShiftLS, 12, 12, Bits)                                                   \
+                                                                               \
+/* Other immediates */                                                         \
+V_(ImmUncondBranch, 25, 0, SignedBits)                                         \
+V_(ImmCmpBranch, 23, 5, SignedBits)                                            \
+V_(ImmLLiteral, 23, 5, SignedBits)                                             \
+V_(ImmException, 20, 5, Bits)                                                  \
+V_(ImmHint, 11, 5, Bits)                                                       \
+                                                                               \
+/* System (MRS, MSR) */                                                        \
+V_(ImmSystemRegister, 19, 5, Bits)                                             \
+V_(SysO0, 19, 19, Bits)                                                        \
+V_(SysOp1, 18, 16, Bits)                                                       \
+V_(SysOp2, 7, 5, Bits)                                                         \
+V_(CRn, 15, 12, Bits)                                                          \
+V_(CRm, 11, 8, Bits)                                                           \
+
+
+#define SYSTEM_REGISTER_FIELDS_LIST(V_, M_)                                    \
+/* NZCV */                                                                     \
+V_(Flags, 31, 28, Bits)                                                        \
+V_(N, 31, 31, Bits)                                                            \
+V_(Z, 30, 30, Bits)                                                            \
+V_(C, 29, 29, Bits)                                                            \
+V_(V, 28, 28, Bits)                                                            \
+M_(NZCV, Flags_mask)                                                           \
+                                                                               \
+/* FPCR */                                                                     \
+V_(AHP, 26, 26, Bits)                                                          \
+V_(DN, 25, 25, Bits)                                                           \
+V_(FZ, 24, 24, Bits)                                                           \
+V_(RMode, 23, 22, Bits)                                                        \
+M_(FPCR, AHP_mask | DN_mask | FZ_mask | RMode_mask)
+
+
+// Fields offsets.
+#define DECLARE_FIELDS_OFFSETS(Name, HighBit, LowBit, X)                       \
+const int Name##_offset = LowBit;                                              \
+const int Name##_width = HighBit - LowBit + 1;                                 \
+const uint32_t Name##_mask = ((1 << Name##_width) - 1) << LowBit;
+#define NOTHING(A, B)
+INSTRUCTION_FIELDS_LIST(DECLARE_FIELDS_OFFSETS)
+SYSTEM_REGISTER_FIELDS_LIST(DECLARE_FIELDS_OFFSETS, NOTHING)
+#undef NOTHING
+#undef DECLARE_FIELDS_BITS
+
+// ImmPCRel is a compound field (not present in INSTRUCTION_FIELDS_LIST), formed
+// from ImmPCRelLo and ImmPCRelHi.
+const int ImmPCRel_mask = ImmPCRelLo_mask | ImmPCRelHi_mask;
+
+// Condition codes.
+enum Condition {
+  eq = 0,
+  ne = 1,
+  hs = 2,
+  lo = 3,
+  mi = 4,
+  pl = 5,
+  vs = 6,
+  vc = 7,
+  hi = 8,
+  ls = 9,
+  ge = 10,
+  lt = 11,
+  gt = 12,
+  le = 13,
+  al = 14,
+  nv = 15  // Behaves as always/al.
+};
+
+inline Condition InvertCondition(Condition cond) {
+  // Conditions al and nv behave identically, as "always true". They can't be
+  // inverted, because there is no "always false" condition.
+  ASSERT((cond != al) && (cond != nv));
+  return static_cast<Condition>(cond ^ 1);
+}
+
+enum FlagsUpdate {
+  SetFlags   = 1,
+  LeaveFlags = 0
+};
+
+enum StatusFlags {
+  NoFlag    = 0,
+
+  // Derive the flag combinations from the system register bit descriptions.
+  NFlag     = N_mask,
+  ZFlag     = Z_mask,
+  CFlag     = C_mask,
+  VFlag     = V_mask,
+  NZFlag    = NFlag | ZFlag,
+  NCFlag    = NFlag | CFlag,
+  NVFlag    = NFlag | VFlag,
+  ZCFlag    = ZFlag | CFlag,
+  ZVFlag    = ZFlag | VFlag,
+  CVFlag    = CFlag | VFlag,
+  NZCFlag   = NFlag | ZFlag | CFlag,
+  NZVFlag   = NFlag | ZFlag | VFlag,
+  NCVFlag   = NFlag | CFlag | VFlag,
+  ZCVFlag   = ZFlag | CFlag | VFlag,
+  NZCVFlag  = NFlag | ZFlag | CFlag | VFlag,
+
+  // Floating-point comparison results.
+  FPEqualFlag       = ZCFlag,
+  FPLessThanFlag    = NFlag,
+  FPGreaterThanFlag = CFlag,
+  FPUnorderedFlag   = CVFlag
+};
+
+enum Shift {
+  NO_SHIFT = -1,
+  LSL = 0x0,
+  LSR = 0x1,
+  ASR = 0x2,
+  ROR = 0x3
+};
+
+enum Extend {
+  NO_EXTEND = -1,
+  UXTB      = 0,
+  UXTH      = 1,
+  UXTW      = 2,
+  UXTX      = 3,
+  SXTB      = 4,
+  SXTH      = 5,
+  SXTW      = 6,
+  SXTX      = 7
+};
+
+enum SystemHint {
+  NOP   = 0,
+  YIELD = 1,
+  WFE   = 2,
+  WFI   = 3,
+  SEV   = 4,
+  SEVL  = 5
+};
+
+// 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).
+enum SystemRegister {
+  NZCV = ((0x1 << SysO0_offset) |
+          (0x3 << SysOp1_offset) |
+          (0x4 << CRn_offset) |
+          (0x2 << CRm_offset) |
+          (0x0 << SysOp2_offset)) >> ImmSystemRegister_offset,
+  FPCR = ((0x1 << SysO0_offset) |
+          (0x3 << SysOp1_offset) |
+          (0x4 << CRn_offset) |
+          (0x4 << CRm_offset) |
+          (0x0 << SysOp2_offset)) >> ImmSystemRegister_offset
+};
+
+// Instruction enumerations.
+//
+// These are the masks that define a class of instructions, and the list of
+// instructions within each class. Each enumeration has a Fixed, FMask and
+// Mask value.
+//
+// Fixed: The fixed bits in this instruction class.
+// FMask: The mask used to extract the fixed bits in the class.
+// Mask:  The mask used to identify the instructions within a class.
+//
+// The enumerations can be used like this:
+//
+// ASSERT(instr->Mask(PCRelAddressingFMask) == PCRelAddressingFixed);
+// switch(instr->Mask(PCRelAddressingMask)) {
+//   case ADR:  Format("adr 'Xd, 'AddrPCRelByte"); break;
+//   case ADRP: Format("adrp 'Xd, 'AddrPCRelPage"); break;
+//   default:   printf("Unknown instruction\n");
+// }
+
+
+// Generic fields.
+enum GenericInstrField {
+  SixtyFourBits        = 0x80000000,
+  ThirtyTwoBits        = 0x00000000,
+  FP32                 = 0x00000000,
+  FP64                 = 0x00400000
+};
+
+// PC relative addressing.
+enum PCRelAddressingOp {
+  PCRelAddressingFixed = 0x10000000,
+  PCRelAddressingFMask = 0x1F000000,
+  PCRelAddressingMask  = 0x9F000000,
+  ADR                  = PCRelAddressingFixed | 0x00000000,
+  ADRP                 = PCRelAddressingFixed | 0x80000000
+};
+
+// Add/sub (immediate, shifted and extended.)
+const int kSFOffset = 31;
+enum AddSubOp {
+  AddSubOpMask      = 0x60000000,
+  AddSubSetFlagsBit = 0x20000000,
+  ADD               = 0x00000000,
+  ADDS              = ADD | AddSubSetFlagsBit,
+  SUB               = 0x40000000,
+  SUBS              = SUB | AddSubSetFlagsBit
+};
+
+#define ADD_SUB_OP_LIST(V)  \
+  V(ADD),                   \
+  V(ADDS),                  \
+  V(SUB),                   \
+  V(SUBS)
+
+enum AddSubImmediateOp {
+  AddSubImmediateFixed = 0x11000000,
+  AddSubImmediateFMask = 0x1F000000,
+  AddSubImmediateMask  = 0xFF000000,
+  #define ADD_SUB_IMMEDIATE(A)           \
+  A##_w_imm = AddSubImmediateFixed | A,  \
+  A##_x_imm = AddSubImmediateFixed | A | SixtyFourBits
+  ADD_SUB_OP_LIST(ADD_SUB_IMMEDIATE)
+  #undef ADD_SUB_IMMEDIATE
+};
+
+enum AddSubShiftedOp {
+  AddSubShiftedFixed   = 0x0B000000,
+  AddSubShiftedFMask   = 0x1F200000,
+  AddSubShiftedMask    = 0xFF200000,
+  #define ADD_SUB_SHIFTED(A)             \
+  A##_w_shift = AddSubShiftedFixed | A,  \
+  A##_x_shift = AddSubShiftedFixed | A | SixtyFourBits
+  ADD_SUB_OP_LIST(ADD_SUB_SHIFTED)
+  #undef ADD_SUB_SHIFTED
+};
+
+enum AddSubExtendedOp {
+  AddSubExtendedFixed  = 0x0B200000,
+  AddSubExtendedFMask  = 0x1F200000,
+  AddSubExtendedMask   = 0xFFE00000,
+  #define ADD_SUB_EXTENDED(A)           \
+  A##_w_ext = AddSubExtendedFixed | A,  \
+  A##_x_ext = AddSubExtendedFixed | A | SixtyFourBits
+  ADD_SUB_OP_LIST(ADD_SUB_EXTENDED)
+  #undef ADD_SUB_EXTENDED
+};
+
+// Add/sub with carry.
+enum AddSubWithCarryOp {
+  AddSubWithCarryFixed = 0x1A000000,
+  AddSubWithCarryFMask = 0x1FE00000,
+  AddSubWithCarryMask  = 0xFFE0FC00,
+  ADC_w                = AddSubWithCarryFixed | ADD,
+  ADC_x                = AddSubWithCarryFixed | ADD | SixtyFourBits,
+  ADC                  = ADC_w,
+  ADCS_w               = AddSubWithCarryFixed | ADDS,
+  ADCS_x               = AddSubWithCarryFixed | ADDS | SixtyFourBits,
+  SBC_w                = AddSubWithCarryFixed | SUB,
+  SBC_x                = AddSubWithCarryFixed | SUB | SixtyFourBits,
+  SBC                  = SBC_w,
+  SBCS_w               = AddSubWithCarryFixed | SUBS,
+  SBCS_x               = AddSubWithCarryFixed | SUBS | SixtyFourBits
+};
+
+
+// Logical (immediate and shifted register).
+enum LogicalOp {
+  LogicalOpMask = 0x60200000,
+  NOT   = 0x00200000,
+  AND   = 0x00000000,
+  BIC   = AND | NOT,
+  ORR   = 0x20000000,
+  ORN   = ORR | NOT,
+  EOR   = 0x40000000,
+  EON   = EOR | NOT,
+  ANDS  = 0x60000000,
+  BICS  = ANDS | NOT
+};
+
+// Logical immediate.
+enum LogicalImmediateOp {
+  LogicalImmediateFixed = 0x12000000,
+  LogicalImmediateFMask = 0x1F800000,
+  LogicalImmediateMask  = 0xFF800000,
+  AND_w_imm   = LogicalImmediateFixed | AND,
+  AND_x_imm   = LogicalImmediateFixed | AND | SixtyFourBits,
+  ORR_w_imm   = LogicalImmediateFixed | ORR,
+  ORR_x_imm   = LogicalImmediateFixed | ORR | SixtyFourBits,
+  EOR_w_imm   = LogicalImmediateFixed | EOR,
+  EOR_x_imm   = LogicalImmediateFixed | EOR | SixtyFourBits,
+  ANDS_w_imm  = LogicalImmediateFixed | ANDS,
+  ANDS_x_imm  = LogicalImmediateFixed | ANDS | SixtyFourBits
+};
+
+// Logical shifted register.
+enum LogicalShiftedOp {
+  LogicalShiftedFixed = 0x0A000000,
+  LogicalShiftedFMask = 0x1F000000,
+  LogicalShiftedMask  = 0xFF200000,
+  AND_w               = LogicalShiftedFixed | AND,
+  AND_x               = LogicalShiftedFixed | AND | SixtyFourBits,
+  AND_shift           = AND_w,
+  BIC_w               = LogicalShiftedFixed | BIC,
+  BIC_x               = LogicalShiftedFixed | BIC | SixtyFourBits,
+  BIC_shift           = BIC_w,
+  ORR_w               = LogicalShiftedFixed | ORR,
+  ORR_x               = LogicalShiftedFixed | ORR | SixtyFourBits,
+  ORR_shift           = ORR_w,
+  ORN_w               = LogicalShiftedFixed | ORN,
+  ORN_x               = LogicalShiftedFixed | ORN | SixtyFourBits,
+  ORN_shift           = ORN_w,
+  EOR_w               = LogicalShiftedFixed | EOR,
+  EOR_x               = LogicalShiftedFixed | EOR | SixtyFourBits,
+  EOR_shift           = EOR_w,
+  EON_w               = LogicalShiftedFixed | EON,
+  EON_x               = LogicalShiftedFixed | EON | SixtyFourBits,
+  EON_shift           = EON_w,
+  ANDS_w              = LogicalShiftedFixed | ANDS,
+  ANDS_x              = LogicalShiftedFixed | ANDS | SixtyFourBits,
+  ANDS_shift          = ANDS_w,
+  BICS_w              = LogicalShiftedFixed | BICS,
+  BICS_x              = LogicalShiftedFixed | BICS | SixtyFourBits,
+  BICS_shift          = BICS_w
+};
+
+// Move wide immediate.
+enum MoveWideImmediateOp {
+  MoveWideImmediateFixed = 0x12800000,
+  MoveWideImmediateFMask = 0x1F800000,
+  MoveWideImmediateMask  = 0xFF800000,
+  MOVN                   = 0x00000000,
+  MOVZ                   = 0x40000000,
+  MOVK                   = 0x60000000,
+  MOVN_w                 = MoveWideImmediateFixed | MOVN,
+  MOVN_x                 = MoveWideImmediateFixed | MOVN | SixtyFourBits,
+  MOVZ_w                 = MoveWideImmediateFixed | MOVZ,
+  MOVZ_x                 = MoveWideImmediateFixed | MOVZ | SixtyFourBits,
+  MOVK_w                 = MoveWideImmediateFixed | MOVK,
+  MOVK_x                 = MoveWideImmediateFixed | MOVK | SixtyFourBits
+};
+
+// Bitfield.
+const int kBitfieldNOffset = 22;
+enum BitfieldOp {
+  BitfieldFixed = 0x13000000,
+  BitfieldFMask = 0x1F800000,
+  BitfieldMask  = 0xFF800000,
+  SBFM_w        = BitfieldFixed | 0x00000000,
+  SBFM_x        = BitfieldFixed | 0x80000000,
+  SBFM          = SBFM_w,
+  BFM_w         = BitfieldFixed | 0x20000000,
+  BFM_x         = BitfieldFixed | 0xA0000000,
+  BFM           = BFM_w,
+  UBFM_w        = BitfieldFixed | 0x40000000,
+  UBFM_x        = BitfieldFixed | 0xC0000000,
+  UBFM          = UBFM_w
+  // Bitfield N field.
+};
+
+// Extract.
+enum ExtractOp {
+  ExtractFixed = 0x13800000,
+  ExtractFMask = 0x1F800000,
+  ExtractMask  = 0xFFA00000,
+  EXTR_w       = ExtractFixed | 0x00000000,
+  EXTR_x       = ExtractFixed | 0x80000000,
+  EXTR         = EXTR_w
+};
+
+// Unconditional branch.
+enum UnconditionalBranchOp {
+  UnconditionalBranchFixed = 0x14000000,
+  UnconditionalBranchFMask = 0x7C000000,
+  UnconditionalBranchMask  = 0xFC000000,
+  B                        = UnconditionalBranchFixed | 0x00000000,
+  BL                       = UnconditionalBranchFixed | 0x80000000
+};
+
+// Unconditional branch to register.
+enum UnconditionalBranchToRegisterOp {
+  UnconditionalBranchToRegisterFixed = 0xD6000000,
+  UnconditionalBranchToRegisterFMask = 0xFE000000,
+  UnconditionalBranchToRegisterMask  = 0xFFFFFC1F,
+  BR      = UnconditionalBranchToRegisterFixed | 0x001F0000,
+  BLR     = UnconditionalBranchToRegisterFixed | 0x003F0000,
+  RET     = UnconditionalBranchToRegisterFixed | 0x005F0000
+};
+
+// Compare and branch.
+enum CompareBranchOp {
+  CompareBranchFixed = 0x34000000,
+  CompareBranchFMask = 0x7E000000,
+  CompareBranchMask  = 0xFF000000,
+  CBZ_w              = CompareBranchFixed | 0x00000000,
+  CBZ_x              = CompareBranchFixed | 0x80000000,
+  CBZ                = CBZ_w,
+  CBNZ_w             = CompareBranchFixed | 0x01000000,
+  CBNZ_x             = CompareBranchFixed | 0x81000000,
+  CBNZ               = CBNZ_w
+};
+
+// Test and branch.
+enum TestBranchOp {
+  TestBranchFixed = 0x36000000,
+  TestBranchFMask = 0x7E000000,
+  TestBranchMask  = 0x7F000000,
+  TBZ             = TestBranchFixed | 0x00000000,
+  TBNZ            = TestBranchFixed | 0x01000000
+};
+
+// Conditional branch.
+enum ConditionalBranchOp {
+  ConditionalBranchFixed = 0x54000000,
+  ConditionalBranchFMask = 0xFE000000,
+  ConditionalBranchMask  = 0xFF000010,
+  B_cond                 = ConditionalBranchFixed | 0x00000000
+};
+
+// System.
+// System instruction encoding is complicated because some instructions use op
+// and CR fields to encode parameters. To handle this cleanly, the system
+// instructions are split into more than one enum.
+
+enum SystemOp {
+  SystemFixed = 0xD5000000,
+  SystemFMask = 0xFFC00000
+};
+
+enum SystemSysRegOp {
+  SystemSysRegFixed = 0xD5100000,
+  SystemSysRegFMask = 0xFFD00000,
+  SystemSysRegMask  = 0xFFF00000,
+  MRS               = SystemSysRegFixed | 0x00200000,
+  MSR               = SystemSysRegFixed | 0x00000000
+};
+
+enum SystemHintOp {
+  SystemHintFixed = 0xD503201F,
+  SystemHintFMask = 0xFFFFF01F,
+  SystemHintMask  = 0xFFFFF01F,
+  HINT            = SystemHintFixed | 0x00000000
+};
+
+// Exception.
+enum ExceptionOp {
+  ExceptionFixed = 0xD4000000,
+  ExceptionFMask = 0xFF000000,
+  ExceptionMask  = 0xFFE0001F,
+  HLT            = ExceptionFixed | 0x00400000,
+  BRK            = ExceptionFixed | 0x00200000,
+  SVC            = ExceptionFixed | 0x00000001,
+  HVC            = ExceptionFixed | 0x00000002,
+  SMC            = ExceptionFixed | 0x00000003,
+  DCPS1          = ExceptionFixed | 0x00A00001,
+  DCPS2          = ExceptionFixed | 0x00A00002,
+  DCPS3          = ExceptionFixed | 0x00A00003
+};
+
+// Any load or store.
+enum LoadStoreAnyOp {
+  LoadStoreAnyFMask = 0x0a000000,
+  LoadStoreAnyFixed = 0x08000000
+};
+
+#define LOAD_STORE_PAIR_OP_LIST(V)  \
+  V(STP, w,   0x00000000),          \
+  V(LDP, w,   0x00400000),          \
+  V(LDPSW, x, 0x40400000),          \
+  V(STP, x,   0x80000000),          \
+  V(LDP, x,   0x80400000),          \
+  V(STP, s,   0x04000000),          \
+  V(LDP, s,   0x04400000),          \
+  V(STP, d,   0x44000000),          \
+  V(LDP, d,   0x44400000)
+
+// Load/store pair (post, pre and offset.)
+enum LoadStorePairOp {
+  LoadStorePairMask = 0xC4400000,
+  LoadStorePairLBit = 1 << 22,
+  #define LOAD_STORE_PAIR(A, B, C) \
+  A##_##B = C
+  LOAD_STORE_PAIR_OP_LIST(LOAD_STORE_PAIR)
+  #undef LOAD_STORE_PAIR
+};
+
+enum LoadStorePairPostIndexOp {
+  LoadStorePairPostIndexFixed = 0x28800000,
+  LoadStorePairPostIndexFMask = 0x3B800000,
+  LoadStorePairPostIndexMask  = 0xFFC00000,
+  #define LOAD_STORE_PAIR_POST_INDEX(A, B, C)  \
+  A##_##B##_post = LoadStorePairPostIndexFixed | A##_##B
+  LOAD_STORE_PAIR_OP_LIST(LOAD_STORE_PAIR_POST_INDEX)
+  #undef LOAD_STORE_PAIR_POST_INDEX
+};
+
+enum LoadStorePairPreIndexOp {
+  LoadStorePairPreIndexFixed = 0x29800000,
+  LoadStorePairPreIndexFMask = 0x3B800000,
+  LoadStorePairPreIndexMask  = 0xFFC00000,
+  #define LOAD_STORE_PAIR_PRE_INDEX(A, B, C)  \
+  A##_##B##_pre = LoadStorePairPreIndexFixed | A##_##B
+  LOAD_STORE_PAIR_OP_LIST(LOAD_STORE_PAIR_PRE_INDEX)
+  #undef LOAD_STORE_PAIR_PRE_INDEX
+};
+
+enum LoadStorePairOffsetOp {
+  LoadStorePairOffsetFixed = 0x29000000,
+  LoadStorePairOffsetFMask = 0x3B800000,
+  LoadStorePairOffsetMask  = 0xFFC00000,
+  #define LOAD_STORE_PAIR_OFFSET(A, B, C)  \
+  A##_##B##_off = LoadStorePairOffsetFixed | A##_##B
+  LOAD_STORE_PAIR_OP_LIST(LOAD_STORE_PAIR_OFFSET)
+  #undef LOAD_STORE_PAIR_OFFSET
+};
+
+enum LoadStorePairNonTemporalOp {
+  LoadStorePairNonTemporalFixed = 0x28000000,
+  LoadStorePairNonTemporalFMask = 0x3B800000,
+  LoadStorePairNonTemporalMask  = 0xFFC00000,
+  STNP_w = LoadStorePairNonTemporalFixed | STP_w,
+  LDNP_w = LoadStorePairNonTemporalFixed | LDP_w,
+  STNP_x = LoadStorePairNonTemporalFixed | STP_x,
+  LDNP_x = LoadStorePairNonTemporalFixed | LDP_x,
+  STNP_s = LoadStorePairNonTemporalFixed | STP_s,
+  LDNP_s = LoadStorePairNonTemporalFixed | LDP_s,
+  STNP_d = LoadStorePairNonTemporalFixed | STP_d,
+  LDNP_d = LoadStorePairNonTemporalFixed | LDP_d
+};
+
+// Load literal.
+enum LoadLiteralOp {
+  LoadLiteralFixed = 0x18000000,
+  LoadLiteralFMask = 0x3B000000,
+  LoadLiteralMask  = 0xFF000000,
+  LDR_w_lit        = LoadLiteralFixed | 0x00000000,
+  LDR_x_lit        = LoadLiteralFixed | 0x40000000,
+  LDRSW_x_lit      = LoadLiteralFixed | 0x80000000,
+  PRFM_lit         = LoadLiteralFixed | 0xC0000000,
+  LDR_s_lit        = LoadLiteralFixed | 0x04000000,
+  LDR_d_lit        = LoadLiteralFixed | 0x44000000
+};
+
+#define LOAD_STORE_OP_LIST(V)     \
+  V(ST, RB, w,  0x00000000),  \
+  V(ST, RH, w,  0x40000000),  \
+  V(ST, R, w,   0x80000000),  \
+  V(ST, R, x,   0xC0000000),  \
+  V(LD, RB, w,  0x00400000),  \
+  V(LD, RH, w,  0x40400000),  \
+  V(LD, R, w,   0x80400000),  \
+  V(LD, R, x,   0xC0400000),  \
+  V(LD, RSB, x, 0x00800000),  \
+  V(LD, RSH, x, 0x40800000),  \
+  V(LD, RSW, x, 0x80800000),  \
+  V(LD, RSB, w, 0x00C00000),  \
+  V(LD, RSH, w, 0x40C00000),  \
+  V(ST, R, s,   0x84000000),  \
+  V(ST, R, d,   0xC4000000),  \
+  V(LD, R, s,   0x84400000),  \
+  V(LD, R, d,   0xC4400000)
+
+
+// Load/store unscaled offset.
+enum LoadStoreUnscaledOffsetOp {
+  LoadStoreUnscaledOffsetFixed = 0x38000000,
+  LoadStoreUnscaledOffsetFMask = 0x3B200C00,
+  LoadStoreUnscaledOffsetMask  = 0xFFE00C00,
+  #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,
+  LoadStorePostIndexFMask = 0x3B200C00,
+  LoadStorePostIndexMask  = 0xFFE00C00,
+  #define LOAD_STORE_POST_INDEX(A, B, C, D)  \
+  A##B##_##C##_post = LoadStorePostIndexFixed | D
+  LOAD_STORE_OP_LIST(LOAD_STORE_POST_INDEX)
+  #undef LOAD_STORE_POST_INDEX
+};
+
+// Load/store pre index.
+enum LoadStorePreIndex {
+  LoadStorePreIndexFixed = 0x38000C00,
+  LoadStorePreIndexFMask = 0x3B200C00,
+  LoadStorePreIndexMask  = 0xFFE00C00,
+  #define LOAD_STORE_PRE_INDEX(A, B, C, D)  \
+  A##B##_##C##_pre = LoadStorePreIndexFixed | D
+  LOAD_STORE_OP_LIST(LOAD_STORE_PRE_INDEX)
+  #undef LOAD_STORE_PRE_INDEX
+};
+
+// Load/store unsigned offset.
+enum LoadStoreUnsignedOffset {
+  LoadStoreUnsignedOffsetFixed = 0x39000000,
+  LoadStoreUnsignedOffsetFMask = 0x3B000000,
+  LoadStoreUnsignedOffsetMask  = 0xFFC00000,
+  PRFM_unsigned                = LoadStoreUnsignedOffsetFixed | PRFM,
+  #define LOAD_STORE_UNSIGNED_OFFSET(A, B, C, D) \
+  A##B##_##C##_unsigned = LoadStoreUnsignedOffsetFixed | D
+  LOAD_STORE_OP_LIST(LOAD_STORE_UNSIGNED_OFFSET)
+  #undef LOAD_STORE_UNSIGNED_OFFSET
+};
+
+// Load/store register offset.
+enum LoadStoreRegisterOffset {
+  LoadStoreRegisterOffsetFixed = 0x38200800,
+  LoadStoreRegisterOffsetFMask = 0x3B200C00,
+  LoadStoreRegisterOffsetMask  = 0xFFE00C00,
+  PRFM_reg                     = LoadStoreRegisterOffsetFixed | PRFM,
+  #define LOAD_STORE_REGISTER_OFFSET(A, B, C, D) \
+  A##B##_##C##_reg = LoadStoreRegisterOffsetFixed | D
+  LOAD_STORE_OP_LIST(LOAD_STORE_REGISTER_OFFSET)
+  #undef LOAD_STORE_REGISTER_OFFSET
+};
+
+// Conditional compare.
+enum ConditionalCompareOp {
+  ConditionalCompareMask = 0x60000000,
+  CCMN                   = 0x20000000,
+  CCMP                   = 0x60000000
+};
+
+// Conditional compare register.
+enum ConditionalCompareRegisterOp {
+  ConditionalCompareRegisterFixed = 0x1A400000,
+  ConditionalCompareRegisterFMask = 0x1FE00800,
+  ConditionalCompareRegisterMask  = 0xFFE00C10,
+  CCMN_w = ConditionalCompareRegisterFixed | CCMN,
+  CCMN_x = ConditionalCompareRegisterFixed | SixtyFourBits | CCMN,
+  CCMP_w = ConditionalCompareRegisterFixed | CCMP,
+  CCMP_x = ConditionalCompareRegisterFixed | SixtyFourBits | CCMP
+};
+
+// Conditional compare immediate.
+enum ConditionalCompareImmediateOp {
+  ConditionalCompareImmediateFixed = 0x1A400800,
+  ConditionalCompareImmediateFMask = 0x1FE00800,
+  ConditionalCompareImmediateMask  = 0xFFE00C10,
+  CCMN_w_imm = ConditionalCompareImmediateFixed | CCMN,
+  CCMN_x_imm = ConditionalCompareImmediateFixed | SixtyFourBits | CCMN,
+  CCMP_w_imm = ConditionalCompareImmediateFixed | CCMP,
+  CCMP_x_imm = ConditionalCompareImmediateFixed | SixtyFourBits | CCMP
+};
+
+// Conditional select.
+enum ConditionalSelectOp {
+  ConditionalSelectFixed = 0x1A800000,
+  ConditionalSelectFMask = 0x1FE00000,
+  ConditionalSelectMask  = 0xFFE00C00,
+  CSEL_w                 = ConditionalSelectFixed | 0x00000000,
+  CSEL_x                 = ConditionalSelectFixed | 0x80000000,
+  CSEL                   = CSEL_w,
+  CSINC_w                = ConditionalSelectFixed | 0x00000400,
+  CSINC_x                = ConditionalSelectFixed | 0x80000400,
+  CSINC                  = CSINC_w,
+  CSINV_w                = ConditionalSelectFixed | 0x40000000,
+  CSINV_x                = ConditionalSelectFixed | 0xC0000000,
+  CSINV                  = CSINV_w,
+  CSNEG_w                = ConditionalSelectFixed | 0x40000400,
+  CSNEG_x                = ConditionalSelectFixed | 0xC0000400,
+  CSNEG                  = CSNEG_w
+};
+
+// Data processing 1 source.
+enum DataProcessing1SourceOp {
+  DataProcessing1SourceFixed = 0x5AC00000,
+  DataProcessing1SourceFMask = 0x5FE00000,
+  DataProcessing1SourceMask  = 0xFFFFFC00,
+  RBIT    = DataProcessing1SourceFixed | 0x00000000,
+  RBIT_w  = RBIT,
+  RBIT_x  = RBIT | SixtyFourBits,
+  REV16   = DataProcessing1SourceFixed | 0x00000400,
+  REV16_w = REV16,
+  REV16_x = REV16 | SixtyFourBits,
+  REV     = DataProcessing1SourceFixed | 0x00000800,
+  REV_w   = REV,
+  REV32_x = REV | SixtyFourBits,
+  REV_x   = DataProcessing1SourceFixed | SixtyFourBits | 0x00000C00,
+  CLZ     = DataProcessing1SourceFixed | 0x00001000,
+  CLZ_w   = CLZ,
+  CLZ_x   = CLZ | SixtyFourBits,
+  CLS     = DataProcessing1SourceFixed | 0x00001400,
+  CLS_w   = CLS,
+  CLS_x   = CLS | SixtyFourBits
+};
+
+// Data processing 2 source.
+enum DataProcessing2SourceOp {
+  DataProcessing2SourceFixed = 0x1AC00000,
+  DataProcessing2SourceFMask = 0x5FE00000,
+  DataProcessing2SourceMask  = 0xFFE0FC00,
+  UDIV_w  = DataProcessing2SourceFixed | 0x00000800,
+  UDIV_x  = DataProcessing2SourceFixed | 0x80000800,
+  UDIV    = UDIV_w,
+  SDIV_w  = DataProcessing2SourceFixed | 0x00000C00,
+  SDIV_x  = DataProcessing2SourceFixed | 0x80000C00,
+  SDIV    = SDIV_w,
+  LSLV_w  = DataProcessing2SourceFixed | 0x00002000,
+  LSLV_x  = DataProcessing2SourceFixed | 0x80002000,
+  LSLV    = LSLV_w,
+  LSRV_w  = DataProcessing2SourceFixed | 0x00002400,
+  LSRV_x  = DataProcessing2SourceFixed | 0x80002400,
+  LSRV    = LSRV_w,
+  ASRV_w  = DataProcessing2SourceFixed | 0x00002800,
+  ASRV_x  = DataProcessing2SourceFixed | 0x80002800,
+  ASRV    = ASRV_w,
+  RORV_w  = DataProcessing2SourceFixed | 0x00002C00,
+  RORV_x  = DataProcessing2SourceFixed | 0x80002C00,
+  RORV    = RORV_w,
+  CRC32B  = DataProcessing2SourceFixed | 0x00004000,
+  CRC32H  = DataProcessing2SourceFixed | 0x00004400,
+  CRC32W  = DataProcessing2SourceFixed | 0x00004800,
+  CRC32X  = DataProcessing2SourceFixed | SixtyFourBits | 0x00004C00,
+  CRC32CB = DataProcessing2SourceFixed | 0x00005000,
+  CRC32CH = DataProcessing2SourceFixed | 0x00005400,
+  CRC32CW = DataProcessing2SourceFixed | 0x00005800,
+  CRC32CX = DataProcessing2SourceFixed | SixtyFourBits | 0x00005C00
+};
+
+// Data processing 3 source.
+enum DataProcessing3SourceOp {
+  DataProcessing3SourceFixed = 0x1B000000,
+  DataProcessing3SourceFMask = 0x1F000000,
+  DataProcessing3SourceMask  = 0xFFE08000,
+  MADD_w                     = DataProcessing3SourceFixed | 0x00000000,
+  MADD_x                     = DataProcessing3SourceFixed | 0x80000000,
+  MADD                       = MADD_w,
+  MSUB_w                     = DataProcessing3SourceFixed | 0x00008000,
+  MSUB_x                     = DataProcessing3SourceFixed | 0x80008000,
+  MSUB                       = MSUB_w,
+  SMADDL_x                   = DataProcessing3SourceFixed | 0x80200000,
+  SMSUBL_x                   = DataProcessing3SourceFixed | 0x80208000,
+  SMULH_x                    = DataProcessing3SourceFixed | 0x80400000,
+  UMADDL_x                   = DataProcessing3SourceFixed | 0x80A00000,
+  UMSUBL_x                   = DataProcessing3SourceFixed | 0x80A08000,
+  UMULH_x                    = DataProcessing3SourceFixed | 0x80C00000
+};
+
+// Floating point compare.
+enum FPCompareOp {
+  FPCompareFixed = 0x1E202000,
+  FPCompareFMask = 0x5F203C00,
+  FPCompareMask  = 0xFFE0FC1F,
+  FCMP_s         = FPCompareFixed | 0x00000000,
+  FCMP_d         = FPCompareFixed | FP64 | 0x00000000,
+  FCMP           = FCMP_s,
+  FCMP_s_zero    = FPCompareFixed | 0x00000008,
+  FCMP_d_zero    = FPCompareFixed | FP64 | 0x00000008,
+  FCMP_zero      = FCMP_s_zero,
+  FCMPE_s        = FPCompareFixed | 0x00000010,
+  FCMPE_d        = FPCompareFixed | FP64 | 0x00000010,
+  FCMPE_s_zero   = FPCompareFixed | 0x00000018,
+  FCMPE_d_zero   = FPCompareFixed | FP64 | 0x00000018
+};
+
+// Floating point conditional compare.
+enum FPConditionalCompareOp {
+  FPConditionalCompareFixed = 0x1E200400,
+  FPConditionalCompareFMask = 0x5F200C00,
+  FPConditionalCompareMask  = 0xFFE00C10,
+  FCCMP_s                   = FPConditionalCompareFixed | 0x00000000,
+  FCCMP_d                   = FPConditionalCompareFixed | FP64 | 0x00000000,
+  FCCMP                     = FCCMP_s,
+  FCCMPE_s                  = FPConditionalCompareFixed | 0x00000010,
+  FCCMPE_d                  = FPConditionalCompareFixed | FP64 | 0x00000010,
+  FCCMPE                    = FCCMPE_s
+};
+
+// Floating point conditional select.
+enum FPConditionalSelectOp {
+  FPConditionalSelectFixed = 0x1E200C00,
+  FPConditionalSelectFMask = 0x5F200C00,
+  FPConditionalSelectMask  = 0xFFE00C00,
+  FCSEL_s                  = FPConditionalSelectFixed | 0x00000000,
+  FCSEL_d                  = FPConditionalSelectFixed | FP64 | 0x00000000,
+  FCSEL                    = FCSEL_s
+};
+
+// Floating point immediate.
+enum FPImmediateOp {
+  FPImmediateFixed = 0x1E201000,
+  FPImmediateFMask = 0x5F201C00,
+  FPImmediateMask  = 0xFFE01C00,
+  FMOV_s_imm       = FPImmediateFixed | 0x00000000,
+  FMOV_d_imm       = FPImmediateFixed | FP64 | 0x00000000
+};
+
+// Floating point data processing 1 source.
+enum FPDataProcessing1SourceOp {
+  FPDataProcessing1SourceFixed = 0x1E204000,
+  FPDataProcessing1SourceFMask = 0x5F207C00,
+  FPDataProcessing1SourceMask  = 0xFFFFFC00,
+  FMOV_s   = FPDataProcessing1SourceFixed | 0x00000000,
+  FMOV_d   = FPDataProcessing1SourceFixed | FP64 | 0x00000000,
+  FMOV     = FMOV_s,
+  FABS_s   = FPDataProcessing1SourceFixed | 0x00008000,
+  FABS_d   = FPDataProcessing1SourceFixed | FP64 | 0x00008000,
+  FABS     = FABS_s,
+  FNEG_s   = FPDataProcessing1SourceFixed | 0x00010000,
+  FNEG_d   = FPDataProcessing1SourceFixed | FP64 | 0x00010000,
+  FNEG     = FNEG_s,
+  FSQRT_s  = FPDataProcessing1SourceFixed | 0x00018000,
+  FSQRT_d  = FPDataProcessing1SourceFixed | FP64 | 0x00018000,
+  FSQRT    = FSQRT_s,
+  FCVT_ds  = FPDataProcessing1SourceFixed | 0x00028000,
+  FCVT_sd  = FPDataProcessing1SourceFixed | FP64 | 0x00020000,
+  FRINTN_s = FPDataProcessing1SourceFixed | 0x00040000,
+  FRINTN_d = FPDataProcessing1SourceFixed | FP64 | 0x00040000,
+  FRINTN   = FRINTN_s,
+  FRINTP_s = FPDataProcessing1SourceFixed | 0x00048000,
+  FRINTP_d = FPDataProcessing1SourceFixed | FP64 | 0x00048000,
+  FRINTM_s = FPDataProcessing1SourceFixed | 0x00050000,
+  FRINTM_d = FPDataProcessing1SourceFixed | FP64 | 0x00050000,
+  FRINTZ_s = FPDataProcessing1SourceFixed | 0x00058000,
+  FRINTZ_d = FPDataProcessing1SourceFixed | FP64 | 0x00058000,
+  FRINTZ   = FRINTZ_s,
+  FRINTA_s = FPDataProcessing1SourceFixed | 0x00060000,
+  FRINTA_d = FPDataProcessing1SourceFixed | FP64 | 0x00060000,
+  FRINTX_s = FPDataProcessing1SourceFixed | 0x00070000,
+  FRINTX_d = FPDataProcessing1SourceFixed | FP64 | 0x00070000,
+  FRINTI_s = FPDataProcessing1SourceFixed | 0x00078000,
+  FRINTI_d = FPDataProcessing1SourceFixed | FP64 | 0x00078000
+};
+
+// Floating point data processing 2 source.
+enum FPDataProcessing2SourceOp {
+  FPDataProcessing2SourceFixed = 0x1E200800,
+  FPDataProcessing2SourceFMask = 0x5F200C00,
+  FPDataProcessing2SourceMask  = 0xFFE0FC00,
+  FMUL     = FPDataProcessing2SourceFixed | 0x00000000,
+  FMUL_s   = FMUL,
+  FMUL_d   = FMUL | FP64,
+  FDIV     = FPDataProcessing2SourceFixed | 0x00001000,
+  FDIV_s   = FDIV,
+  FDIV_d   = FDIV | FP64,
+  FADD     = FPDataProcessing2SourceFixed | 0x00002000,
+  FADD_s   = FADD,
+  FADD_d   = FADD | FP64,
+  FSUB     = FPDataProcessing2SourceFixed | 0x00003000,
+  FSUB_s   = FSUB,
+  FSUB_d   = FSUB | FP64,
+  FMAX     = FPDataProcessing2SourceFixed | 0x00004000,
+  FMAX_s   = FMAX,
+  FMAX_d   = FMAX | FP64,
+  FMIN     = FPDataProcessing2SourceFixed | 0x00005000,
+  FMIN_s   = FMIN,
+  FMIN_d   = FMIN | FP64,
+  FMAXNM   = FPDataProcessing2SourceFixed | 0x00006000,
+  FMAXNM_s = FMAXNM,
+  FMAXNM_d = FMAXNM | FP64,
+  FMINNM   = FPDataProcessing2SourceFixed | 0x00007000,
+  FMINNM_s = FMINNM,
+  FMINNM_d = FMINNM | FP64,
+  FNMUL    = FPDataProcessing2SourceFixed | 0x00008000,
+  FNMUL_s  = FNMUL,
+  FNMUL_d  = FNMUL | FP64
+};
+
+// Floating point data processing 3 source.
+enum FPDataProcessing3SourceOp {
+  FPDataProcessing3SourceFixed = 0x1F000000,
+  FPDataProcessing3SourceFMask = 0x5F000000,
+  FPDataProcessing3SourceMask  = 0xFFE08000,
+  FMADD_s                      = FPDataProcessing3SourceFixed | 0x00000000,
+  FMSUB_s                      = FPDataProcessing3SourceFixed | 0x00008000,
+  FNMADD_s                     = FPDataProcessing3SourceFixed | 0x00200000,
+  FNMSUB_s                     = FPDataProcessing3SourceFixed | 0x00208000,
+  FMADD_d                      = FPDataProcessing3SourceFixed | 0x00400000,
+  FMSUB_d                      = FPDataProcessing3SourceFixed | 0x00408000,
+  FNMADD_d                     = FPDataProcessing3SourceFixed | 0x00600000,
+  FNMSUB_d                     = FPDataProcessing3SourceFixed | 0x00608000
+};
+
+// Conversion between floating point and integer.
+enum FPIntegerConvertOp {
+  FPIntegerConvertFixed = 0x1E200000,
+  FPIntegerConvertFMask = 0x5F20FC00,
+  FPIntegerConvertMask  = 0xFFFFFC00,
+  FCVTNS    = FPIntegerConvertFixed | 0x00000000,
+  FCVTNS_ws = FCVTNS,
+  FCVTNS_xs = FCVTNS | SixtyFourBits,
+  FCVTNS_wd = FCVTNS | FP64,
+  FCVTNS_xd = FCVTNS | SixtyFourBits | FP64,
+  FCVTNU    = FPIntegerConvertFixed | 0x00010000,
+  FCVTNU_ws = FCVTNU,
+  FCVTNU_xs = FCVTNU | SixtyFourBits,
+  FCVTNU_wd = FCVTNU | FP64,
+  FCVTNU_xd = FCVTNU | SixtyFourBits | FP64,
+  FCVTPS    = FPIntegerConvertFixed | 0x00080000,
+  FCVTPS_ws = FCVTPS,
+  FCVTPS_xs = FCVTPS | SixtyFourBits,
+  FCVTPS_wd = FCVTPS | FP64,
+  FCVTPS_xd = FCVTPS | SixtyFourBits | FP64,
+  FCVTPU    = FPIntegerConvertFixed | 0x00090000,
+  FCVTPU_ws = FCVTPU,
+  FCVTPU_xs = FCVTPU | SixtyFourBits,
+  FCVTPU_wd = FCVTPU | FP64,
+  FCVTPU_xd = FCVTPU | SixtyFourBits | FP64,
+  FCVTMS    = FPIntegerConvertFixed | 0x00100000,
+  FCVTMS_ws = FCVTMS,
+  FCVTMS_xs = FCVTMS | SixtyFourBits,
+  FCVTMS_wd = FCVTMS | FP64,
+  FCVTMS_xd = FCVTMS | SixtyFourBits | FP64,
+  FCVTMU    = FPIntegerConvertFixed | 0x00110000,
+  FCVTMU_ws = FCVTMU,
+  FCVTMU_xs = FCVTMU | SixtyFourBits,
+  FCVTMU_wd = FCVTMU | FP64,
+  FCVTMU_xd = FCVTMU | SixtyFourBits | FP64,
+  FCVTZS    = FPIntegerConvertFixed | 0x00180000,
+  FCVTZS_ws = FCVTZS,
+  FCVTZS_xs = FCVTZS | SixtyFourBits,
+  FCVTZS_wd = FCVTZS | FP64,
+  FCVTZS_xd = FCVTZS | SixtyFourBits | FP64,
+  FCVTZU    = FPIntegerConvertFixed | 0x00190000,
+  FCVTZU_ws = FCVTZU,
+  FCVTZU_xs = FCVTZU | SixtyFourBits,
+  FCVTZU_wd = FCVTZU | FP64,
+  FCVTZU_xd = FCVTZU | SixtyFourBits | FP64,
+  SCVTF     = FPIntegerConvertFixed | 0x00020000,
+  SCVTF_sw  = SCVTF,
+  SCVTF_sx  = SCVTF | SixtyFourBits,
+  SCVTF_dw  = SCVTF | FP64,
+  SCVTF_dx  = SCVTF | SixtyFourBits | FP64,
+  UCVTF     = FPIntegerConvertFixed | 0x00030000,
+  UCVTF_sw  = UCVTF,
+  UCVTF_sx  = UCVTF | SixtyFourBits,
+  UCVTF_dw  = UCVTF | FP64,
+  UCVTF_dx  = UCVTF | SixtyFourBits | FP64,
+  FCVTAS    = FPIntegerConvertFixed | 0x00040000,
+  FCVTAS_ws = FCVTAS,
+  FCVTAS_xs = FCVTAS | SixtyFourBits,
+  FCVTAS_wd = FCVTAS | FP64,
+  FCVTAS_xd = FCVTAS | SixtyFourBits | FP64,
+  FCVTAU    = FPIntegerConvertFixed | 0x00050000,
+  FCVTAU_ws = FCVTAU,
+  FCVTAU_xs = FCVTAU | SixtyFourBits,
+  FCVTAU_wd = FCVTAU | FP64,
+  FCVTAU_xd = FCVTAU | SixtyFourBits | FP64,
+  FMOV_ws   = FPIntegerConvertFixed | 0x00060000,
+  FMOV_sw   = FPIntegerConvertFixed | 0x00070000,
+  FMOV_xd   = FMOV_ws | SixtyFourBits | FP64,
+  FMOV_dx   = FMOV_sw | SixtyFourBits | FP64
+};
+
+// Conversion between fixed point and floating point.
+enum FPFixedPointConvertOp {
+  FPFixedPointConvertFixed = 0x1E000000,
+  FPFixedPointConvertFMask = 0x5F200000,
+  FPFixedPointConvertMask  = 0xFFFF0000,
+  FCVTZS_fixed    = FPFixedPointConvertFixed | 0x00180000,
+  FCVTZS_ws_fixed = FCVTZS_fixed,
+  FCVTZS_xs_fixed = FCVTZS_fixed | SixtyFourBits,
+  FCVTZS_wd_fixed = FCVTZS_fixed | FP64,
+  FCVTZS_xd_fixed = FCVTZS_fixed | SixtyFourBits | FP64,
+  FCVTZU_fixed    = FPFixedPointConvertFixed | 0x00190000,
+  FCVTZU_ws_fixed = FCVTZU_fixed,
+  FCVTZU_xs_fixed = FCVTZU_fixed | SixtyFourBits,
+  FCVTZU_wd_fixed = FCVTZU_fixed | FP64,
+  FCVTZU_xd_fixed = FCVTZU_fixed | SixtyFourBits | FP64,
+  SCVTF_fixed     = FPFixedPointConvertFixed | 0x00020000,
+  SCVTF_sw_fixed  = SCVTF_fixed,
+  SCVTF_sx_fixed  = SCVTF_fixed | SixtyFourBits,
+  SCVTF_dw_fixed  = SCVTF_fixed | FP64,
+  SCVTF_dx_fixed  = SCVTF_fixed | SixtyFourBits | FP64,
+  UCVTF_fixed     = FPFixedPointConvertFixed | 0x00030000,
+  UCVTF_sw_fixed  = UCVTF_fixed,
+  UCVTF_sx_fixed  = UCVTF_fixed | SixtyFourBits,
+  UCVTF_dw_fixed  = UCVTF_fixed | FP64,
+  UCVTF_dx_fixed  = UCVTF_fixed | SixtyFourBits | FP64
+};
+
+// Unimplemented and unallocated instructions. These are defined to make fixed
+// bit assertion easier.
+enum UnimplementedOp {
+  UnimplementedFixed = 0x00000000,
+  UnimplementedFMask = 0x00000000
+};
+
+enum UnallocatedOp {
+  UnallocatedFixed = 0x00000000,
+  UnallocatedFMask = 0x00000000
+};
+
+}  // namespace vixl
+
+#endif  // VIXL_A64_CONSTANTS_A64_H_
diff --git a/disas/libvixl/src/a64/cpu-a64.cc b/disas/libvixl/src/a64/cpu-a64.cc
new file mode 100644
index 0000000..8586563
--- /dev/null
+++ b/disas/libvixl/src/a64/cpu-a64.cc
@@ -0,0 +1,148 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "utils.h"
+#include "a64/cpu-a64.h"
+
+namespace vixl {
+
+// Initialise to smallest possible cache size.
+unsigned CPU::dcache_line_size_ = 1;
+unsigned CPU::icache_line_size_ = 1;
+
+
+// Currently computes I and D cache line size.
+void CPU::SetUp() {
+  uint32_t cache_type_register = GetCacheType();
+
+  // The cache type register holds information about the caches, including I
+  // D caches line size.
+  static const int kDCacheLineSizeShift = 16;
+  static const int kICacheLineSizeShift = 0;
+  static const uint32_t kDCacheLineSizeMask = 0xf << kDCacheLineSizeShift;
+  static const uint32_t kICacheLineSizeMask = 0xf << kICacheLineSizeShift;
+
+  // The cache type register holds the size of the I and D caches as a power of
+  // two.
+  uint32_t dcache_line_size_power_of_two =
+      (cache_type_register & kDCacheLineSizeMask) >> kDCacheLineSizeShift;
+  uint32_t icache_line_size_power_of_two =
+      (cache_type_register & kICacheLineSizeMask) >> kICacheLineSizeShift;
+
+  dcache_line_size_ = 1 << dcache_line_size_power_of_two;
+  icache_line_size_ = 1 << icache_line_size_power_of_two;
+}
+
+
+uint32_t CPU::GetCacheType() {
+#ifdef USE_SIMULATOR
+  // This will lead to a cache with 1 byte long lines, which is fine since the
+  // simulator will not need this information.
+  return 0;
+#else
+  uint32_t cache_type_register;
+  // Copy the content of the cache type register to a core register.
+  __asm__ __volatile__ ("mrs %[ctr], ctr_el0"  // NOLINT
+                        : [ctr] "=r" (cache_type_register));
+  return cache_type_register;
+#endif
+}
+
+
+void CPU::EnsureIAndDCacheCoherency(void *address, size_t length) {
+#ifdef USE_SIMULATOR
+  USE(address);
+  USE(length);
+  // TODO: consider adding cache simulation to ensure every address run has been
+  // synchronised.
+#else
+  // The code below assumes user space cache operations are allowed.
+
+  uintptr_t start = reinterpret_cast<uintptr_t>(address);
+  // Sizes will be used to generate a mask big enough to cover a pointer.
+  uintptr_t dsize = static_cast<uintptr_t>(dcache_line_size_);
+  uintptr_t isize = static_cast<uintptr_t>(icache_line_size_);
+  // Cache line sizes are always a power of 2.
+  ASSERT(CountSetBits(dsize, 64) == 1);
+  ASSERT(CountSetBits(isize, 64) == 1);
+  uintptr_t dstart = start & ~(dsize - 1);
+  uintptr_t istart = start & ~(isize - 1);
+  uintptr_t end = start + length;
+
+  __asm__ __volatile__ (  // NOLINT
+    // Clean every line of the D cache containing the target data.
+    "0:                                \n\t"
+    // dc      : Data Cache maintenance
+    //    c    : Clean
+    //     va  : by (Virtual) Address
+    //       u : to the point of Unification
+    // The point of unification for a processor is the point by which the
+    // instruction and data caches are guaranteed to see the same copy of a
+    // memory location. See ARM DDI 0406B page B2-12 for more information.
+    "dc   cvau, %[dline]                \n\t"
+    "add  %[dline], %[dline], %[dsize]  \n\t"
+    "cmp  %[dline], %[end]              \n\t"
+    "b.lt 0b                            \n\t"
+    // Barrier to make sure the effect of the code above is visible to the rest
+    // of the world.
+    // dsb    : Data Synchronisation Barrier
+    //    ish : Inner SHareable domain
+    // The point of unification for an Inner Shareable shareability domain is
+    // the point by which the instruction and data caches of all the processors
+    // in that Inner Shareable shareability domain are guaranteed to see the
+    // same copy of a memory location.  See ARM DDI 0406B page B2-12 for more
+    // information.
+    "dsb  ish                           \n\t"
+    // Invalidate every line of the I cache containing the target data.
+    "1:                                 \n\t"
+    // ic      : instruction cache maintenance
+    //    i    : invalidate
+    //     va  : by address
+    //       u : to the point of unification
+    "ic   ivau, %[iline]                \n\t"
+    "add  %[iline], %[iline], %[isize]  \n\t"
+    "cmp  %[iline], %[end]              \n\t"
+    "b.lt 1b                            \n\t"
+    // Barrier to make sure the effect of the code above is visible to the rest
+    // of the world.
+    "dsb  ish                           \n\t"
+    // Barrier to ensure any prefetching which happened before this code is
+    // discarded.
+    // isb : Instruction Synchronisation Barrier
+    "isb                                \n\t"
+    : [dline] "+r" (dstart),
+      [iline] "+r" (istart)
+    : [dsize] "r"  (dsize),
+      [isize] "r"  (isize),
+      [end]   "r"  (end)
+    // This code does not write to memory but without the dependency gcc might
+    // move this code before the code is generated.
+    : "cc", "memory"
+  );  // NOLINT
+#endif
+}
+
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/cpu-a64.h b/disas/libvixl/src/a64/cpu-a64.h
new file mode 100644
index 0000000..dfd8f01
--- /dev/null
+++ b/disas/libvixl/src/a64/cpu-a64.h
@@ -0,0 +1,56 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_CPU_A64_H
+#define VIXL_CPU_A64_H
+
+#include "globals.h"
+
+namespace vixl {
+
+class CPU {
+ public:
+  // Initialise CPU support.
+  static void SetUp();
+
+  // Ensures the data at a given address and with a given size is the same for
+  // the I and D caches. I and D caches are not automatically coherent on ARM
+  // so this operation is required before any dynamically generated code can
+  // safely run.
+  static void EnsureIAndDCacheCoherency(void *address, size_t length);
+
+ private:
+  // Return the content of the cache type register.
+  static uint32_t GetCacheType();
+
+  // I and D cache line size in bytes.
+  static unsigned icache_line_size_;
+  static unsigned dcache_line_size_;
+};
+
+}  // namespace vixl
+
+#endif  // VIXL_CPU_A64_H
diff --git a/disas/libvixl/src/a64/debugger-a64.cc b/disas/libvixl/src/a64/debugger-a64.cc
new file mode 100644
index 0000000..ecdc835
--- /dev/null
+++ b/disas/libvixl/src/a64/debugger-a64.cc
@@ -0,0 +1,1511 @@ 
+// Copyright 2013 ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY ARM LIMITED AND CONTRIBUTORS "AS IS" AND ANY
+// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL ARM LIMITED BE LIABLE FOR ANY DIRECT, INDIRECT,
+// INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+// OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
+// EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "a64/debugger-a64.h"
+
+namespace vixl {
+
+// List of commands supported by the debugger.
+#define DEBUG_COMMAND_LIST(C)  \
+C(HelpCommand)                 \
+C(ContinueCommand)             \
+C(StepCommand)                 \
+C(DisasmCommand)               \
+C(PrintCommand)                \
+C(MemCommand)
+
+// Debugger command lines are broken up in token of different type to make
+// processing easier later on.
+class Token {
+ public:
+  virtual ~Token() {}
+
+  // Token type.
+  virtual bool IsRegister() const { return false; }
+  virtual bool IsFPRegister() const { return false; }
+  virtual bool IsIdentifier() const { return false; }
+  virtual bool IsAddress() const { return false; }
+  virtual bool IsInteger() const { return false; }
+  virtual bool IsFormat() const { return false; }
+  virtual bool IsUnknown() const { return false; }
+  // Token properties.
+  virtual bool CanAddressMemory() const { return false; }
+  virtual uint8_t* ToAddress(Debugger* debugger) const;
+  virtual void Print(FILE* out = stdout) const = 0;
+
+  static Token* Tokenize(const char* arg);
+};
+
+// Tokens often hold one value.
+template<typename T> class ValueToken : public Token {
+ public:
+  explicit ValueToken(T value) : value_(value) {}
+  ValueToken() {}
+
+  T value() const { return value_; }
+
+ protected:
+  T value_;
+};
+
+// Integer registers (X or W) and their aliases.
+// Format: wn or xn with 0 <= n < 32 or a name in the aliases list.
+class RegisterToken : public ValueToken<const Register> {
+ public:
+  explicit RegisterToken(const Register reg)
+      : ValueToken<const Register>(reg) {}
+
+  virtual bool IsRegister() const { return true; }
+  virtual bool CanAddressMemory() const { return value().Is64Bits(); }
+  virtual uint8_t* ToAddress(Debugger* debugger) const;
+  virtual void Print(FILE* out = stdout) const ;
+  const char* Name() const;
+
+  static Token* Tokenize(const char* arg);
+  static RegisterToken* Cast(Token* tok) {
+    ASSERT(tok->IsRegister());
+    return reinterpret_cast<RegisterToken*>(tok);
+  }
+
+ private:
+  static const int kMaxAliasNumber = 4;
+  static const char* kXAliases[kNumberOfRegisters][kMaxAliasNumber];
+  static const char* kWAliases[kNumberOfRegisters][kMaxAliasNumber];
+};
+
+// Floating point registers (D or S).
+// Format: sn or dn with 0 <= n < 32.
+class FPRegisterToken : public ValueToken<const FPRegister> {
+ public:
+  explicit FPRegisterToken(const FPRegister fpreg)
+      : ValueToken<const FPRegister>(fpreg) {}
+
+  virtual bool IsFPRegister() const { return true; }
+  virtual void Print(FILE* out = stdout) const ;
+
+  static Token* Tokenize(const char* arg);
+  static FPRegisterToken* Cast(Token* tok) {
+    ASSERT(tok->IsFPRegister());
+    return reinterpret_cast<FPRegisterToken*>(tok);
+  }
+};
+
+
+// Non-register identifiers.
+// Format: Alphanumeric string starting with a letter.
+class IdentifierToken : public ValueToken<char*> {
+ public:
+  explicit IdentifierToken(const char* name) {
+    int size = strlen(name) + 1;
+    value_ = new char[size];
+    strncpy(value_, name, size);
+  }
+  virtual ~IdentifierToken() { delete[] value_; }
+
+  virtual bool IsIdentifier() const { return true; }
+  virtual bool CanAddressMemory() const { return strcmp(value(), "pc") == 0; }
+  virtual uint8_t* ToAddress(Debugger* debugger) const;
+  virtual void Print(FILE* out = stdout) const;
+
+  static Token* Tokenize(const char* arg);
+  static IdentifierToken* Cast(Token* tok) {
+    ASSERT(tok->IsIdentifier());
+    return reinterpret_cast<IdentifierToken*>(tok);
+  }
+};
+
+// 64-bit address literal.
+// Format: 0x... with up to 16 hexadecimal digits.
+class AddressToken : public ValueToken<uint8_t*> {
+ public:
+  explicit AddressToken(uint8_t* address) : ValueToken<uint8_t*>(address) {}
+
+  virtual bool IsAddress() const { return true; }
+  virtual bool CanAddressMemory() const { return true; }
+  virtual uint8_t* ToAddress(Debugger* debugger) const;
+  virtual void Print(FILE* out = stdout) const ;
+
+  static Token* Tokenize(const char* arg);
+  static AddressToken* Cast(Token* tok) {
+    ASSERT(tok->IsAddress());
+    return reinterpret_cast<AddressToken*>(tok);
+  }
+};
+
+
+// 64-bit decimal integer literal.
+// Format: n.
+class IntegerToken : public ValueToken<int64_t> {
+ public:
+  explicit IntegerToken(int value) : ValueToken<int64_t>(value) {}
+
+  virtual bool IsInteger() const { return true; }
+  virtual void Print(FILE* out = stdout) const;
+
+  static Token* Tokenize(const char* arg);
+  static IntegerToken* Cast(Token* tok) {
+    ASSERT(tok->IsInteger());
+    return reinterpret_cast<IntegerToken*>(tok);
+  }
+};
+
+// Literal describing how to print a chunk of data (up to 64 bits).
+// Format: %qt
+// where q (qualifier) is one of
+//  * s: signed integer
+//  * u: unsigned integer
+//  * a: hexadecimal floating point
+// and t (type) is one of
+//  * x: 64-bit integer
+//  * w: 32-bit integer
+//  * h: 16-bit integer
+//  * b: 8-bit integer
+//  * c: character
+//  * d: double
+//  * s: float
+// When no qualifier is given for integers, they are printed in hexadecinal.
+class FormatToken : public Token {
+ public:
+  FormatToken() {}
+
+  virtual bool IsFormat() const { return true; }
+  virtual int SizeOf() const = 0;
+  virtual void PrintData(void* data, FILE* out = stdout) const = 0;
+  virtual void Print(FILE* out = stdout) const = 0;
+
+  static Token* Tokenize(const char* arg);
+  static FormatToken* Cast(Token* tok) {
+    ASSERT(tok->IsFormat());
+    return reinterpret_cast<FormatToken*>(tok);
+  }
+};
+
+
+template<typename T> class Format : public FormatToken {
+ public:
+  explicit Format(const char* fmt) : fmt_(fmt) {}
+
+  virtual int SizeOf() const { return sizeof(T); }
+  virtual void PrintData(void* data, FILE* out = stdout) const {
+    T value;
+    memcpy(&value, data, sizeof(value));
+    fprintf(out, fmt_, value);
+  }
+  virtual void Print(FILE* out = stdout) const;
+
+ private:
+  const char* fmt_;
+};
+
+// Tokens which don't fit any of the above.
+class UnknownToken : public Token {
+ public:
+  explicit UnknownToken(const char* arg) {
+    int size = strlen(arg) + 1;
+    unknown_ = new char[size];
+    strncpy(unknown_, arg, size);
+  }
+  virtual ~UnknownToken() { delete[] unknown_; }
+
+  virtual bool IsUnknown() const { return true; }
+  virtual void Print(FILE* out = stdout) const;
+
+ private:
+  char* unknown_;
+};
+
+
+// All debugger commands must subclass DebugCommand and implement Run, Print
+// and Build. Commands must also define kHelp and kAliases.
+class DebugCommand {
+ public:
+  explicit DebugCommand(Token* name) : name_(IdentifierToken::Cast(name)) {}
+  DebugCommand() : name_(NULL) {}
+  virtual ~DebugCommand() { delete name_; }
+
+  const char* name() { return name_->value(); }
+  // Run the command on the given debugger. The command returns true if
+  // execution should move to the next instruction.
+  virtual bool Run(Debugger * debugger) = 0;
+  virtual void Print(FILE* out = stdout);
+
+  static bool Match(const char* name, const char** aliases);
+  static DebugCommand* Parse(char* line);
+  static void PrintHelp(const char** aliases,
+                        const char* args,
+                        const char* help);
+
+ private:
+  IdentifierToken* name_;
+};
+
+// For all commands below see their respective kHelp and kAliases in
+// debugger-a64.cc
+class HelpCommand : public DebugCommand {
+ public:
+  explicit HelpCommand(Token* name) : DebugCommand(name) {}
+
+  virtual bool Run(Debugger* debugger);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+};
+
+
+class ContinueCommand : public DebugCommand {
+ public:
+  explicit ContinueCommand(Token* name) : DebugCommand(name) {}
+
+  virtual bool Run(Debugger* debugger);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+};
+
+
+class StepCommand : public DebugCommand {
+ public:
+  StepCommand(Token* name, IntegerToken* count)
+      : DebugCommand(name), count_(count) {}
+  virtual ~StepCommand() { delete count_; }
+
+  int64_t count() { return count_->value(); }
+  virtual bool Run(Debugger* debugger);
+  virtual void Print(FILE* out = stdout);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+
+ private:
+  IntegerToken* count_;
+};
+
+class DisasmCommand : public DebugCommand {
+ public:
+  DisasmCommand(Token* name, Token* target, IntegerToken* count)
+      : DebugCommand(name), target_(target), count_(count) {}
+  virtual ~DisasmCommand() {
+    delete target_;
+    delete count_;
+  }
+
+  Token* target() { return target_; }
+  int64_t count() { return count_->value(); }
+  virtual bool Run(Debugger* debugger);
+  virtual void Print(FILE* out = stdout);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+
+ private:
+  Token* target_;
+  IntegerToken* count_;
+};
+
+
+class PrintCommand : public DebugCommand {
+ public:
+  PrintCommand(Token* name, Token* target)
+      : DebugCommand(name), target_(target) {}
+  virtual ~PrintCommand() { delete target_; }
+
+  Token* target() { return target_; }
+  virtual bool Run(Debugger* debugger);
+  virtual void Print(FILE* out = stdout);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+
+ private:
+  Token* target_;
+};
+
+class MemCommand : public DebugCommand {
+ public:
+  MemCommand(Token* name,
+             Token* target,
+             IntegerToken* count,
+             FormatToken* format)
+      : DebugCommand(name), target_(target), count_(count), format_(format) {}
+  virtual ~MemCommand() {
+    delete target_;
+    delete count_;
+    delete format_;
+  }
+
+  Token* target() { return target_; }
+  int64_t count() { return count_->value(); }
+  FormatToken* format() { return format_; }
+  virtual bool Run(Debugger* debugger);
+  virtual void Print(FILE* out = stdout);
+
+  static DebugCommand* Build(std::vector<Token*> args);
+
+  static const char* kHelp;
+  static const char* kAliases[];
+  static const char* kArguments;
+
+ private:
+  Token* target_;
+  IntegerToken* count_;
+  FormatToken* format_;
+};
+
+// Commands which name does not match any of the known commnand.
+class UnknownCommand : public DebugCommand {
+ public:
+  explicit UnknownCommand(std::vector<Token*> args) : args_(args) {}
+  virtual ~UnknownCommand();
+
+  virtual bool Run(Debugger* debugger);
+
+ private:
+  std::vector<Token*> args_;
+};
+
+// Commands which name match a known command but the syntax is invalid.
+class InvalidCommand : public DebugCommand {
+ public:
+  InvalidCommand(std::vector<Token*> args, int index, const char* cause)
+      : args_(args), index_(index), cause_(cause) {}
+  virtual ~InvalidCommand();
+
+  virtual bool Run(Debugger* debugger);
+
+ private:
+  std::vector<Token*> args_;
+  int index_;
+  const char* cause_;
+};
+
+const char* HelpCommand::kAliases[] = { "help", NULL };
+const char* HelpCommand::kArguments = NULL;
+const char* HelpCommand::kHelp = "  print this help";
+
+const char* ContinueCommand::kAliases[] = { "continue", "c", NULL };
+const char* ContinueCommand::kArguments = NULL;
+const char* ContinueCommand::kHelp = "  resume execution";
+
+const char* StepCommand::kAliases[] = { "stepi", "si", NULL };
+const char* StepCommand::kArguments = "[n = 1]";
+const char* StepCommand::kHelp = "  execute n next instruction(s)";
+
+const char* DisasmCommand::kAliases[] = { "dis", "d", NULL };
+const char* DisasmCommand::kArguments = "[addr = pc] [n = 1]";
+const char* DisasmCommand::kHelp =
+  "  disassemble n instruction(s) at address addr.\n"
+  "  addr can be an immediate address, a register or the pc."
+;
+
+const char* PrintCommand::kAliases[] = { "print", "p", NULL };
+const char* PrintCommand::kArguments =  "<entity>";
+const char* PrintCommand::kHelp =
+  "  print the given entity\n"
+  "  entity can be 'regs' for W and X registers, 'fpregs' for S and D\n"
+  "  registers, 'sysregs' for system registers (including NZCV) or 'pc'."
+;
+
+const char* MemCommand::kAliases[] = { "mem", "m", NULL };
+const char* MemCommand::kArguments = "<addr> [n = 1] [format = %x]";
+const char* MemCommand::kHelp =
+  "  print n memory item(s) at address addr according to the given format.\n"
+  "  addr can be an immediate address, a register or the pc.\n"
+  "  format is made of a qualifer: 's', 'u', 'a' (signed, unsigned, hexa)\n"
+  "  and a type 'x', 'w', 'h', 'b' (64- to 8-bit integer), 'c' (character),\n"
+  "  's' (float) or 'd' (double). E.g 'mem sp %w' will print a 32-bit word\n"
+  "  from the stack as an hexadecimal number."
+;
+
+const char* RegisterToken::kXAliases[kNumberOfRegisters][kMaxAliasNumber] = {
+  { "x0", NULL },
+  { "x1", NULL },
+  { "x2", NULL },
+  { "x3", NULL },
+  { "x4", NULL },
+  { "x5", NULL },
+  { "x6", NULL },
+  { "x7", NULL },
+  { "x8", NULL },
+  { "x9", NULL },
+  { "x10", NULL },
+  { "x11", NULL },
+  { "x12", NULL },
+  { "x13", NULL },
+  { "x14", NULL },
+  { "x15", NULL },
+  { "ip0", "x16", NULL },
+  { "ip1", "x17", NULL },
+  { "x18", "pr", NULL },
+  { "x19", NULL },
+  { "x20", NULL },
+  { "x21", NULL },
+  { "x22", NULL },
+  { "x23", NULL },
+  { "x24", NULL },
+  { "x25", NULL },
+  { "x26", NULL },
+  { "x27", NULL },
+  { "x28", NULL },
+  { "fp", "x29", NULL },
+  { "lr", "x30", NULL },
+  { "sp", NULL}
+};
+
+const char* RegisterToken::kWAliases[kNumberOfRegisters][kMaxAliasNumber] = {
+  { "w0", NULL },
+  { "w1", NULL },
+  { "w2", NULL },
+  { "w3", NULL },
+  { "w4", NULL },
+  { "w5", NULL },
+  { "w6", NULL },
+  { "w7", NULL },
+  { "w8", NULL },
+  { "w9", NULL },
+  { "w10", NULL },
+  { "w11", NULL },
+  { "w12", NULL },
+  { "w13", NULL },
+  { "w14", NULL },
+  { "w15", NULL },
+  { "w16", NULL },
+  { "w17", NULL },
+  { "w18", NULL },
+  { "w19", NULL },
+  { "w20", NULL },
+  { "w21", NULL },
+  { "w22", NULL },
+  { "w23", NULL },
+  { "w24", NULL },
+  { "w25", NULL },
+  { "w26", NULL },
+  { "w27", NULL },
+  { "w28", NULL },
+  { "w29", NULL },
+  { "w30", NULL },
+  { "wsp", NULL }
+};
+
+
+Debugger::Debugger(Decoder* decoder, FILE* stream)
+    : Simulator(decoder, stream),
+      log_parameters_(0),
+      debug_parameters_(0),
+      pending_request_(false),
+      steps_(0),
+      last_command_(NULL) {
+  disasm_ = new PrintDisassembler(stdout);
+  printer_ = new Decoder();
+  printer_->AppendVisitor(disasm_);
+}
+
+
+void Debugger::Run() {
+  while (pc_ != kEndOfSimAddress) {
+    if (pending_request()) {
+      LogProcessorState();
+      RunDebuggerShell();
+    }
+
+    ExecuteInstruction();
+  }
+}
+
+
+void Debugger::PrintInstructions(void* address, int64_t count) {
+  if (count == 0) {
+    return;
+  }
+
+  Instruction* from = Instruction::Cast(address);
+  if (count < 0) {
+    count = -count;
+    from -= (count - 1) * kInstructionSize;
+  }
+  Instruction* to = from + count * kInstructionSize;
+
+  for (Instruction* current = from;
+       current < to;
+       current = current->NextInstruction()) {
+    printer_->Decode(current);
+  }
+}
+
+
+void Debugger::PrintMemory(const uint8_t* address,
+                           int64_t count,
+                           const FormatToken* format) {
+  if (count == 0) {
+    return;
+  }
+
+  const uint8_t* from = address;
+  int size = format->SizeOf();
+  if (count < 0) {
+    count = -count;
+    from -= (count - 1) * size;
+  }
+  const uint8_t* to = from + count * size;
+
+  for (const uint8_t* current = from; current < to; current += size) {
+    if (((current - from) % 16) == 0) {
+      printf("\n%p: ", current);
+    }
+
+    uint64_t data = MemoryRead(current, size);
+    format->PrintData(&data);
+    printf(" ");
+  }
+  printf("\n\n");
+}
+
+
+void Debugger::VisitException(Instruction* instr) {
+  switch (instr->Mask(ExceptionMask)) {
+    case BRK:
+      DoBreakpoint(instr);
+      return;
+    case HLT:
+      switch (instr->ImmException()) {
+        case kUnreachableOpcode:
+          DoUnreachable(instr);
+          return;
+        case kTraceOpcode:
+          DoTrace(instr);
+          return;
+        case kLogOpcode:
+          DoLog(instr);
+          return;
+      }
+      // Fall through
+    default: Simulator::VisitException(instr);
+  }
+}
+
+
+void Debugger::LogSystemRegisters() {
+  if (log_parameters_ & LOG_SYS_REGS) PrintSystemRegisters();
+}
+
+
+void Debugger::LogRegisters() {
+  if (log_parameters_ & LOG_REGS) PrintRegisters();
+}
+
+
+void Debugger::LogFPRegisters() {
+  if (log_parameters_ & LOG_FP_REGS) PrintFPRegisters();
+}
+
+
+void Debugger::LogProcessorState() {
+  LogSystemRegisters();
+  LogRegisters();
+  LogFPRegisters();
+}
+
+
+// Read a command. A command will be at most kMaxDebugShellLine char long and
+// ends with '\n\0'.
+// TODO: Should this be a utility function?
+char* Debugger::ReadCommandLine(const char* prompt, char* buffer, int length) {
+  int fgets_calls = 0;
+  char* end = NULL;
+
+  printf("%s", prompt);
+  fflush(stdout);
+
+  do {
+    if (fgets(buffer, length, stdin) == NULL) {
+      printf(" ** Error while reading command. **\n");
+      return NULL;
+    }
+
+    fgets_calls++;
+    end = strchr(buffer, '\n');
+  } while (end == NULL);
+
+  if (fgets_calls != 1) {
+    printf(" ** Command too long. **\n");
+    return NULL;
+  }
+
+  // Remove the newline from the end of the command.
+  ASSERT(end[1] == '\0');
+  ASSERT((end - buffer) < (length - 1));
+  end[0] = '\0';
+
+  return buffer;
+}
+
+
+void Debugger::RunDebuggerShell() {
+  if (IsDebuggerRunning()) {
+    if (steps_ > 0) {
+      // Finish stepping first.
+      --steps_;
+      return;
+    }
+
+    printf("Next: ");
+    PrintInstructions(pc());
+    bool done = false;
+    while (!done) {
+      char buffer[kMaxDebugShellLine];
+      char* line = ReadCommandLine("vixl> ", buffer, kMaxDebugShellLine);
+
+      if (line == NULL) continue;  // An error occurred.
+
+      DebugCommand* command = DebugCommand::Parse(line);
+      if (command != NULL) {
+        last_command_ = command;
+      }
+
+      if (last_command_ != NULL) {
+        done = last_command_->Run(this);
+      } else {
+        printf("No previous command to run!\n");
+      }
+    }
+
+    if ((debug_parameters_ & DBG_BREAK) != 0) {
+      // The break request has now been handled, move to next instruction.
+      debug_parameters_ &= ~DBG_BREAK;
+      increment_pc();
+    }
+  }
+}
+
+
+void Debugger::DoBreakpoint(Instruction* instr) {
+  ASSERT(instr->Mask(ExceptionMask) == BRK);
+
+  printf("Hit breakpoint at pc=%p.\n", reinterpret_cast<void*>(instr));
+  set_debug_parameters(debug_parameters() | DBG_BREAK | DBG_ACTIVE);
+  // Make the shell point to the brk instruction.
+  set_pc(instr);
+}
+
+
+void Debugger::DoUnreachable(Instruction* instr) {
+  ASSERT((instr->Mask(ExceptionMask) == HLT) &&
+         (instr->ImmException() == kUnreachableOpcode));
+
+  fprintf(stream_, "Hit UNREACHABLE marker at pc=%p.\n",
+          reinterpret_cast<void*>(instr));
+  abort();
+}
+
+
+void Debugger::DoTrace(Instruction* instr) {
+  ASSERT((instr->Mask(ExceptionMask) == HLT) &&
+         (instr->ImmException() == kTraceOpcode));
+
+  // Read the arguments encoded inline in the instruction stream.
+  uint32_t parameters;
+  uint32_t command;
+
+  ASSERT(sizeof(*instr) == 1);
+  memcpy(&parameters, instr + kTraceParamsOffset, sizeof(parameters));
+  memcpy(&command, instr + kTraceCommandOffset, sizeof(command));
+
+  switch (command) {
+    case TRACE_ENABLE:
+      set_log_parameters(log_parameters() | parameters);
+      break;
+    case TRACE_DISABLE:
+      set_log_parameters(log_parameters() & ~parameters);
+      break;
+    default:
+      UNREACHABLE();
+  }
+
+  set_pc(instr->InstructionAtOffset(kTraceLength));
+}
+
+
+void Debugger::DoLog(Instruction* instr) {
+  ASSERT((instr->Mask(ExceptionMask) == HLT) &&
+         (instr->ImmException() == kLogOpcode));
+
+  // Read the arguments encoded inline in the instruction stream.
+  uint32_t parameters;
+
+  ASSERT(sizeof(*instr) == 1);
+  memcpy(&parameters, instr + kTraceParamsOffset, sizeof(parameters));
+
+  // We don't support a one-shot LOG_DISASM.
+  ASSERT((parameters & LOG_DISASM) == 0);
+  // Print the requested information.
+  if (parameters & LOG_SYS_REGS) PrintSystemRegisters(true);
+  if (parameters & LOG_REGS) PrintRegisters(true);
+  if (parameters & LOG_FP_REGS) PrintFPRegisters(true);
+
+  set_pc(instr->InstructionAtOffset(kLogLength));
+}
+
+
+static bool StringToUInt64(uint64_t* value, const char* line, int base = 10) {
+  char* endptr = NULL;
+  errno = 0;  // Reset errors.
+  uint64_t parsed = strtoul(line, &endptr, base);
+
+  if (errno == ERANGE) {
+    // Overflow.
+    return false;
+  }
+
+  if (endptr == line) {
+    // No digits were parsed.
+    return false;
+  }
+
+  if (*endptr != '\0') {
+    // Non-digit characters present at the end.
+    return false;
+  }
+
+  *value = parsed;
+  return true;
+}
+
+
+static bool StringToInt64(int64_t* value, const char* line, int base = 10) {
+  char* endptr = NULL;
+  errno = 0;  // Reset errors.
+  int64_t parsed = strtol(line, &endptr, base);
+
+  if (errno == ERANGE) {
+    // Overflow, undeflow.
+    return false;
+  }
+
+  if (endptr == line) {
+    // No digits were parsed.
+    return false;
+  }
+
+  if (*endptr != '\0') {
+    // Non-digit characters present at the end.
+    return false;
+  }
+
+  *value = parsed;
+  return true;
+}
+
+
+uint8_t* Token::ToAddress(Debugger* debugger) const {
+  USE(debugger);
+  UNREACHABLE();
+  return NULL;
+}
+
+
+Token* Token::Tokenize(const char* arg) {
+  if ((arg == NULL) || (*arg == '\0')) {
+    return NULL;
+  }
+
+  // The order is important. For example Identifier::Tokenize would consider
+  // any register to be a valid identifier.
+
+  Token* token = RegisterToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  token = FPRegisterToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  token = IdentifierToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  token = AddressToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  token = IntegerToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  token = FormatToken::Tokenize(arg);
+  if (token != NULL) {
+    return token;
+  }
+
+  return new UnknownToken(arg);
+}
+
+
+uint8_t* RegisterToken::ToAddress(Debugger* debugger) const {
+  ASSERT(CanAddressMemory());
+  uint64_t reg_value = debugger->xreg(value().code(), Reg31IsStackPointer);
+  uint8_t* address = NULL;
+  memcpy(&address, &reg_value, sizeof(address));
+  return address;
+}
+
+
+void RegisterToken::Print(FILE* out) const {
+  ASSERT(value().IsValid());
+  fprintf(out, "[Register %s]", Name());
+}
+
+
+const char* RegisterToken::Name() const {
+  if (value().Is32Bits()) {
+    return kWAliases[value().code()][0];
+  } else {
+    return kXAliases[value().code()][0];
+  }
+}
+
+
+Token* RegisterToken::Tokenize(const char* arg) {
+  for (unsigned i = 0; i < kNumberOfRegisters; i++) {
+    // Is it a X register or alias?
+    for (const char** current = kXAliases[i]; *current != NULL; current++) {
+      if (strcmp(arg, *current) == 0) {
+        return new RegisterToken(Register::XRegFromCode(i));
+      }
+    }
+
+    // Is it a W register or alias?
+    for (const char** current = kWAliases[i]; *current != NULL; current++) {
+      if (strcmp(arg, *current) == 0) {
+        return new RegisterToken(Register::WRegFromCode(i));
+      }
+    }
+  }
+
+  return NULL;
+}
+
+
+void FPRegisterToken::Print(FILE* out) const {
+  ASSERT(value().IsValid());
+  char prefix = value().Is32Bits() ? 's' : 'd';
+  fprintf(out, "[FPRegister %c%" PRIu32 "]", prefix, value().code());
+}
+
+
+Token* FPRegisterToken::Tokenize(const char* arg) {
+  if (strlen(arg) < 2) {
+    return NULL;
+  }
+
+  switch (*arg) {
+    case 's':
+    case 'd':
+      const char* cursor = arg + 1;
+      uint64_t code = 0;
+      if (!StringToUInt64(&code, cursor)) {
+        return NULL;
+      }
+
+      if (code > kNumberOfFPRegisters) {
+        return NULL;
+      }
+
+      FPRegister fpreg = NoFPReg;
+      switch (*arg) {
+        case 's': fpreg = FPRegister::SRegFromCode(code); break;
+        case 'd': fpreg = FPRegister::DRegFromCode(code); break;
+        default: UNREACHABLE();
+      }
+
+      return new FPRegisterToken(fpreg);
+  }
+
+  return NULL;
+}
+
+
+uint8_t* IdentifierToken::ToAddress(Debugger* debugger) const {
+  ASSERT(CanAddressMemory());
+  Instruction* pc_value = debugger->pc();
+  uint8_t* address = NULL;
+  memcpy(&address, &pc_value, sizeof(address));
+  return address;
+}
+
+void IdentifierToken::Print(FILE* out) const {
+  fprintf(out, "[Identifier %s]", value());
+}
+
+
+Token* IdentifierToken::Tokenize(const char* arg) {
+  if (!isalpha(arg[0])) {
+    return NULL;
+  }
+
+  const char* cursor = arg + 1;
+  while ((*cursor != '\0') && isalnum(*cursor)) {
+    ++cursor;
+  }
+
+  if (*cursor == '\0') {
+    return new IdentifierToken(arg);
+  }
+
+  return NULL;
+}
+
+
+uint8_t* AddressToken::ToAddress(Debugger* debugger) const {
+  USE(debugger);
+  return value();
+}
+
+
+void AddressToken::Print(FILE* out) const {
+  fprintf(out, "[Address %p]", value());
+}
+
+
+Token* AddressToken::Tokenize(const char* arg) {
+  if ((strlen(arg) < 3) || (arg[0] != '0') || (arg[1] != 'x')) {
+    return NULL;
+  }
+
+  uint64_t ptr = 0;
+  if (!StringToUInt64(&ptr, arg, 16)) {
+    return NULL;
+  }
+
+  uint8_t* address = reinterpret_cast<uint8_t*>(ptr);
+  return new AddressToken(address);
+}
+
+
+void IntegerToken::Print(FILE* out) const {
+  fprintf(out, "[Integer %" PRId64 "]", value());
+}
+
+
+Token* IntegerToken::Tokenize(const char* arg) {
+  int64_t value = 0;
+  if (!StringToInt64(&value, arg)) {
+    return NULL;
+  }
+
+  return new IntegerToken(value);
+}
+
+
+Token* FormatToken::Tokenize(const char* arg) {
+  if (arg[0] != '%') {
+    return NULL;
+  }
+
+  int length = strlen(arg);
+  if ((length < 2) || (length > 3)) {
+    return NULL;
+  }
+
+  char type = arg[length - 1];
+  if (length == 2) {
+    switch (type) {
+      case 'x': return new Format<uint64_t>("%016" PRIx64);
+      case 'w': return new Format<uint32_t>("%08" PRIx32);
+      case 'h': return new Format<uint16_t>("%04" PRIx16);
+      case 'b': return new Format<uint8_t>("%02" PRIx8);
+      case 'c': return new Format<char>("%c");
+      case 'd': return new Format<double>("%g");
+      case 's': return new Format<float>("%g");
+      default: return NULL;
+    }
+  }
+
+  ASSERT(length == 3);
+  switch (arg[1]) {
+    case 's':
+      switch (type) {
+        case 'x': return new Format<int64_t>("%+20" PRId64);
+        case 'w': return new Format<int32_t>("%+11" PRId32);
+        case 'h': return new Format<int16_t>("%+6" PRId16);
+        case 'b': return new Format<int8_t>("%+4" PRId8);
+        default: return NULL;
+      }
+    case 'u':
+      switch (type) {
+        case 'x': return new Format<uint64_t>("%20" PRIu64);
+        case 'w': return new Format<uint32_t>("%10" PRIu32);
+        case 'h': return new Format<uint16_t>("%5" PRIu16);
+        case 'b': return new Format<uint8_t>("%3" PRIu8);
+        default: return NULL;
+      }
+    case 'a':
+      switch (type) {
+        case 'd': return new Format<double>("%a");
+        case 's': return new Format<float>("%a");
+        default: return NULL;
+      }
+    default: return NULL;
+  }
+}
+
+
+template<typename T>
+void Format<T>::Print(FILE* out) const {
+  fprintf(out, "[Format %s - %lu byte(s)]", fmt_, sizeof(T));
+}
+
+
+void UnknownToken::Print(FILE* out) const {
+  fprintf(out, "[Unknown %s]", unknown_);
+}
+
+
+void DebugCommand::Print(FILE* out) {
+  fprintf(out, "%s", name());
+}
+
+
+bool DebugCommand::Match(const char* name, const char** aliases) {
+  for (const char** current = aliases; *current != NULL; current++) {
+    if (strcmp(name, *current) == 0) {
+       return true;
+    }
+  }
+
+  return false;
+}
+
+
+DebugCommand* DebugCommand::Parse(char* line) {
+  std::vector<Token*> args;
+
+  for (char* chunk = strtok(line, " ");
+       chunk != NULL;
+       chunk = strtok(NULL, " ")) {
+    args.push_back(Token::Tokenize(chunk));
+  }
+
+  if (args.size() == 0) {
+    return NULL;
+  }
+
+  if (!args[0]->IsIdentifier()) {
+    return new InvalidCommand(args, 0, "command name is not an identifier");
+  }
+
+  const char* name = IdentifierToken::Cast(args[0])->value();
+  #define RETURN_IF_MATCH(Command)       \
+  if (Match(name, Command::kAliases)) {  \
+    return Command::Build(args);         \
+  }
+  DEBUG_COMMAND_LIST(RETURN_IF_MATCH);
+  #undef RETURN_IF_MATCH
+
+  return new UnknownCommand(args);
+}
+
+
+void DebugCommand::PrintHelp(const char** aliases,
+                             const char* args,
+                             const char* help) {
+  ASSERT(aliases[0] != NULL);
+  ASSERT(help != NULL);
+
+  printf("\n----\n\n");
+  for (const char** current = aliases; *current != NULL; current++) {
+    if (args != NULL) {
+      printf("%s %s\n", *current, args);
+    } else {
+      printf("%s\n", *current);
+    }
+  }
+  printf("\n%s\n", help);
+}
+
+
+bool HelpCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+  USE(debugger);
+
+  #define PRINT_HELP(Command)                     \
+    DebugCommand::PrintHelp(Command::kAliases,    \
+                            Command::kArguments,  \
+                            Command::kHelp);
+  DEBUG_COMMAND_LIST(PRINT_HELP);
+  #undef PRINT_HELP
+  printf("\n----\n\n");
+
+  return false;
+}
+
+
+DebugCommand* HelpCommand::Build(std::vector<Token*> args) {
+  if (args.size() != 1) {
+    return new InvalidCommand(args, -1, "too many arguments");
+  }
+
+  return new HelpCommand(args[0]);
+}
+
+
+bool ContinueCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+
+  debugger->set_debug_parameters(debugger->debug_parameters() & ~DBG_ACTIVE);
+  return true;
+}
+
+
+DebugCommand* ContinueCommand::Build(std::vector<Token*> args) {
+  if (args.size() != 1) {
+    return new InvalidCommand(args, -1, "too many arguments");
+  }
+
+  return new ContinueCommand(args[0]);
+}
+
+
+bool StepCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+
+  int64_t steps = count();
+  if (steps < 0) {
+    printf(" ** invalid value for steps: %" PRId64 " (<0) **\n", steps);
+  } else if (steps > 1) {
+    debugger->set_steps(steps - 1);
+  }
+
+  return true;
+}
+
+
+void StepCommand::Print(FILE* out) {
+  fprintf(out, "%s %" PRId64 "", name(), count());
+}
+
+
+DebugCommand* StepCommand::Build(std::vector<Token*> args) {
+  IntegerToken* count = NULL;
+  switch (args.size()) {
+    case 1: {  // step [1]
+      count = new IntegerToken(1);
+      break;
+    }
+    case 2: {  // step n
+      Token* first = args[1];
+      if (!first->IsInteger()) {
+        return new InvalidCommand(args, 1, "expects int");
+      }
+      count = IntegerToken::Cast(first);
+      break;
+    }
+    default:
+      return new InvalidCommand(args, -1, "too many arguments");
+  }
+
+  return new StepCommand(args[0], count);
+}
+
+
+bool DisasmCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+
+  uint8_t* from = target()->ToAddress(debugger);
+  debugger->PrintInstructions(from, count());
+
+  return false;
+}
+
+
+void DisasmCommand::Print(FILE* out) {
+  fprintf(out, "%s ", name());
+  target()->Print(out);
+  fprintf(out, " %" PRId64 "", count());
+}
+
+
+DebugCommand* DisasmCommand::Build(std::vector<Token*> args) {
+  Token* address = NULL;
+  IntegerToken* count = NULL;
+  switch (args.size()) {
+    case 1: {  // disasm [pc] [1]
+      address = new IdentifierToken("pc");
+      count = new IntegerToken(1);
+      break;
+    }
+    case 2: {  // disasm [pc] n or disasm address [1]
+      Token* first = args[1];
+      if (first->IsInteger()) {
+        address = new IdentifierToken("pc");
+        count = IntegerToken::Cast(first);
+      } else if (first->CanAddressMemory()) {
+        address = first;
+        count = new IntegerToken(1);
+      } else {
+        return new InvalidCommand(args, 1, "expects int or addr");
+      }
+      break;
+    }
+    case 3: {  // disasm address count
+      Token* first = args[1];
+      Token* second = args[2];
+      if (!first->CanAddressMemory() || !second->IsInteger()) {
+        return new InvalidCommand(args, -1, "disasm addr int");
+      }
+      address = first;
+      count = IntegerToken::Cast(second);
+      break;
+    }
+    default:
+      return new InvalidCommand(args, -1, "wrong arguments number");
+  }
+
+  return new DisasmCommand(args[0], address, count);
+}
+
+
+void PrintCommand::Print(FILE* out) {
+  fprintf(out, "%s ", name());
+  target()->Print(out);
+}
+
+
+bool PrintCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+
+  Token* tok = target();
+  if (tok->IsIdentifier()) {
+    char* identifier = IdentifierToken::Cast(tok)->value();
+    if (strcmp(identifier, "regs") == 0) {
+      debugger->PrintRegisters(true);
+    } else if (strcmp(identifier, "fpregs") == 0) {
+      debugger->PrintFPRegisters(true);
+    } else if (strcmp(identifier, "sysregs") == 0) {
+      debugger->PrintSystemRegisters(true);
+    } else if (strcmp(identifier, "pc") == 0) {
+      printf("pc = %16p\n", reinterpret_cast<void*>(debugger->pc()));
+    } else {
+      printf(" ** Unknown identifier to print: %s **\n", identifier);
+    }
+
+    return false;
+  }
+
+  if (tok->IsRegister()) {
+    RegisterToken* reg_tok = RegisterToken::Cast(tok);
+    Register reg = reg_tok->value();
+    if (reg.Is32Bits()) {
+      printf("%s = %" PRId32 "\n",
+             reg_tok->Name(),
+             debugger->wreg(reg.code(), Reg31IsStackPointer));
+    } else {
+      printf("%s = %" PRId64 "\n",
+             reg_tok->Name(),
+             debugger->xreg(reg.code(), Reg31IsStackPointer));
+    }
+
+    return false;
+  }
+
+  if (tok->IsFPRegister()) {
+    FPRegister fpreg = FPRegisterToken::Cast(tok)->value();
+    if (fpreg.Is32Bits()) {
+      printf("s%u = %g\n", fpreg.code(), debugger->sreg(fpreg.code()));
+    } else {
+      printf("d%u = %g\n", fpreg.code(), debugger->dreg(fpreg.code()));
+    }
+
+    return false;
+  }
+
+  UNREACHABLE();
+  return false;
+}
+
+
+DebugCommand* PrintCommand::Build(std::vector<Token*> args) {
+  Token* target = NULL;
+  switch (args.size()) {
+    case 2: {
+      target = args[1];
+      if (!target->IsRegister()
+          && !target->IsFPRegister()
+          && !target->IsIdentifier()) {
+        return new InvalidCommand(args, 1, "expects reg or identifier");
+      }
+      break;
+    }
+    default:
+      return new InvalidCommand(args, -1, "too many arguments");
+  }
+
+  return new PrintCommand(args[0], target);
+}
+
+
+bool MemCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+
+  uint8_t* address = target()->ToAddress(debugger);
+  debugger->PrintMemory(address, count(), format());
+
+  return false;
+}
+
+
+void MemCommand::Print(FILE* out) {
+  fprintf(out, "%s ", name());
+  target()->Print(out);
+  fprintf(out, " %" PRId64 " ", count());
+  format()->Print(out);
+}
+
+
+DebugCommand* MemCommand::Build(std::vector<Token*> args) {
+  if (args.size() < 2) {
+    return new InvalidCommand(args, -1, "too few arguments");
+  }
+
+  Token* target = args[1];
+  IntegerToken* count = NULL;
+  FormatToken* format = NULL;
+
+  if (!target->CanAddressMemory()) {
+    return new InvalidCommand(args, 1, "expects address");
+  }
+
+  switch (args.size()) {
+    case 2: {  // mem addressable [1] [%x]
+      count = new IntegerToken(1);
+      format = new Format<uint64_t>("%016x");
+      break;
+    }
+    case 3: {  // mem addr n [%x] or mem addr [n] %f
+      Token* second = args[2];
+      if (second->IsInteger()) {
+        count = IntegerToken::Cast(second);
+        format = new Format<uint64_t>("%016x");
+      } else if (second->IsFormat()) {
+        count = new IntegerToken(1);
+        format = FormatToken::Cast(second);
+      } else {
+        return new InvalidCommand(args, 2, "expects int or format");
+      }
+      break;
+    }
+    case 4: {  // mem addr n %f
+      Token* second = args[2];
+      Token* third = args[3];
+      if (!second->IsInteger() || !third->IsFormat()) {
+        return new InvalidCommand(args, -1, "mem addr >>int<< %F");
+      }
+
+      count = IntegerToken::Cast(second);
+      format = FormatToken::Cast(third);
+      break;
+    }
+    default:
+      return new InvalidCommand(args, -1, "too many arguments");
+  }
+
+  return new MemCommand(args[0], target, count, format);
+}
+
+
+UnknownCommand::~UnknownCommand() {
+  const int size = args_.size();
+  for (int i = 0; i < size; ++i) {
+    delete args_[i];
+  }
+}
+
+
+bool UnknownCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+  USE(debugger);
+
+  printf(" ** Unknown Command:");
+  const int size = args_.size();
+  for (int i = 0; i < size; ++i) {
+    printf(" ");
+    args_[i]->Print(stdout);
+  }
+  printf(" **\n");
+
+  return false;
+}
+
+
+InvalidCommand::~InvalidCommand() {
+  const int size = args_.size();
+  for (int i = 0; i < size; ++i) {
+    delete args_[i];
+  }
+}
+
+
+bool InvalidCommand::Run(Debugger* debugger) {
+  ASSERT(debugger->IsDebuggerRunning());
+  USE(debugger);
+
+  printf(" ** Invalid Command:");
+  const int size = args_.size();
+  for (int i = 0; i < size; ++i) {
+    printf(" ");
+    if (i == index_) {
+      printf(">>");
+      args_[i]->Print(stdout);
+      printf("<<");
+    } else {
+      args_[i]->Print(stdout);
+    }
+  }
+  printf(" **\n");
+  printf(" ** %s\n", cause_);
+
+  return false;
+}
+
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/debugger-a64.h b/disas/libvixl/src/a64/debugger-a64.h
new file mode 100644
index 0000000..542d202
--- /dev/null
+++ b/disas/libvixl/src/a64/debugger-a64.h
@@ -0,0 +1,188 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_DEBUGGER_A64_H_
+#define VIXL_A64_DEBUGGER_A64_H_
+
+#include <ctype.h>
+#include <limits.h>
+#include <errno.h>
+#include <vector>
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/constants-a64.h"
+#include "a64/simulator-a64.h"
+
+namespace vixl {
+
+// Debug instructions.
+//
+// VIXL's macro-assembler and debugger support a few pseudo instructions to
+// make debugging easier. These pseudo instructions do not exist on real
+// hardware.
+//
+// Each debug pseudo instruction is represented by a HLT instruction. The HLT
+// immediate field is used to identify the type of debug pseudo isntruction.
+// Each pseudo instruction use a custom encoding for additional arguments, as
+// described below.
+
+// Unreachable
+//
+// Instruction which should never be executed. This is used as a guard in parts
+// of the code that should not be reachable, such as in data encoded inline in
+// the instructions.
+const Instr kUnreachableOpcode = 0xdeb0;
+
+// Trace
+//  - parameter: TraceParameter stored as a uint32_t
+//  - command: TraceCommand stored as a uint32_t
+//
+// Allow for trace management in the generated code. See the corresponding
+// enums for more information on permitted actions.
+const Instr kTraceOpcode = 0xdeb2;
+const unsigned kTraceParamsOffset = 1 * kInstructionSize;
+const unsigned kTraceCommandOffset = 2 * kInstructionSize;
+const unsigned kTraceLength = 3 * kInstructionSize;
+
+// Log
+//  - parameter: TraceParameter stored as a uint32_t
+//
+// Output the requested information.
+const Instr kLogOpcode = 0xdeb3;
+const unsigned kLogParamsOffset = 1 * kInstructionSize;
+const unsigned kLogLength = 2 * kInstructionSize;
+
+// Trace commands.
+enum TraceCommand {
+  TRACE_ENABLE   = 1,
+  TRACE_DISABLE  = 2
+};
+
+// Trace parameters.
+enum TraceParameters {
+  LOG_DISASM     = 1 << 0,  // Log disassembly.
+  LOG_REGS       = 1 << 1,  // Log general purpose registers.
+  LOG_FP_REGS    = 1 << 2,  // Log floating-point registers.
+  LOG_SYS_REGS   = 1 << 3,  // Log the flags and system registers.
+
+  LOG_STATE      = LOG_REGS | LOG_FP_REGS | LOG_SYS_REGS,
+  LOG_ALL        = LOG_DISASM | LOG_REGS | LOG_FP_REGS | LOG_SYS_REGS
+};
+
+// Debugger parameters
+enum DebugParameters {
+  DBG_ACTIVE = 1 << 0,  // The debugger is active.
+  DBG_BREAK  = 1 << 1   // The debugger is at a breakpoint.
+};
+
+// Forward declarations.
+class DebugCommand;
+class Token;
+class FormatToken;
+
+class Debugger : public Simulator {
+ public:
+  Debugger(Decoder* decoder, FILE* stream = stdout);
+
+  virtual void Run();
+  void VisitException(Instruction* instr);
+
+  inline int log_parameters() {
+    // The simulator can control disassembly, so make sure that the Debugger's
+    // log parameters agree with it.
+    if (disasm_trace()) {
+      log_parameters_ |= LOG_DISASM;
+    }
+    return log_parameters_;
+  }
+  inline void set_log_parameters(int parameters) {
+    set_disasm_trace((parameters & LOG_DISASM) != 0);
+    log_parameters_ = parameters;
+
+    update_pending_request();
+  }
+
+  inline int debug_parameters() { return debug_parameters_; }
+  inline void set_debug_parameters(int parameters) {
+    debug_parameters_ = parameters;
+
+    update_pending_request();
+  }
+
+  // Numbers of instructions to execute before the debugger shell is given
+  // back control.
+  inline int steps() { return steps_; }
+  inline void set_steps(int value) {
+    ASSERT(value > 1);
+    steps_ = value;
+  }
+
+  inline bool IsDebuggerRunning() {
+    return (debug_parameters_ & DBG_ACTIVE) != 0;
+  }
+
+  inline bool pending_request() { return pending_request_; }
+  inline void update_pending_request() {
+    const int kLoggingMask = LOG_STATE;
+    const bool logging = (log_parameters_ & kLoggingMask) != 0;
+    const bool debugging = IsDebuggerRunning();
+
+    pending_request_ = logging || debugging;
+  }
+
+  void PrintInstructions(void* address, int64_t count = 1);
+  void PrintMemory(const uint8_t* address,
+                   int64_t count,
+                   const FormatToken* format);
+
+ private:
+  void LogSystemRegisters();
+  void LogRegisters();
+  void LogFPRegisters();
+  void LogProcessorState();
+  char* ReadCommandLine(const char* prompt, char* buffer, int length);
+  void RunDebuggerShell();
+  void DoBreakpoint(Instruction* instr);
+  void DoUnreachable(Instruction* instr);
+  void DoTrace(Instruction* instr);
+  void DoLog(Instruction* instr);
+
+  int  log_parameters_;
+  int  debug_parameters_;
+  bool pending_request_;
+  int steps_;
+  DebugCommand* last_command_;
+  PrintDisassembler* disasm_;
+  Decoder* printer_;
+
+  // Length of the biggest command line accepted by the debugger shell.
+  static const int kMaxDebugShellLine = 256;
+};
+
+}  // namespace vixl
+
+#endif  // VIXL_A64_DEBUGGER_A64_H_
diff --git a/disas/libvixl/src/a64/decoder-a64.cc b/disas/libvixl/src/a64/decoder-a64.cc
new file mode 100644
index 0000000..9e9033c
--- /dev/null
+++ b/disas/libvixl/src/a64/decoder-a64.cc
@@ -0,0 +1,712 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/decoder-a64.h"
+
+namespace vixl {
+// Top-level instruction decode function.
+void Decoder::Decode(Instruction *instr) {
+  if (instr->Bits(28, 27) == 0) {
+    VisitUnallocated(instr);
+  } else {
+    switch (instr->Bits(27, 24)) {
+      // 0:   PC relative addressing.
+      case 0x0: DecodePCRelAddressing(instr); break;
+
+      // 1:   Add/sub immediate.
+      case 0x1: DecodeAddSubImmediate(instr); break;
+
+      // A:   Logical shifted register.
+      //      Add/sub with carry.
+      //      Conditional compare register.
+      //      Conditional compare immediate.
+      //      Conditional select.
+      //      Data processing 1 source.
+      //      Data processing 2 source.
+      // B:   Add/sub shifted register.
+      //      Add/sub extended register.
+      //      Data processing 3 source.
+      case 0xA:
+      case 0xB: DecodeDataProcessing(instr); break;
+
+      // 2:   Logical immediate.
+      //      Move wide immediate.
+      case 0x2: DecodeLogical(instr); break;
+
+      // 3:   Bitfield.
+      //      Extract.
+      case 0x3: DecodeBitfieldExtract(instr); break;
+
+      // 4:   Unconditional branch immediate.
+      //      Exception generation.
+      //      Compare and branch immediate.
+      // 5:   Compare and branch immediate.
+      //      Conditional branch.
+      //      System.
+      // 6,7: Unconditional branch.
+      //      Test and branch immediate.
+      case 0x4:
+      case 0x5:
+      case 0x6:
+      case 0x7: DecodeBranchSystemException(instr); break;
+
+      // 8,9: Load/store register pair post-index.
+      //      Load register literal.
+      //      Load/store register unscaled immediate.
+      //      Load/store register immediate post-index.
+      //      Load/store register immediate pre-index.
+      //      Load/store register offset.
+      //      Load/store exclusive.
+      // C,D: Load/store register pair offset.
+      //      Load/store register pair pre-index.
+      //      Load/store register unsigned immediate.
+      //      Advanced SIMD.
+      case 0x8:
+      case 0x9:
+      case 0xC:
+      case 0xD: DecodeLoadStore(instr); break;
+
+      // E:   FP fixed point conversion.
+      //      FP integer conversion.
+      //      FP data processing 1 source.
+      //      FP compare.
+      //      FP immediate.
+      //      FP data processing 2 source.
+      //      FP conditional compare.
+      //      FP conditional select.
+      //      Advanced SIMD.
+      // F:   FP data processing 3 source.
+      //      Advanced SIMD.
+      case 0xE:
+      case 0xF: DecodeFP(instr); break;
+    }
+  }
+}
+
+void Decoder::AppendVisitor(DecoderVisitor* new_visitor) {
+  visitors_.remove(new_visitor);
+  visitors_.push_front(new_visitor);
+}
+
+
+void Decoder::PrependVisitor(DecoderVisitor* new_visitor) {
+  visitors_.remove(new_visitor);
+  visitors_.push_back(new_visitor);
+}
+
+
+void Decoder::InsertVisitorBefore(DecoderVisitor* new_visitor,
+                                  DecoderVisitor* registered_visitor) {
+  visitors_.remove(new_visitor);
+  std::list<DecoderVisitor*>::iterator it;
+  for (it = visitors_.begin(); it != visitors_.end(); it++) {
+    if (*it == registered_visitor) {
+      visitors_.insert(it, new_visitor);
+      return;
+    }
+  }
+  // We reached the end of the list. The last element must be
+  // registered_visitor.
+  ASSERT(*it == registered_visitor);
+  visitors_.insert(it, new_visitor);
+}
+
+
+void Decoder::InsertVisitorAfter(DecoderVisitor* new_visitor,
+                                 DecoderVisitor* registered_visitor) {
+  visitors_.remove(new_visitor);
+  std::list<DecoderVisitor*>::iterator it;
+  for (it = visitors_.begin(); it != visitors_.end(); it++) {
+    if (*it == registered_visitor) {
+      it++;
+      visitors_.insert(it, new_visitor);
+      return;
+    }
+  }
+  // We reached the end of the list. The last element must be
+  // registered_visitor.
+  ASSERT(*it == registered_visitor);
+  visitors_.push_back(new_visitor);
+}
+
+
+void Decoder::RemoveVisitor(DecoderVisitor* visitor) {
+  visitors_.remove(visitor);
+}
+
+
+void Decoder::DecodePCRelAddressing(Instruction* instr) {
+  ASSERT(instr->Bits(27, 24) == 0x0);
+  // We know bit 28 is set, as <b28:b27> = 0 is filtered out at the top level
+  // decode.
+  ASSERT(instr->Bit(28) == 0x1);
+  VisitPCRelAddressing(instr);
+}
+
+
+void Decoder::DecodeBranchSystemException(Instruction* instr) {
+  ASSERT((instr->Bits(27, 24) == 0x4) ||
+         (instr->Bits(27, 24) == 0x5) ||
+         (instr->Bits(27, 24) == 0x6) ||
+         (instr->Bits(27, 24) == 0x7) );
+
+  switch (instr->Bits(31, 29)) {
+    case 0:
+    case 4: {
+      VisitUnconditionalBranch(instr);
+      break;
+    }
+    case 1:
+    case 5: {
+      if (instr->Bit(25) == 0) {
+        VisitCompareBranch(instr);
+      } else {
+        VisitTestBranch(instr);
+      }
+      break;
+    }
+    case 2: {
+      if (instr->Bit(25) == 0) {
+        if ((instr->Bit(24) == 0x1) ||
+            (instr->Mask(0x01000010) == 0x00000010)) {
+          VisitUnallocated(instr);
+        } else {
+          VisitConditionalBranch(instr);
+        }
+      } else {
+        VisitUnallocated(instr);
+      }
+      break;
+    }
+    case 6: {
+      if (instr->Bit(25) == 0) {
+        if (instr->Bit(24) == 0) {
+          if ((instr->Bits(4, 2) != 0) ||
+              (instr->Mask(0x00E0001D) == 0x00200001) ||
+              (instr->Mask(0x00E0001D) == 0x00400001) ||
+              (instr->Mask(0x00E0001E) == 0x00200002) ||
+              (instr->Mask(0x00E0001E) == 0x00400002) ||
+              (instr->Mask(0x00E0001C) == 0x00600000) ||
+              (instr->Mask(0x00E0001C) == 0x00800000) ||
+              (instr->Mask(0x00E0001F) == 0x00A00000) ||
+              (instr->Mask(0x00C0001C) == 0x00C00000)) {
+            VisitUnallocated(instr);
+          } else {
+            VisitException(instr);
+          }
+        } else {
+          if (instr->Bits(23, 22) == 0) {
+            const Instr masked_003FF0E0 = instr->Mask(0x003FF0E0);
+            if ((instr->Bits(21, 19) == 0x4) ||
+                (masked_003FF0E0 == 0x00033000) ||
+                (masked_003FF0E0 == 0x003FF020) ||
+                (masked_003FF0E0 == 0x003FF060) ||
+                (masked_003FF0E0 == 0x003FF0E0) ||
+                (instr->Mask(0x00388000) == 0x00008000) ||
+                (instr->Mask(0x0038E000) == 0x00000000) ||
+                (instr->Mask(0x0039E000) == 0x00002000) ||
+                (instr->Mask(0x003AE000) == 0x00002000) ||
+                (instr->Mask(0x003CE000) == 0x00042000) ||
+                (instr->Mask(0x003FFFC0) == 0x000320C0) ||
+                (instr->Mask(0x003FF100) == 0x00032100) ||
+                (instr->Mask(0x003FF200) == 0x00032200) ||
+                (instr->Mask(0x003FF400) == 0x00032400) ||
+                (instr->Mask(0x003FF800) == 0x00032800) ||
+                (instr->Mask(0x0038F000) == 0x00005000) ||
+                (instr->Mask(0x0038E000) == 0x00006000)) {
+              VisitUnallocated(instr);
+            } else {
+              VisitSystem(instr);
+            }
+          } else {
+            VisitUnallocated(instr);
+          }
+        }
+      } else {
+        if ((instr->Bit(24) == 0x1) ||
+            (instr->Bits(20, 16) != 0x1F) ||
+            (instr->Bits(15, 10) != 0) ||
+            (instr->Bits(4, 0) != 0) ||
+            (instr->Bits(24, 21) == 0x3) ||
+            (instr->Bits(24, 22) == 0x3)) {
+          VisitUnallocated(instr);
+        } else {
+          VisitUnconditionalBranchToRegister(instr);
+        }
+      }
+      break;
+    }
+    case 3:
+    case 7: {
+      VisitUnallocated(instr);
+      break;
+    }
+  }
+}
+
+
+void Decoder::DecodeLoadStore(Instruction* instr) {
+  ASSERT((instr->Bits(27, 24) == 0x8) ||
+         (instr->Bits(27, 24) == 0x9) ||
+         (instr->Bits(27, 24) == 0xC) ||
+         (instr->Bits(27, 24) == 0xD) );
+
+  if (instr->Bit(24) == 0) {
+    if (instr->Bit(28) == 0) {
+      if (instr->Bit(29) == 0) {
+        if (instr->Bit(26) == 0) {
+          // TODO: VisitLoadStoreExclusive.
+          VisitUnimplemented(instr);
+        } else {
+          DecodeAdvSIMDLoadStore(instr);
+        }
+      } else {
+        if ((instr->Bits(31, 30) == 0x3) ||
+            (instr->Mask(0xC4400000) == 0x40000000)) {
+          VisitUnallocated(instr);
+        } else {
+          if (instr->Bit(23) == 0) {
+            if (instr->Mask(0xC4400000) == 0xC0400000) {
+              VisitUnallocated(instr);
+            } else {
+              VisitLoadStorePairNonTemporal(instr);
+            }
+          } else {
+            VisitLoadStorePairPostIndex(instr);
+          }
+        }
+      }
+    } else {
+      if (instr->Bit(29) == 0) {
+        if (instr->Mask(0xC4000000) == 0xC4000000) {
+          VisitUnallocated(instr);
+        } else {
+          VisitLoadLiteral(instr);
+        }
+      } else {
+        if ((instr->Mask(0x84C00000) == 0x80C00000) ||
+            (instr->Mask(0x44800000) == 0x44800000) ||
+            (instr->Mask(0x84800000) == 0x84800000)) {
+          VisitUnallocated(instr);
+        } else {
+          if (instr->Bit(21) == 0) {
+            switch (instr->Bits(11, 10)) {
+              case 0: {
+                VisitLoadStoreUnscaledOffset(instr);
+                break;
+              }
+              case 1: {
+                if (instr->Mask(0xC4C00000) == 0xC0800000) {
+                  VisitUnallocated(instr);
+                } else {
+                  VisitLoadStorePostIndex(instr);
+                }
+                break;
+              }
+              case 2: {
+                // TODO: VisitLoadStoreRegisterOffsetUnpriv.
+                VisitUnimplemented(instr);
+                break;
+              }
+              case 3: {
+                if (instr->Mask(0xC4C00000) == 0xC0800000) {
+                  VisitUnallocated(instr);
+                } else {
+                  VisitLoadStorePreIndex(instr);
+                }
+                break;
+              }
+            }
+          } else {
+            if (instr->Bits(11, 10) == 0x2) {
+              if (instr->Bit(14) == 0) {
+                VisitUnallocated(instr);
+              } else {
+                VisitLoadStoreRegisterOffset(instr);
+              }
+            } else {
+              VisitUnallocated(instr);
+            }
+          }
+        }
+      }
+    }
+  } else {
+    if (instr->Bit(28) == 0) {
+      if (instr->Bit(29) == 0) {
+        VisitUnallocated(instr);
+      } else {
+        if ((instr->Bits(31, 30) == 0x3) ||
+            (instr->Mask(0xC4400000) == 0x40000000)) {
+          VisitUnallocated(instr);
+        } else {
+          if (instr->Bit(23) == 0) {
+            VisitLoadStorePairOffset(instr);
+          } else {
+            VisitLoadStorePairPreIndex(instr);
+          }
+        }
+      }
+    } else {
+      if (instr->Bit(29) == 0) {
+        VisitUnallocated(instr);
+      } else {
+        if ((instr->Mask(0x84C00000) == 0x80C00000) ||
+            (instr->Mask(0x44800000) == 0x44800000) ||
+            (instr->Mask(0x84800000) == 0x84800000)) {
+          VisitUnallocated(instr);
+        } else {
+          VisitLoadStoreUnsignedOffset(instr);
+        }
+      }
+    }
+  }
+}
+
+
+void Decoder::DecodeLogical(Instruction* instr) {
+  ASSERT(instr->Bits(27, 24) == 0x2);
+
+  if (instr->Mask(0x80400000) == 0x00400000) {
+    VisitUnallocated(instr);
+  } else {
+    if (instr->Bit(23) == 0) {
+      VisitLogicalImmediate(instr);
+    } else {
+      if (instr->Bits(30, 29) == 0x1) {
+        VisitUnallocated(instr);
+      } else {
+        VisitMoveWideImmediate(instr);
+      }
+    }
+  }
+}
+
+
+void Decoder::DecodeBitfieldExtract(Instruction* instr) {
+  ASSERT(instr->Bits(27, 24) == 0x3);
+
+  if ((instr->Mask(0x80400000) == 0x80000000) ||
+      (instr->Mask(0x80400000) == 0x00400000) ||
+      (instr->Mask(0x80008000) == 0x00008000)) {
+    VisitUnallocated(instr);
+  } else if (instr->Bit(23) == 0) {
+    if ((instr->Mask(0x80200000) == 0x00200000) ||
+        (instr->Mask(0x60000000) == 0x60000000)) {
+      VisitUnallocated(instr);
+    } else {
+      VisitBitfield(instr);
+    }
+  } else {
+    if ((instr->Mask(0x60200000) == 0x00200000) ||
+        (instr->Mask(0x60000000) != 0x00000000)) {
+      VisitUnallocated(instr);
+    } else {
+      VisitExtract(instr);
+    }
+  }
+}
+
+
+void Decoder::DecodeAddSubImmediate(Instruction* instr) {
+  ASSERT(instr->Bits(27, 24) == 0x1);
+  if (instr->Bit(23) == 1) {
+    VisitUnallocated(instr);
+  } else {
+    VisitAddSubImmediate(instr);
+  }
+}
+
+
+void Decoder::DecodeDataProcessing(Instruction* instr) {
+  ASSERT((instr->Bits(27, 24) == 0xA) ||
+         (instr->Bits(27, 24) == 0xB) );
+
+  if (instr->Bit(24) == 0) {
+    if (instr->Bit(28) == 0) {
+      if (instr->Mask(0x80008000) == 0x00008000) {
+        VisitUnallocated(instr);
+      } else {
+        VisitLogicalShifted(instr);
+      }
+    } else {
+      switch (instr->Bits(23, 21)) {
+        case 0: {
+          if (instr->Mask(0x0000FC00) != 0) {
+            VisitUnallocated(instr);
+          } else {
+            VisitAddSubWithCarry(instr);
+          }
+          break;
+        }
+        case 2: {
+          if ((instr->Bit(29) == 0) ||
+              (instr->Mask(0x00000410) != 0)) {
+            VisitUnallocated(instr);
+          } else {
+            if (instr->Bit(11) == 0) {
+              VisitConditionalCompareRegister(instr);
+            } else {
+              VisitConditionalCompareImmediate(instr);
+            }
+          }
+          break;
+        }
+        case 4: {
+          if (instr->Mask(0x20000800) != 0x00000000) {
+            VisitUnallocated(instr);
+          } else {
+            VisitConditionalSelect(instr);
+          }
+          break;
+        }
+        case 6: {
+          if (instr->Bit(29) == 0x1) {
+            VisitUnallocated(instr);
+          } else {
+            if (instr->Bit(30) == 0) {
+              if ((instr->Bit(15) == 0x1) ||
+                  (instr->Bits(15, 11) == 0) ||
+                  (instr->Bits(15, 12) == 0x1) ||
+                  (instr->Bits(15, 12) == 0x3) ||
+                  (instr->Bits(15, 13) == 0x3) ||
+                  (instr->Mask(0x8000EC00) == 0x00004C00) ||
+                  (instr->Mask(0x8000E800) == 0x80004000) ||
+                  (instr->Mask(0x8000E400) == 0x80004000)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitDataProcessing2Source(instr);
+              }
+            } else {
+              if ((instr->Bit(13) == 1) ||
+                  (instr->Bits(20, 16) != 0) ||
+                  (instr->Bits(15, 14) != 0) ||
+                  (instr->Mask(0xA01FFC00) == 0x00000C00) ||
+                  (instr->Mask(0x201FF800) == 0x00001800)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitDataProcessing1Source(instr);
+              }
+            }
+            break;
+          }
+        }
+        case 1:
+        case 3:
+        case 5:
+        case 7: VisitUnallocated(instr); break;
+      }
+    }
+  } else {
+    if (instr->Bit(28) == 0) {
+     if (instr->Bit(21) == 0) {
+        if ((instr->Bits(23, 22) == 0x3) ||
+            (instr->Mask(0x80008000) == 0x00008000)) {
+          VisitUnallocated(instr);
+        } else {
+          VisitAddSubShifted(instr);
+        }
+      } else {
+        if ((instr->Mask(0x00C00000) != 0x00000000) ||
+            (instr->Mask(0x00001400) == 0x00001400) ||
+            (instr->Mask(0x00001800) == 0x00001800)) {
+          VisitUnallocated(instr);
+        } else {
+          VisitAddSubExtended(instr);
+        }
+      }
+    } else {
+      if ((instr->Bit(30) == 0x1) ||
+          (instr->Bits(30, 29) == 0x1) ||
+          (instr->Mask(0xE0600000) == 0x00200000) ||
+          (instr->Mask(0xE0608000) == 0x00400000) ||
+          (instr->Mask(0x60608000) == 0x00408000) ||
+          (instr->Mask(0x60E00000) == 0x00E00000) ||
+          (instr->Mask(0x60E00000) == 0x00800000) ||
+          (instr->Mask(0x60E00000) == 0x00600000)) {
+        VisitUnallocated(instr);
+      } else {
+        VisitDataProcessing3Source(instr);
+      }
+    }
+  }
+}
+
+
+void Decoder::DecodeFP(Instruction* instr) {
+  ASSERT((instr->Bits(27, 24) == 0xE) ||
+         (instr->Bits(27, 24) == 0xF) );
+
+  if (instr->Bit(28) == 0) {
+    DecodeAdvSIMDDataProcessing(instr);
+  } else {
+    if (instr->Bit(29) == 1) {
+      VisitUnallocated(instr);
+    } else {
+      if (instr->Bits(31, 30) == 0x3) {
+        VisitUnallocated(instr);
+      } else if (instr->Bits(31, 30) == 0x1) {
+        DecodeAdvSIMDDataProcessing(instr);
+      } else {
+        if (instr->Bit(24) == 0) {
+          if (instr->Bit(21) == 0) {
+            if ((instr->Bit(23) == 1) ||
+                (instr->Bit(18) == 1) ||
+                (instr->Mask(0x80008000) == 0x00000000) ||
+                (instr->Mask(0x000E0000) == 0x00000000) ||
+                (instr->Mask(0x000E0000) == 0x000A0000) ||
+                (instr->Mask(0x00160000) == 0x00000000) ||
+                (instr->Mask(0x00160000) == 0x00120000)) {
+              VisitUnallocated(instr);
+            } else {
+              VisitFPFixedPointConvert(instr);
+            }
+          } else {
+            if (instr->Bits(15, 10) == 32) {
+              VisitUnallocated(instr);
+            } else if (instr->Bits(15, 10) == 0) {
+              if ((instr->Bits(23, 22) == 0x3) ||
+                  (instr->Mask(0x000E0000) == 0x000A0000) ||
+                  (instr->Mask(0x000E0000) == 0x000C0000) ||
+                  (instr->Mask(0x00160000) == 0x00120000) ||
+                  (instr->Mask(0x00160000) == 0x00140000) ||
+                  (instr->Mask(0x20C40000) == 0x00800000) ||
+                  (instr->Mask(0x20C60000) == 0x00840000) ||
+                  (instr->Mask(0xA0C60000) == 0x80060000) ||
+                  (instr->Mask(0xA0C60000) == 0x00860000) ||
+                  (instr->Mask(0xA0C60000) == 0x00460000) ||
+                  (instr->Mask(0xA0CE0000) == 0x80860000) ||
+                  (instr->Mask(0xA0CE0000) == 0x804E0000) ||
+                  (instr->Mask(0xA0CE0000) == 0x000E0000) ||
+                  (instr->Mask(0xA0D60000) == 0x00160000) ||
+                  (instr->Mask(0xA0D60000) == 0x80560000) ||
+                  (instr->Mask(0xA0D60000) == 0x80960000)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitFPIntegerConvert(instr);
+              }
+            } else if (instr->Bits(14, 10) == 16) {
+              const Instr masked_A0DF8000 = instr->Mask(0xA0DF8000);
+              if ((instr->Mask(0x80180000) != 0) ||
+                  (masked_A0DF8000 == 0x00020000) ||
+                  (masked_A0DF8000 == 0x00030000) ||
+                  (masked_A0DF8000 == 0x00068000) ||
+                  (masked_A0DF8000 == 0x00428000) ||
+                  (masked_A0DF8000 == 0x00430000) ||
+                  (masked_A0DF8000 == 0x00468000) ||
+                  (instr->Mask(0xA0D80000) == 0x00800000) ||
+                  (instr->Mask(0xA0DE0000) == 0x00C00000) ||
+                  (instr->Mask(0xA0DF0000) == 0x00C30000) ||
+                  (instr->Mask(0xA0DC0000) == 0x00C40000)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitFPDataProcessing1Source(instr);
+              }
+            } else if (instr->Bits(13, 10) == 8) {
+              if ((instr->Bits(15, 14) != 0) ||
+                  (instr->Bits(2, 0) != 0) ||
+                  (instr->Mask(0x80800000) != 0x00000000)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitFPCompare(instr);
+              }
+            } else if (instr->Bits(12, 10) == 4) {
+              if ((instr->Bits(9, 5) != 0) ||
+                  (instr->Mask(0x80800000) != 0x00000000)) {
+                VisitUnallocated(instr);
+              } else {
+                VisitFPImmediate(instr);
+              }
+            } else {
+              if (instr->Mask(0x80800000) != 0x00000000) {
+                VisitUnallocated(instr);
+              } else {
+                switch (instr->Bits(11, 10)) {
+                  case 1: {
+                    VisitFPConditionalCompare(instr);
+                    break;
+                  }
+                  case 2: {
+                    if ((instr->Bits(15, 14) == 0x3) ||
+                        (instr->Mask(0x00009000) == 0x00009000) ||
+                        (instr->Mask(0x0000A000) == 0x0000A000)) {
+                      VisitUnallocated(instr);
+                    } else {
+                      VisitFPDataProcessing2Source(instr);
+                    }
+                    break;
+                  }
+                  case 3: {
+                    VisitFPConditionalSelect(instr);
+                    break;
+                  }
+                  default: UNREACHABLE();
+                }
+              }
+            }
+          }
+        } else {
+          // Bit 30 == 1 has been handled earlier.
+          ASSERT(instr->Bit(30) == 0);
+          if (instr->Mask(0xA0800000) != 0) {
+            VisitUnallocated(instr);
+          } else {
+            VisitFPDataProcessing3Source(instr);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+void Decoder::DecodeAdvSIMDLoadStore(Instruction* instr) {
+  // TODO: Implement Advanced SIMD load/store instruction decode.
+  ASSERT(instr->Bits(29, 25) == 0x6);
+  VisitUnimplemented(instr);
+}
+
+
+void Decoder::DecodeAdvSIMDDataProcessing(Instruction* instr) {
+  // TODO: Implement Advanced SIMD data processing instruction decode.
+  ASSERT(instr->Bits(27, 25) == 0x7);
+  VisitUnimplemented(instr);
+}
+
+
+#define DEFINE_VISITOR_CALLERS(A)                                              \
+  void Decoder::Visit##A(Instruction *instr) {                                 \
+    ASSERT(instr->Mask(A##FMask) == A##Fixed);                                 \
+    std::list<DecoderVisitor*>::iterator it;                                   \
+    for (it = visitors_.begin(); it != visitors_.end(); it++) {                \
+      (*it)->Visit##A(instr);                                                  \
+    }                                                                          \
+  }
+VISITOR_LIST(DEFINE_VISITOR_CALLERS)
+#undef DEFINE_VISITOR_CALLERS
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/decoder-a64.h b/disas/libvixl/src/a64/decoder-a64.h
new file mode 100644
index 0000000..bbbbd81
--- /dev/null
+++ b/disas/libvixl/src/a64/decoder-a64.h
@@ -0,0 +1,198 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_DECODER_A64_H_
+#define VIXL_A64_DECODER_A64_H_
+
+#include <list>
+
+#include "globals.h"
+#include "a64/instructions-a64.h"
+
+
+// List macro containing all visitors needed by the decoder class.
+
+#define VISITOR_LIST(V)             \
+  V(PCRelAddressing)                \
+  V(AddSubImmediate)                \
+  V(LogicalImmediate)               \
+  V(MoveWideImmediate)              \
+  V(Bitfield)                       \
+  V(Extract)                        \
+  V(UnconditionalBranch)            \
+  V(UnconditionalBranchToRegister)  \
+  V(CompareBranch)                  \
+  V(TestBranch)                     \
+  V(ConditionalBranch)              \
+  V(System)                         \
+  V(Exception)                      \
+  V(LoadStorePairPostIndex)         \
+  V(LoadStorePairOffset)            \
+  V(LoadStorePairPreIndex)          \
+  V(LoadStorePairNonTemporal)       \
+  V(LoadLiteral)                    \
+  V(LoadStoreUnscaledOffset)        \
+  V(LoadStorePostIndex)             \
+  V(LoadStorePreIndex)              \
+  V(LoadStoreRegisterOffset)        \
+  V(LoadStoreUnsignedOffset)        \
+  V(LogicalShifted)                 \
+  V(AddSubShifted)                  \
+  V(AddSubExtended)                 \
+  V(AddSubWithCarry)                \
+  V(ConditionalCompareRegister)     \
+  V(ConditionalCompareImmediate)    \
+  V(ConditionalSelect)              \
+  V(DataProcessing1Source)          \
+  V(DataProcessing2Source)          \
+  V(DataProcessing3Source)          \
+  V(FPCompare)                      \
+  V(FPConditionalCompare)           \
+  V(FPConditionalSelect)            \
+  V(FPImmediate)                    \
+  V(FPDataProcessing1Source)        \
+  V(FPDataProcessing2Source)        \
+  V(FPDataProcessing3Source)        \
+  V(FPIntegerConvert)               \
+  V(FPFixedPointConvert)            \
+  V(Unallocated)                    \
+  V(Unimplemented)
+
+namespace vixl {
+
+// The Visitor interface. Disassembler and simulator (and other tools)
+// must provide implementations for all of these functions.
+class DecoderVisitor {
+ public:
+  #define DECLARE(A) virtual void Visit##A(Instruction* instr) = 0;
+  VISITOR_LIST(DECLARE)
+  #undef DECLARE
+
+  virtual ~DecoderVisitor() {}
+
+ private:
+  // Visitors are registered in a list.
+  std::list<DecoderVisitor*> visitors_;
+
+  friend class Decoder;
+};
+
+
+class Decoder: public DecoderVisitor {
+ public:
+  Decoder() {}
+
+  // Top-level instruction decoder function. Decodes an instruction and calls
+  // the visitor functions registered with the Decoder class.
+  void Decode(Instruction *instr);
+
+  // Register a new visitor class with the decoder.
+  // Decode() will call the corresponding visitor method from all registered
+  // visitor classes when decoding reaches the leaf node of the instruction
+  // decode tree.
+  // Visitors are called in the order.
+  // A visitor can only be registered once.
+  // Registering an already registered visitor will update its position.
+  //
+  //   d.AppendVisitor(V1);
+  //   d.AppendVisitor(V2);
+  //   d.PrependVisitor(V2);            // Move V2 at the start of the list.
+  //   d.InsertVisitorBefore(V3, V2);
+  //   d.AppendVisitor(V4);
+  //   d.AppendVisitor(V4);             // No effect.
+  //
+  //   d.Decode(i);
+  //
+  // will call in order visitor methods in V3, V2, V1, V4.
+  void AppendVisitor(DecoderVisitor* visitor);
+  void PrependVisitor(DecoderVisitor* visitor);
+  void InsertVisitorBefore(DecoderVisitor* new_visitor,
+                           DecoderVisitor* registered_visitor);
+  void InsertVisitorAfter(DecoderVisitor* new_visitor,
+                          DecoderVisitor* registered_visitor);
+
+  // Remove a previously registered visitor class from the list of visitors
+  // stored by the decoder.
+  void RemoveVisitor(DecoderVisitor* visitor);
+
+  #define DECLARE(A) void Visit##A(Instruction* instr);
+  VISITOR_LIST(DECLARE)
+  #undef DECLARE
+
+ private:
+  // Decode the PC relative addressing instruction, and call the corresponding
+  // visitors.
+  // On entry, instruction bits 27:24 = 0x0.
+  void DecodePCRelAddressing(Instruction* instr);
+
+  // Decode the add/subtract immediate instruction, and call the correspoding
+  // visitors.
+  // On entry, instruction bits 27:24 = 0x1.
+  void DecodeAddSubImmediate(Instruction* instr);
+
+  // Decode the branch, system command, and exception generation parts of
+  // the instruction tree, and call the corresponding visitors.
+  // On entry, instruction bits 27:24 = {0x4, 0x5, 0x6, 0x7}.
+  void DecodeBranchSystemException(Instruction* instr);
+
+  // Decode the load and store parts of the instruction tree, and call
+  // the corresponding visitors.
+  // On entry, instruction bits 27:24 = {0x8, 0x9, 0xC, 0xD}.
+  void DecodeLoadStore(Instruction* instr);
+
+  // Decode the logical immediate and move wide immediate parts of the
+  // instruction tree, and call the corresponding visitors.
+  // On entry, instruction bits 27:24 = 0x2.
+  void DecodeLogical(Instruction* instr);
+
+  // Decode the bitfield and extraction parts of the instruction tree,
+  // and call the corresponding visitors.
+  // On entry, instruction bits 27:24 = 0x3.
+  void DecodeBitfieldExtract(Instruction* instr);
+
+  // Decode the data processing parts of the instruction tree, and call the
+  // corresponding visitors.
+  // On entry, instruction bits 27:24 = {0x1, 0xA, 0xB}.
+  void DecodeDataProcessing(Instruction* instr);
+
+  // Decode the floating point parts of the instruction tree, and call the
+  // corresponding visitors.
+  // On entry, instruction bits 27:24 = {0xE, 0xF}.
+  void DecodeFP(Instruction* instr);
+
+  // Decode the Advanced SIMD (NEON) load/store part of the instruction tree,
+  // and call the corresponding visitors.
+  // On entry, instruction bits 29:25 = 0x6.
+  void DecodeAdvSIMDLoadStore(Instruction* instr);
+
+  // Decode the Advanced SIMD (NEON) data processing part of the instruction
+  // tree, and call the corresponding visitors.
+  // On entry, instruction bits 27:25 = 0x7.
+  void DecodeAdvSIMDDataProcessing(Instruction* instr);
+};
+}  // namespace vixl
+
+#endif  // VIXL_A64_DECODER_A64_H_
diff --git a/disas/libvixl/src/a64/disasm-a64.cc b/disas/libvixl/src/a64/disasm-a64.cc
new file mode 100644
index 0000000..4a49748
--- /dev/null
+++ b/disas/libvixl/src/a64/disasm-a64.cc
@@ -0,0 +1,1678 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "a64/disasm-a64.h"
+
+namespace vixl {
+
+Disassembler::Disassembler() {
+  buffer_size_ = 256;
+  buffer_ = reinterpret_cast<char*>(malloc(buffer_size_));
+  buffer_pos_ = 0;
+  own_buffer_ = true;
+}
+
+
+Disassembler::Disassembler(char* text_buffer, int buffer_size) {
+  buffer_size_ = buffer_size;
+  buffer_ = text_buffer;
+  buffer_pos_ = 0;
+  own_buffer_ = false;
+}
+
+
+Disassembler::~Disassembler() {
+  if (own_buffer_) {
+    free(buffer_);
+  }
+}
+
+
+char* Disassembler::GetOutput() {
+  return buffer_;
+}
+
+
+void Disassembler::VisitAddSubImmediate(Instruction* instr) {
+  bool rd_is_zr = RdIsZROrSP(instr);
+  bool stack_op = (rd_is_zr || RnIsZROrSP(instr)) &&
+                  (instr->ImmAddSub() == 0) ? true : false;
+  const char *mnemonic = "";
+  const char *form = "'Rds, 'Rns, 'IAddSub";
+  const char *form_cmp = "'Rns, 'IAddSub";
+  const char *form_mov = "'Rds, 'Rns";
+
+  switch (instr->Mask(AddSubImmediateMask)) {
+    case ADD_w_imm:
+    case ADD_x_imm: {
+      mnemonic = "add";
+      if (stack_op) {
+        mnemonic = "mov";
+        form = form_mov;
+      }
+      break;
+    }
+    case ADDS_w_imm:
+    case ADDS_x_imm: {
+      mnemonic = "adds";
+      if (rd_is_zr) {
+        mnemonic = "cmn";
+        form = form_cmp;
+      }
+      break;
+    }
+    case SUB_w_imm:
+    case SUB_x_imm: mnemonic = "sub"; break;
+    case SUBS_w_imm:
+    case SUBS_x_imm: {
+      mnemonic = "subs";
+      if (rd_is_zr) {
+        mnemonic = "cmp";
+        form = form_cmp;
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitAddSubShifted(Instruction* instr) {
+  bool rd_is_zr = RdIsZROrSP(instr);
+  bool rn_is_zr = RnIsZROrSP(instr);
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn, 'Rm'HDP";
+  const char *form_cmp = "'Rn, 'Rm'HDP";
+  const char *form_neg = "'Rd, 'Rm'HDP";
+
+  switch (instr->Mask(AddSubShiftedMask)) {
+    case ADD_w_shift:
+    case ADD_x_shift: mnemonic = "add"; break;
+    case ADDS_w_shift:
+    case ADDS_x_shift: {
+      mnemonic = "adds";
+      if (rd_is_zr) {
+        mnemonic = "cmn";
+        form = form_cmp;
+      }
+      break;
+    }
+    case SUB_w_shift:
+    case SUB_x_shift: {
+      mnemonic = "sub";
+      if (rn_is_zr) {
+        mnemonic = "neg";
+        form = form_neg;
+      }
+      break;
+    }
+    case SUBS_w_shift:
+    case SUBS_x_shift: {
+      mnemonic = "subs";
+      if (rd_is_zr) {
+        mnemonic = "cmp";
+        form = form_cmp;
+      } else if (rn_is_zr) {
+        mnemonic = "negs";
+        form = form_neg;
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitAddSubExtended(Instruction* instr) {
+  bool rd_is_zr = RdIsZROrSP(instr);
+  const char *mnemonic = "";
+  Extend mode = static_cast<Extend>(instr->ExtendMode());
+  const char *form = ((mode == UXTX) || (mode == SXTX)) ?
+                     "'Rds, 'Rns, 'Xm'Ext" : "'Rds, 'Rns, 'Wm'Ext";
+  const char *form_cmp = ((mode == UXTX) || (mode == SXTX)) ?
+                         "'Rns, 'Xm'Ext" : "'Rns, 'Wm'Ext";
+
+  switch (instr->Mask(AddSubExtendedMask)) {
+    case ADD_w_ext:
+    case ADD_x_ext: mnemonic = "add"; break;
+    case ADDS_w_ext:
+    case ADDS_x_ext: {
+      mnemonic = "adds";
+      if (rd_is_zr) {
+        mnemonic = "cmn";
+        form = form_cmp;
+      }
+      break;
+    }
+    case SUB_w_ext:
+    case SUB_x_ext: mnemonic = "sub"; break;
+    case SUBS_w_ext:
+    case SUBS_x_ext: {
+      mnemonic = "subs";
+      if (rd_is_zr) {
+        mnemonic = "cmp";
+        form = form_cmp;
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitAddSubWithCarry(Instruction* instr) {
+  bool rn_is_zr = RnIsZROrSP(instr);
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn, 'Rm";
+  const char *form_neg = "'Rd, 'Rm";
+
+  switch (instr->Mask(AddSubWithCarryMask)) {
+    case ADC_w:
+    case ADC_x: mnemonic = "adc"; break;
+    case ADCS_w:
+    case ADCS_x: mnemonic = "adcs"; break;
+    case SBC_w:
+    case SBC_x: {
+      mnemonic = "sbc";
+      if (rn_is_zr) {
+        mnemonic = "ngc";
+        form = form_neg;
+      }
+      break;
+    }
+    case SBCS_w:
+    case SBCS_x: {
+      mnemonic = "sbcs";
+      if (rn_is_zr) {
+        mnemonic = "ngcs";
+        form = form_neg;
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLogicalImmediate(Instruction* instr) {
+  bool rd_is_zr = RdIsZROrSP(instr);
+  bool rn_is_zr = RnIsZROrSP(instr);
+  const char *mnemonic = "";
+  const char *form = "'Rds, 'Rn, 'ITri";
+
+  if (instr->ImmLogical() == 0) {
+    // The immediate encoded in the instruction is not in the expected format.
+    Format(instr, "unallocated", "(LogicalImmediate)");
+    return;
+  }
+
+  switch (instr->Mask(LogicalImmediateMask)) {
+    case AND_w_imm:
+    case AND_x_imm: mnemonic = "and"; break;
+    case ORR_w_imm:
+    case ORR_x_imm: {
+      mnemonic = "orr";
+      unsigned reg_size = (instr->SixtyFourBits() == 1) ? kXRegSize
+                                                        : kWRegSize;
+      if (rn_is_zr && !IsMovzMovnImm(reg_size, instr->ImmLogical())) {
+        mnemonic = "mov";
+        form = "'Rds, 'ITri";
+      }
+      break;
+    }
+    case EOR_w_imm:
+    case EOR_x_imm: mnemonic = "eor"; break;
+    case ANDS_w_imm:
+    case ANDS_x_imm: {
+      mnemonic = "ands";
+      if (rd_is_zr) {
+        mnemonic = "tst";
+        form = "'Rn, 'ITri";
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+bool Disassembler::IsMovzMovnImm(unsigned reg_size, uint64_t value) {
+  ASSERT((reg_size == kXRegSize) ||
+         ((reg_size == kWRegSize) && (value <= 0xffffffff)));
+
+  // Test for movz: 16 bits set at positions 0, 16, 32 or 48.
+  if (((value & 0xffffffffffff0000UL) == 0UL) ||
+      ((value & 0xffffffff0000ffffUL) == 0UL) ||
+      ((value & 0xffff0000ffffffffUL) == 0UL) ||
+      ((value & 0x0000ffffffffffffUL) == 0UL)) {
+    return true;
+  }
+
+  // Test for movn: NOT(16 bits set at positions 0, 16, 32 or 48).
+  if ((reg_size == kXRegSize) &&
+      (((value & 0xffffffffffff0000UL) == 0xffffffffffff0000UL) ||
+       ((value & 0xffffffff0000ffffUL) == 0xffffffff0000ffffUL) ||
+       ((value & 0xffff0000ffffffffUL) == 0xffff0000ffffffffUL) ||
+       ((value & 0x0000ffffffffffffUL) == 0x0000ffffffffffffUL))) {
+    return true;
+  }
+  if ((reg_size == kWRegSize) &&
+      (((value & 0xffff0000) == 0xffff0000) ||
+       ((value & 0x0000ffff) == 0x0000ffff))) {
+    return true;
+  }
+  return false;
+}
+
+
+void Disassembler::VisitLogicalShifted(Instruction* instr) {
+  bool rd_is_zr = RdIsZROrSP(instr);
+  bool rn_is_zr = RnIsZROrSP(instr);
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn, 'Rm'HLo";
+
+  switch (instr->Mask(LogicalShiftedMask)) {
+    case AND_w:
+    case AND_x: mnemonic = "and"; break;
+    case BIC_w:
+    case BIC_x: mnemonic = "bic"; break;
+    case EOR_w:
+    case EOR_x: mnemonic = "eor"; break;
+    case EON_w:
+    case EON_x: mnemonic = "eon"; break;
+    case BICS_w:
+    case BICS_x: mnemonic = "bics"; break;
+    case ANDS_w:
+    case ANDS_x: {
+      mnemonic = "ands";
+      if (rd_is_zr) {
+        mnemonic = "tst";
+        form = "'Rn, 'Rm'HLo";
+      }
+      break;
+    }
+    case ORR_w:
+    case ORR_x: {
+      mnemonic = "orr";
+      if (rn_is_zr && (instr->ImmDPShift() == 0) && (instr->ShiftDP() == LSL)) {
+        mnemonic = "mov";
+        form = "'Rd, 'Rm";
+      }
+      break;
+    }
+    case ORN_w:
+    case ORN_x: {
+      mnemonic = "orn";
+      if (rn_is_zr) {
+        mnemonic = "mvn";
+        form = "'Rd, 'Rm'HLo";
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitConditionalCompareRegister(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rn, 'Rm, 'INzcv, 'Cond";
+
+  switch (instr->Mask(ConditionalCompareRegisterMask)) {
+    case CCMN_w:
+    case CCMN_x: mnemonic = "ccmn"; break;
+    case CCMP_w:
+    case CCMP_x: mnemonic = "ccmp"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitConditionalCompareImmediate(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rn, 'IP, 'INzcv, 'Cond";
+
+  switch (instr->Mask(ConditionalCompareImmediateMask)) {
+    case CCMN_w_imm:
+    case CCMN_x_imm: mnemonic = "ccmn"; break;
+    case CCMP_w_imm:
+    case CCMP_x_imm: mnemonic = "ccmp"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitConditionalSelect(Instruction* instr) {
+  bool rnm_is_zr = (RnIsZROrSP(instr) && RmIsZROrSP(instr));
+  bool rn_is_rm = (instr->Rn() == instr->Rm());
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn, 'Rm, 'Cond";
+  const char *form_test = "'Rd, 'CInv";
+  const char *form_update = "'Rd, 'Rn, 'CInv";
+
+  Condition cond = static_cast<Condition>(instr->Condition());
+  bool invertible_cond = (cond != al) && (cond != nv);
+
+  switch (instr->Mask(ConditionalSelectMask)) {
+    case CSEL_w:
+    case CSEL_x: mnemonic = "csel"; break;
+    case CSINC_w:
+    case CSINC_x: {
+      mnemonic = "csinc";
+      if (rnm_is_zr && invertible_cond) {
+        mnemonic = "cset";
+        form = form_test;
+      } else if (rn_is_rm && invertible_cond) {
+        mnemonic = "cinc";
+        form = form_update;
+      }
+      break;
+    }
+    case CSINV_w:
+    case CSINV_x: {
+      mnemonic = "csinv";
+      if (rnm_is_zr && invertible_cond) {
+        mnemonic = "csetm";
+        form = form_test;
+      } else if (rn_is_rm && invertible_cond) {
+        mnemonic = "cinv";
+        form = form_update;
+      }
+      break;
+    }
+    case CSNEG_w:
+    case CSNEG_x: {
+      mnemonic = "csneg";
+      if (rn_is_rm && invertible_cond) {
+        mnemonic = "cneg";
+        form = form_update;
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitBitfield(Instruction* instr) {
+  unsigned s = instr->ImmS();
+  unsigned r = instr->ImmR();
+  unsigned rd_size_minus_1 =
+    ((instr->SixtyFourBits() == 1) ? kXRegSize : kWRegSize) - 1;
+  const char *mnemonic = "";
+  const char *form = "";
+  const char *form_shift_right = "'Rd, 'Rn, 'IBr";
+  const char *form_extend = "'Rd, 'Wn";
+  const char *form_bfiz = "'Rd, 'Rn, 'IBZ-r, 'IBs+1";
+  const char *form_bfx = "'Rd, 'Rn, 'IBr, 'IBs-r+1";
+  const char *form_lsl = "'Rd, 'Rn, 'IBZ-r";
+
+  switch (instr->Mask(BitfieldMask)) {
+    case SBFM_w:
+    case SBFM_x: {
+      mnemonic = "sbfx";
+      form = form_bfx;
+      if (r == 0) {
+        form = form_extend;
+        if (s == 7) {
+          mnemonic = "sxtb";
+        } else if (s == 15) {
+          mnemonic = "sxth";
+        } else if ((s == 31) && (instr->SixtyFourBits() == 1)) {
+          mnemonic = "sxtw";
+        } else {
+          form = form_bfx;
+        }
+      } else if (s == rd_size_minus_1) {
+        mnemonic = "asr";
+        form = form_shift_right;
+      } else if (s < r) {
+        mnemonic = "sbfiz";
+        form = form_bfiz;
+      }
+      break;
+    }
+    case UBFM_w:
+    case UBFM_x: {
+      mnemonic = "ubfx";
+      form = form_bfx;
+      if (r == 0) {
+        form = form_extend;
+        if (s == 7) {
+          mnemonic = "uxtb";
+        } else if (s == 15) {
+          mnemonic = "uxth";
+        } else {
+          form = form_bfx;
+        }
+      }
+      if (s == rd_size_minus_1) {
+        mnemonic = "lsr";
+        form = form_shift_right;
+      } else if (r == s + 1) {
+        mnemonic = "lsl";
+        form = form_lsl;
+      } else if (s < r) {
+        mnemonic = "ubfiz";
+        form = form_bfiz;
+      }
+      break;
+    }
+    case BFM_w:
+    case BFM_x: {
+      mnemonic = "bfxil";
+      form = form_bfx;
+      if (s < r) {
+        mnemonic = "bfi";
+        form = form_bfiz;
+      }
+    }
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitExtract(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn, 'Rm, 'IExtract";
+
+  switch (instr->Mask(ExtractMask)) {
+    case EXTR_w:
+    case EXTR_x: {
+      if (instr->Rn() == instr->Rm()) {
+        mnemonic = "ror";
+        form = "'Rd, 'Rn, 'IExtract";
+      } else {
+        mnemonic = "extr";
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitPCRelAddressing(Instruction* instr) {
+  switch (instr->Mask(PCRelAddressingMask)) {
+    case ADR: Format(instr, "adr", "'Xd, 'AddrPCRelByte"); break;
+    // ADRP is not implemented.
+    default: Format(instr, "unimplemented", "(PCRelAddressing)");
+  }
+}
+
+
+void Disassembler::VisitConditionalBranch(Instruction* instr) {
+  switch (instr->Mask(ConditionalBranchMask)) {
+    case B_cond: Format(instr, "b.'CBrn", "'BImmCond"); break;
+    default: UNREACHABLE();
+  }
+}
+
+
+void Disassembler::VisitUnconditionalBranchToRegister(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'Xn";
+
+  switch (instr->Mask(UnconditionalBranchToRegisterMask)) {
+    case BR: mnemonic = "br"; break;
+    case BLR: mnemonic = "blr"; break;
+    case RET: {
+      mnemonic = "ret";
+      if (instr->Rn() == kLinkRegCode) {
+        form = NULL;
+      }
+      break;
+    }
+    default: form = "(UnconditionalBranchToRegister)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitUnconditionalBranch(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'BImmUncn";
+
+  switch (instr->Mask(UnconditionalBranchMask)) {
+    case B: mnemonic = "b"; break;
+    case BL: mnemonic = "bl"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitDataProcessing1Source(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Rn";
+
+  switch (instr->Mask(DataProcessing1SourceMask)) {
+    #define FORMAT(A, B)  \
+    case A##_w:           \
+    case A##_x: mnemonic = B; break;
+    FORMAT(RBIT, "rbit");
+    FORMAT(REV16, "rev16");
+    FORMAT(REV, "rev");
+    FORMAT(CLZ, "clz");
+    FORMAT(CLS, "cls");
+    #undef FORMAT
+    case REV32_x: mnemonic = "rev32"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitDataProcessing2Source(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'Rd, 'Rn, 'Rm";
+
+  switch (instr->Mask(DataProcessing2SourceMask)) {
+    #define FORMAT(A, B)  \
+    case A##_w:           \
+    case A##_x: mnemonic = B; break;
+    FORMAT(UDIV, "udiv");
+    FORMAT(SDIV, "sdiv");
+    FORMAT(LSLV, "lsl");
+    FORMAT(LSRV, "lsr");
+    FORMAT(ASRV, "asr");
+    FORMAT(RORV, "ror");
+    #undef FORMAT
+    default: form = "(DataProcessing2Source)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitDataProcessing3Source(Instruction* instr) {
+  bool ra_is_zr = RaIsZROrSP(instr);
+  const char *mnemonic = "";
+  const char *form = "'Xd, 'Wn, 'Wm, 'Xa";
+  const char *form_rrr = "'Rd, 'Rn, 'Rm";
+  const char *form_rrrr = "'Rd, 'Rn, 'Rm, 'Ra";
+  const char *form_xww = "'Xd, 'Wn, 'Wm";
+  const char *form_xxx = "'Xd, 'Xn, 'Xm";
+
+  switch (instr->Mask(DataProcessing3SourceMask)) {
+    case MADD_w:
+    case MADD_x: {
+      mnemonic = "madd";
+      form = form_rrrr;
+      if (ra_is_zr) {
+        mnemonic = "mul";
+        form = form_rrr;
+      }
+      break;
+    }
+    case MSUB_w:
+    case MSUB_x: {
+      mnemonic = "msub";
+      form = form_rrrr;
+      if (ra_is_zr) {
+        mnemonic = "mneg";
+        form = form_rrr;
+      }
+      break;
+    }
+    case SMADDL_x: {
+      mnemonic = "smaddl";
+      if (ra_is_zr) {
+        mnemonic = "smull";
+        form = form_xww;
+      }
+      break;
+    }
+    case SMSUBL_x: {
+      mnemonic = "smsubl";
+      if (ra_is_zr) {
+        mnemonic = "smnegl";
+        form = form_xww;
+      }
+      break;
+    }
+    case UMADDL_x: {
+      mnemonic = "umaddl";
+      if (ra_is_zr) {
+        mnemonic = "umull";
+        form = form_xww;
+      }
+      break;
+    }
+    case UMSUBL_x: {
+      mnemonic = "umsubl";
+      if (ra_is_zr) {
+        mnemonic = "umnegl";
+        form = form_xww;
+      }
+      break;
+    }
+    case SMULH_x: {
+      mnemonic = "smulh";
+      form = form_xxx;
+      break;
+    }
+    case UMULH_x: {
+      mnemonic = "umulh";
+      form = form_xxx;
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitCompareBranch(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rt, 'BImmCmpa";
+
+  switch (instr->Mask(CompareBranchMask)) {
+    case CBZ_w:
+    case CBZ_x: mnemonic = "cbz"; break;
+    case CBNZ_w:
+    case CBNZ_x: mnemonic = "cbnz"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitTestBranch(Instruction* instr) {
+  const char *mnemonic = "";
+  // If the top bit of the immediate is clear, the tested register is
+  // disassembled as Wt, otherwise Xt. As the top bit of the immediate is
+  // encoded in bit 31 of the instruction, we can reuse the Rt form, which
+  // uses bit 31 (normally "sf") to choose the register size.
+  const char *form = "'Rt, 'IS, 'BImmTest";
+
+  switch (instr->Mask(TestBranchMask)) {
+    case TBZ: mnemonic = "tbz"; break;
+    case TBNZ: mnemonic = "tbnz"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitMoveWideImmediate(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'IMoveImm";
+
+  // Print the shift separately for movk, to make it clear which half word will
+  // be overwritten. Movn and movz print the computed immediate, which includes
+  // shift calculation.
+  switch (instr->Mask(MoveWideImmediateMask)) {
+    case MOVN_w:
+    case MOVN_x: mnemonic = "movn"; break;
+    case MOVZ_w:
+    case MOVZ_x: mnemonic = "movz"; break;
+    case MOVK_w:
+    case MOVK_x: mnemonic = "movk"; form = "'Rd, 'IMoveLSL"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+#define LOAD_STORE_LIST(V)    \
+  V(STRB_w, "strb", "'Wt")    \
+  V(STRH_w, "strh", "'Wt")    \
+  V(STR_w, "str", "'Wt")      \
+  V(STR_x, "str", "'Xt")      \
+  V(LDRB_w, "ldrb", "'Wt")    \
+  V(LDRH_w, "ldrh", "'Wt")    \
+  V(LDR_w, "ldr", "'Wt")      \
+  V(LDR_x, "ldr", "'Xt")      \
+  V(LDRSB_x, "ldrsb", "'Xt")  \
+  V(LDRSH_x, "ldrsh", "'Xt")  \
+  V(LDRSW_x, "ldrsw", "'Xt")  \
+  V(LDRSB_w, "ldrsb", "'Wt")  \
+  V(LDRSH_w, "ldrsh", "'Wt")  \
+  V(STR_s, "str", "'St")      \
+  V(STR_d, "str", "'Dt")      \
+  V(LDR_s, "ldr", "'St")      \
+  V(LDR_d, "ldr", "'Dt")
+
+void Disassembler::VisitLoadStorePreIndex(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStorePreIndex)";
+
+  switch (instr->Mask(LoadStorePreIndexMask)) {
+    #define LS_PREINDEX(A, B, C) \
+    case A##_pre: mnemonic = B; form = C ", ['Xns'ILS]!"; break;
+    LOAD_STORE_LIST(LS_PREINDEX)
+    #undef LS_PREINDEX
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStorePostIndex(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStorePostIndex)";
+
+  switch (instr->Mask(LoadStorePostIndexMask)) {
+    #define LS_POSTINDEX(A, B, C) \
+    case A##_post: mnemonic = B; form = C ", ['Xns]'ILS"; break;
+    LOAD_STORE_LIST(LS_POSTINDEX)
+    #undef LS_POSTINDEX
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStoreUnsignedOffset(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStoreUnsignedOffset)";
+
+  switch (instr->Mask(LoadStoreUnsignedOffsetMask)) {
+    #define LS_UNSIGNEDOFFSET(A, B, C) \
+    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]";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStoreRegisterOffset(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStoreRegisterOffset)";
+
+  switch (instr->Mask(LoadStoreRegisterOffsetMask)) {
+    #define LS_REGISTEROFFSET(A, B, C) \
+    case A##_reg: mnemonic = B; form = C ", ['Xns, 'Offsetreg]"; break;
+    LOAD_STORE_LIST(LS_REGISTEROFFSET)
+    #undef LS_REGISTEROFFSET
+    case PRFM_reg: mnemonic = "prfm"; form = "'PrefOp, ['Xns, 'Offsetreg]";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStoreUnscaledOffset(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'Wt, ['Xns'ILS]";
+  const char *form_x = "'Xt, ['Xns'ILS]";
+  const char *form_s = "'St, ['Xns'ILS]";
+  const char *form_d = "'Dt, ['Xns'ILS]";
+
+  switch (instr->Mask(LoadStoreUnscaledOffsetMask)) {
+    case STURB_w:  mnemonic = "sturb"; break;
+    case STURH_w:  mnemonic = "sturh"; break;
+    case STUR_w:   mnemonic = "stur"; break;
+    case STUR_x:   mnemonic = "stur"; form = form_x; break;
+    case STUR_s:   mnemonic = "stur"; form = form_s; break;
+    case STUR_d:   mnemonic = "stur"; form = form_d; break;
+    case LDURB_w:  mnemonic = "ldurb"; break;
+    case LDURH_w:  mnemonic = "ldurh"; break;
+    case LDUR_w:   mnemonic = "ldur"; break;
+    case LDUR_x:   mnemonic = "ldur"; form = form_x; break;
+    case LDUR_s:   mnemonic = "ldur"; form = form_s; break;
+    case LDUR_d:   mnemonic = "ldur"; form = form_d; break;
+    case LDURSB_x: form = form_x;  // Fall through.
+    case LDURSB_w: mnemonic = "ldursb"; break;
+    case LDURSH_x: form = form_x;  // Fall through.
+    case LDURSH_w: mnemonic = "ldursh"; break;
+    case LDURSW_x: mnemonic = "ldursw"; form = form_x; break;
+    default: form = "(LoadStoreUnscaledOffset)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadLiteral(Instruction* instr) {
+  const char *mnemonic = "ldr";
+  const char *form = "(LoadLiteral)";
+
+  switch (instr->Mask(LoadLiteralMask)) {
+    case LDR_w_lit: form = "'Wt, 'ILLiteral 'LValue"; break;
+    case LDR_x_lit: form = "'Xt, 'ILLiteral 'LValue"; break;
+    case LDR_s_lit: form = "'St, 'ILLiteral 'LValue"; break;
+    case LDR_d_lit: form = "'Dt, 'ILLiteral 'LValue"; break;
+    default: mnemonic = "unimplemented";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+#define LOAD_STORE_PAIR_LIST(V)         \
+  V(STP_w, "stp", "'Wt, 'Wt2", "4")     \
+  V(LDP_w, "ldp", "'Wt, 'Wt2", "4")     \
+  V(LDPSW_x, "ldpsw", "'Xt, 'Xt2", "4") \
+  V(STP_x, "stp", "'Xt, 'Xt2", "8")     \
+  V(LDP_x, "ldp", "'Xt, 'Xt2", "8")     \
+  V(STP_s, "stp", "'St, 'St2", "4")     \
+  V(LDP_s, "ldp", "'St, 'St2", "4")     \
+  V(STP_d, "stp", "'Dt, 'Dt2", "8")     \
+  V(LDP_d, "ldp", "'Dt, 'Dt2", "8")
+
+void Disassembler::VisitLoadStorePairPostIndex(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStorePairPostIndex)";
+
+  switch (instr->Mask(LoadStorePairPostIndexMask)) {
+    #define LSP_POSTINDEX(A, B, C, D) \
+    case A##_post: mnemonic = B; form = C ", ['Xns]'ILP" D; break;
+    LOAD_STORE_PAIR_LIST(LSP_POSTINDEX)
+    #undef LSP_POSTINDEX
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStorePairPreIndex(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStorePairPreIndex)";
+
+  switch (instr->Mask(LoadStorePairPreIndexMask)) {
+    #define LSP_PREINDEX(A, B, C, D) \
+    case A##_pre: mnemonic = B; form = C ", ['Xns'ILP" D "]!"; break;
+    LOAD_STORE_PAIR_LIST(LSP_PREINDEX)
+    #undef LSP_PREINDEX
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStorePairOffset(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(LoadStorePairOffset)";
+
+  switch (instr->Mask(LoadStorePairOffsetMask)) {
+    #define LSP_OFFSET(A, B, C, D) \
+    case A##_off: mnemonic = B; form = C ", ['Xns'ILP" D "]"; break;
+    LOAD_STORE_PAIR_LIST(LSP_OFFSET)
+    #undef LSP_OFFSET
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitLoadStorePairNonTemporal(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form;
+
+  switch (instr->Mask(LoadStorePairNonTemporalMask)) {
+    case STNP_w: mnemonic = "stnp"; form = "'Wt, 'Wt2, ['Xns'ILP4]"; break;
+    case LDNP_w: mnemonic = "ldnp"; form = "'Wt, 'Wt2, ['Xns'ILP4]"; break;
+    case STNP_x: mnemonic = "stnp"; form = "'Xt, 'Xt2, ['Xns'ILP8]"; break;
+    case LDNP_x: mnemonic = "ldnp"; form = "'Xt, 'Xt2, ['Xns'ILP8]"; break;
+    case STNP_s: mnemonic = "stnp"; form = "'St, 'St2, ['Xns'ILP4]"; break;
+    case LDNP_s: mnemonic = "ldnp"; form = "'St, 'St2, ['Xns'ILP4]"; break;
+    case STNP_d: mnemonic = "stnp"; form = "'Dt, 'Dt2, ['Xns'ILP8]"; break;
+    case LDNP_d: mnemonic = "ldnp"; form = "'Dt, 'Dt2, ['Xns'ILP8]"; break;
+    default: form = "(LoadStorePairNonTemporal)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPCompare(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'Fn, 'Fm";
+  const char *form_zero = "'Fn, #0.0";
+
+  switch (instr->Mask(FPCompareMask)) {
+    case FCMP_s_zero:
+    case FCMP_d_zero: form = form_zero;  // Fall through.
+    case FCMP_s:
+    case FCMP_d: mnemonic = "fcmp"; break;
+    default: form = "(FPCompare)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPConditionalCompare(Instruction* instr) {
+  const char *mnemonic = "unmplemented";
+  const char *form = "'Fn, 'Fm, 'INzcv, 'Cond";
+
+  switch (instr->Mask(FPConditionalCompareMask)) {
+    case FCCMP_s:
+    case FCCMP_d: mnemonic = "fccmp"; break;
+    case FCCMPE_s:
+    case FCCMPE_d: mnemonic = "fccmpe"; break;
+    default: form = "(FPConditionalCompare)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPConditionalSelect(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Fd, 'Fn, 'Fm, 'Cond";
+
+  switch (instr->Mask(FPConditionalSelectMask)) {
+    case FCSEL_s:
+    case FCSEL_d: mnemonic = "fcsel"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPDataProcessing1Source(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'Fd, 'Fn";
+
+  switch (instr->Mask(FPDataProcessing1SourceMask)) {
+    #define FORMAT(A, B)  \
+    case A##_s:           \
+    case A##_d: mnemonic = B; break;
+    FORMAT(FMOV, "fmov");
+    FORMAT(FABS, "fabs");
+    FORMAT(FNEG, "fneg");
+    FORMAT(FSQRT, "fsqrt");
+    FORMAT(FRINTN, "frintn");
+    FORMAT(FRINTP, "frintp");
+    FORMAT(FRINTM, "frintm");
+    FORMAT(FRINTZ, "frintz");
+    FORMAT(FRINTA, "frinta");
+    FORMAT(FRINTX, "frintx");
+    FORMAT(FRINTI, "frinti");
+    #undef FORMAT
+    case FCVT_ds: mnemonic = "fcvt"; form = "'Dd, 'Sn"; break;
+    case FCVT_sd: mnemonic = "fcvt"; form = "'Sd, 'Dn"; break;
+    default: form = "(FPDataProcessing1Source)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPDataProcessing2Source(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Fd, 'Fn, 'Fm";
+
+  switch (instr->Mask(FPDataProcessing2SourceMask)) {
+    #define FORMAT(A, B)  \
+    case A##_s:           \
+    case A##_d: mnemonic = B; break;
+    FORMAT(FMUL, "fmul");
+    FORMAT(FDIV, "fdiv");
+    FORMAT(FADD, "fadd");
+    FORMAT(FSUB, "fsub");
+    FORMAT(FMAX, "fmax");
+    FORMAT(FMIN, "fmin");
+    FORMAT(FMAXNM, "fmaxnm");
+    FORMAT(FMINNM, "fminnm");
+    FORMAT(FNMUL, "fnmul");
+    #undef FORMAT
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPDataProcessing3Source(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Fd, 'Fn, 'Fm, 'Fa";
+
+  switch (instr->Mask(FPDataProcessing3SourceMask)) {
+    #define FORMAT(A, B)  \
+    case A##_s:           \
+    case A##_d: mnemonic = B; break;
+    FORMAT(FMADD, "fmadd");
+    FORMAT(FMSUB, "fmsub");
+    FORMAT(FNMADD, "fnmadd");
+    FORMAT(FNMSUB, "fnmsub");
+    #undef FORMAT
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPImmediate(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "(FPImmediate)";
+
+  switch (instr->Mask(FPImmediateMask)) {
+    case FMOV_s_imm: mnemonic = "fmov"; form = "'Sd, 'IFPSingle"; break;
+    case FMOV_d_imm: mnemonic = "fmov"; form = "'Dd, 'IFPDouble"; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPIntegerConvert(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "(FPIntegerConvert)";
+  const char *form_rf = "'Rd, 'Fn";
+  const char *form_fr = "'Fd, 'Rn";
+
+  switch (instr->Mask(FPIntegerConvertMask)) {
+    case FMOV_ws:
+    case FMOV_xd: mnemonic = "fmov"; form = form_rf; break;
+    case FMOV_sw:
+    case FMOV_dx: mnemonic = "fmov"; form = form_fr; break;
+    case FCVTMS_ws:
+    case FCVTMS_xs:
+    case FCVTMS_wd:
+    case FCVTMS_xd: mnemonic = "fcvtms"; form = form_rf; break;
+    case FCVTMU_ws:
+    case FCVTMU_xs:
+    case FCVTMU_wd:
+    case FCVTMU_xd: mnemonic = "fcvtmu"; form = form_rf; break;
+    case FCVTNS_ws:
+    case FCVTNS_xs:
+    case FCVTNS_wd:
+    case FCVTNS_xd: mnemonic = "fcvtns"; form = form_rf; break;
+    case FCVTNU_ws:
+    case FCVTNU_xs:
+    case FCVTNU_wd:
+    case FCVTNU_xd: mnemonic = "fcvtnu"; form = form_rf; break;
+    case FCVTZU_xd:
+    case FCVTZU_ws:
+    case FCVTZU_wd:
+    case FCVTZU_xs: mnemonic = "fcvtzu"; form = form_rf; break;
+    case FCVTZS_xd:
+    case FCVTZS_wd:
+    case FCVTZS_xs:
+    case FCVTZS_ws: mnemonic = "fcvtzs"; form = form_rf; break;
+    case SCVTF_sw:
+    case SCVTF_sx:
+    case SCVTF_dw:
+    case SCVTF_dx: mnemonic = "scvtf"; form = form_fr; break;
+    case UCVTF_sw:
+    case UCVTF_sx:
+    case UCVTF_dw:
+    case UCVTF_dx: mnemonic = "ucvtf"; form = form_fr; break;
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitFPFixedPointConvert(Instruction* instr) {
+  const char *mnemonic = "";
+  const char *form = "'Rd, 'Fn, 'IFPFBits";
+  const char *form_fr = "'Fd, 'Rn, 'IFPFBits";
+
+  switch (instr->Mask(FPFixedPointConvertMask)) {
+    case FCVTZS_ws_fixed:
+    case FCVTZS_xs_fixed:
+    case FCVTZS_wd_fixed:
+    case FCVTZS_xd_fixed: mnemonic = "fcvtzs"; break;
+    case FCVTZU_ws_fixed:
+    case FCVTZU_xs_fixed:
+    case FCVTZU_wd_fixed:
+    case FCVTZU_xd_fixed: mnemonic = "fcvtzu"; break;
+    case SCVTF_sw_fixed:
+    case SCVTF_sx_fixed:
+    case SCVTF_dw_fixed:
+    case SCVTF_dx_fixed: mnemonic = "scvtf"; form = form_fr; break;
+    case UCVTF_sw_fixed:
+    case UCVTF_sx_fixed:
+    case UCVTF_dw_fixed:
+    case UCVTF_dx_fixed: mnemonic = "ucvtf"; form = form_fr; break;
+    default: UNREACHABLE();
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitSystem(Instruction* instr) {
+  // Some system instructions hijack their Op and Cp fields to represent a
+  // range of immediates instead of indicating a different instruction. This
+  // makes the decoding tricky.
+  const char *mnemonic = "unimplemented";
+  const char *form = "(System)";
+
+  if (instr->Mask(SystemSysRegFMask) == SystemSysRegFixed) {
+    switch (instr->Mask(SystemSysRegMask)) {
+      case MRS: {
+        mnemonic = "mrs";
+        switch (instr->ImmSystemRegister()) {
+          case NZCV: form = "'Xt, nzcv"; break;
+          case FPCR: form = "'Xt, fpcr"; break;
+          default: form = "'Xt, (unknown)"; break;
+        }
+        break;
+      }
+      case MSR: {
+        mnemonic = "msr";
+        switch (instr->ImmSystemRegister()) {
+          case NZCV: form = "nzcv, 'Xt"; break;
+          case FPCR: form = "fpcr, 'Xt"; break;
+          default: form = "(unknown), 'Xt"; break;
+        }
+        break;
+      }
+    }
+  } else if (instr->Mask(SystemHintFMask) == SystemHintFixed) {
+    ASSERT(instr->Mask(SystemHintMask) == HINT);
+    switch (instr->ImmHint()) {
+      case NOP: {
+        mnemonic = "nop";
+        form = NULL;
+        break;
+      }
+    }
+  }
+
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitException(Instruction* instr) {
+  const char *mnemonic = "unimplemented";
+  const char *form = "'IDebug";
+
+  switch (instr->Mask(ExceptionMask)) {
+    case HLT: mnemonic = "hlt"; break;
+    case BRK: mnemonic = "brk"; break;
+    case SVC: mnemonic = "svc"; break;
+    case HVC: mnemonic = "hvc"; break;
+    case SMC: mnemonic = "smc"; break;
+    case DCPS1: mnemonic = "dcps1"; form = "{'IDebug}"; break;
+    case DCPS2: mnemonic = "dcps2"; form = "{'IDebug}"; break;
+    case DCPS3: mnemonic = "dcps3"; form = "{'IDebug}"; break;
+    default: form = "(Exception)";
+  }
+  Format(instr, mnemonic, form);
+}
+
+
+void Disassembler::VisitUnimplemented(Instruction* instr) {
+  Format(instr, "unimplemented", "(Unimplemented)");
+}
+
+
+void Disassembler::VisitUnallocated(Instruction* instr) {
+  Format(instr, "unallocated", "(Unallocated)");
+}
+
+
+void Disassembler::ProcessOutput(Instruction* /*instr*/) {
+  // The base disasm does nothing more than disassembling into a buffer.
+}
+
+
+void Disassembler::Format(Instruction* instr, const char* mnemonic,
+                          const char* format) {
+  ASSERT(mnemonic != NULL);
+  ResetOutput();
+  Substitute(instr, mnemonic);
+  if (format != NULL) {
+    buffer_[buffer_pos_++] = ' ';
+    Substitute(instr, format);
+  }
+  buffer_[buffer_pos_] = 0;
+  ProcessOutput(instr);
+}
+
+
+void Disassembler::Substitute(Instruction* instr, const char* string) {
+  char chr = *string++;
+  while (chr != '\0') {
+    if (chr == '\'') {
+      string += SubstituteField(instr, string);
+    } else {
+      buffer_[buffer_pos_++] = chr;
+    }
+    chr = *string++;
+  }
+}
+
+
+int Disassembler::SubstituteField(Instruction* instr, const char* format) {
+  switch (format[0]) {
+    case 'R':  // Register. X or W, selected by sf bit.
+    case 'F':  // FP Register. S or D, selected by type field.
+    case 'W':
+    case 'X':
+    case 'S':
+    case 'D': return SubstituteRegisterField(instr, format);
+    case 'I': return SubstituteImmediateField(instr, format);
+    case 'L': return SubstituteLiteralField(instr, format);
+    case 'H': return SubstituteShiftField(instr, format);
+    case 'P': return SubstitutePrefetchField(instr, format);
+    case 'C': return SubstituteConditionField(instr, format);
+    case 'E': return SubstituteExtendField(instr, format);
+    case 'A': return SubstitutePCRelAddressField(instr, format);
+    case 'B': return SubstituteBranchTargetField(instr, format);
+    case 'O': return SubstituteLSRegOffsetField(instr, format);
+    default: {
+      UNREACHABLE();
+      return 1;
+    }
+  }
+}
+
+
+int Disassembler::SubstituteRegisterField(Instruction* instr,
+                                          const char* format) {
+  unsigned reg_num = 0;
+  unsigned field_len = 2;
+  switch (format[1]) {
+    case 'd': reg_num = instr->Rd(); break;
+    case 'n': reg_num = instr->Rn(); break;
+    case 'm': reg_num = instr->Rm(); break;
+    case 'a': reg_num = instr->Ra(); break;
+    case 't': {
+      if (format[2] == '2') {
+        reg_num = instr->Rt2();
+        field_len = 3;
+      } else {
+        reg_num = instr->Rt();
+      }
+      break;
+    }
+    default: UNREACHABLE();
+  }
+
+  // Increase field length for registers tagged as stack.
+  if (format[2] == 's') {
+    field_len = 3;
+  }
+
+  char reg_type;
+  if (format[0] == 'R') {
+    // Register type is R: use sf bit to choose X and W.
+    reg_type = instr->SixtyFourBits() ? 'x' : 'w';
+  } else if (format[0] == 'F') {
+    // Floating-point register: use type field to choose S or D.
+    reg_type = ((instr->FPType() & 1) == 0) ? 's' : 'd';
+  } else {
+    // Register type is specified. Make it lower case.
+    reg_type = format[0] + 0x20;
+  }
+
+  if ((reg_num != kZeroRegCode) || (reg_type == 's') || (reg_type == 'd')) {
+    // A normal register: w0 - w30, x0 - x30, s0 - s31, d0 - d31.
+    AppendToOutput("%c%d", reg_type, reg_num);
+  } else if (format[2] == 's') {
+    // Disassemble w31/x31 as stack pointer wsp/sp.
+    AppendToOutput("%s", (reg_type == 'w') ? "wsp" : "sp");
+  } else {
+    // Disassemble w31/x31 as zero register wzr/xzr.
+    AppendToOutput("%czr", reg_type);
+  }
+
+  return field_len;
+}
+
+
+int Disassembler::SubstituteImmediateField(Instruction* instr,
+                                           const char* format) {
+  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 {
+        ASSERT(format[5] == 'L');
+        AppendToOutput("#0x%" PRIx64, instr->ImmMoveWide());
+        if (instr->ShiftMoveWide() > 0) {
+          AppendToOutput(", lsl #%d", 16 * instr->ShiftMoveWide());
+        }
+      }
+      return 8;
+    }
+    case 'L': {
+      switch (format[2]) {
+        case 'L': {  // ILLiteral - Immediate Load Literal.
+          AppendToOutput("pc%+" PRId64,
+                         instr->ImmLLiteral() << kLiteralEntrySizeLog2);
+          return 9;
+        }
+        case 'S': {  // ILS - Immediate Load/Store.
+          if (instr->ImmLS() != 0) {
+            AppendToOutput(", #%" PRId64, instr->ImmLS());
+          }
+          return 3;
+        }
+        case 'P': {  // ILPx - Immediate Load/Store Pair, x = access size.
+          if (instr->ImmLSPair() != 0) {
+            // format[3] is the scale value. Convert to a number.
+            int scale = format[3] - 0x30;
+            AppendToOutput(", #%" PRId64, instr->ImmLSPair() * scale);
+          }
+          return 4;
+        }
+        case 'U': {  // ILU - Immediate Load/Store Unsigned.
+          if (instr->ImmLSUnsigned() != 0) {
+            AppendToOutput(", #%" PRIu64,
+                           instr->ImmLSUnsigned() << instr->SizeLS());
+          }
+          return 3;
+        }
+      }
+    }
+    case 'C': {  // ICondB - Immediate Conditional Branch.
+      int64_t offset = instr->ImmCondBranch() << 2;
+      char sign = (offset >= 0) ? '+' : '-';
+      AppendToOutput("#%c0x%" PRIx64, sign, offset);
+      return 6;
+    }
+    case 'A': {  // IAddSub.
+      ASSERT(instr->ShiftAddSub() <= 1);
+      int64_t imm = instr->ImmAddSub() << (12 * instr->ShiftAddSub());
+      AppendToOutput("#0x%" PRIx64 " (%" PRId64 ")", imm, imm);
+      return 7;
+    }
+    case 'F': {  // IFPSingle, IFPDouble or IFPFBits.
+      if (format[3] == 'F') {  // IFPFbits.
+        AppendToOutput("#%d", 64 - instr->FPScale());
+        return 8;
+      } else {
+        AppendToOutput("#0x%" PRIx64 " (%.4f)", instr->ImmFP(),
+                       format[3] == 'S' ? instr->ImmFP32() : instr->ImmFP64());
+        return 9;
+      }
+    }
+    case 'T': {  // ITri - Immediate Triangular Encoded.
+      AppendToOutput("#0x%" PRIx64, instr->ImmLogical());
+      return 4;
+    }
+    case 'N': {  // INzcv.
+      int nzcv = (instr->Nzcv() << Flags_offset);
+      AppendToOutput("#%c%c%c%c", ((nzcv & NFlag) == 0) ? 'n' : 'N',
+                                  ((nzcv & ZFlag) == 0) ? 'z' : 'Z',
+                                  ((nzcv & CFlag) == 0) ? 'c' : 'C',
+                                  ((nzcv & VFlag) == 0) ? 'v' : 'V');
+      return 5;
+    }
+    case 'P': {  // IP - Conditional compare.
+      AppendToOutput("#%d", instr->ImmCondCmp());
+      return 2;
+    }
+    case 'B': {  // Bitfields.
+      return SubstituteBitfieldImmediateField(instr, format);
+    }
+    case 'E': {  // IExtract.
+      AppendToOutput("#%d", instr->ImmS());
+      return 8;
+    }
+    case 'S': {  // IS - Test and branch bit.
+      AppendToOutput("#%d", (instr->ImmTestBranchBit5() << 5) |
+                            instr->ImmTestBranchBit40());
+      return 2;
+    }
+    case 'D': {  // IDebug - HLT and BRK instructions.
+      AppendToOutput("#0x%x", instr->ImmException());
+      return 6;
+    }
+    default: {
+      UNIMPLEMENTED();
+      return 0;
+    }
+  }
+}
+
+
+int Disassembler::SubstituteBitfieldImmediateField(Instruction* instr,
+                                                   const char* format) {
+  ASSERT((format[0] == 'I') && (format[1] == 'B'));
+  unsigned r = instr->ImmR();
+  unsigned s = instr->ImmS();
+
+  switch (format[2]) {
+    case 'r': {  // IBr.
+      AppendToOutput("#%d", r);
+      return 3;
+    }
+    case 's': {  // IBs+1 or IBs-r+1.
+      if (format[3] == '+') {
+        AppendToOutput("#%d", s + 1);
+        return 5;
+      } else {
+        ASSERT(format[3] == '-');
+        AppendToOutput("#%d", s - r + 1);
+        return 7;
+      }
+    }
+    case 'Z': {  // IBZ-r.
+      ASSERT((format[3] == '-') && (format[4] == 'r'));
+      unsigned reg_size = (instr->SixtyFourBits() == 1) ? kXRegSize : kWRegSize;
+      AppendToOutput("#%d", reg_size - r);
+      return 5;
+    }
+    default: {
+      UNREACHABLE();
+      return 0;
+    }
+  }
+}
+
+
+int Disassembler::SubstituteLiteralField(Instruction* instr,
+                                         const char* format) {
+  ASSERT(strncmp(format, "LValue", 6) == 0);
+  USE(format);
+
+  switch (instr->Mask(LoadLiteralMask)) {
+    case LDR_w_lit:
+    case LDR_x_lit:
+    case LDR_s_lit:
+    case LDR_d_lit: AppendToOutput("(addr %p)", instr->LiteralAddress()); break;
+    default: UNREACHABLE();
+  }
+
+  return 6;
+}
+
+
+int Disassembler::SubstituteShiftField(Instruction* instr, const char* format) {
+  ASSERT(format[0] == 'H');
+  ASSERT(instr->ShiftDP() <= 0x3);
+
+  switch (format[1]) {
+    case 'D': {  // HDP.
+      ASSERT(instr->ShiftDP() != ROR);
+    }  // Fall through.
+    case 'L': {  // HLo.
+      if (instr->ImmDPShift() != 0) {
+        const char* shift_type[] = {"lsl", "lsr", "asr", "ror"};
+        AppendToOutput(", %s #%" PRId64, shift_type[instr->ShiftDP()],
+                       instr->ImmDPShift());
+      }
+      return 3;
+    }
+    default:
+      UNIMPLEMENTED();
+      return 0;
+  }
+}
+
+
+int Disassembler::SubstituteConditionField(Instruction* instr,
+                                           const char* format) {
+  ASSERT(format[0] == 'C');
+  const char* condition_code[] = { "eq", "ne", "hs", "lo",
+                                   "mi", "pl", "vs", "vc",
+                                   "hi", "ls", "ge", "lt",
+                                   "gt", "le", "al", "nv" };
+  int cond;
+  switch (format[1]) {
+    case 'B': cond = instr->ConditionBranch(); break;
+    case 'I': {
+      cond = InvertCondition(static_cast<Condition>(instr->Condition()));
+      break;
+    }
+    default: cond = instr->Condition();
+  }
+  AppendToOutput("%s", condition_code[cond]);
+  return 4;
+}
+
+
+int Disassembler::SubstitutePCRelAddressField(Instruction* instr,
+                                              const char* format) {
+  USE(format);
+  ASSERT(strncmp(format, "AddrPCRel", 9) == 0);
+
+  int offset = instr->ImmPCRel();
+
+  // Only ADR (AddrPCRelByte) is supported.
+  ASSERT(strcmp(format, "AddrPCRelByte") == 0);
+
+  char sign = '+';
+  if (offset < 0) {
+    offset = -offset;
+    sign = '-';
+  }
+  // TODO: Extend this to support printing the target address.
+  AppendToOutput("#%c0x%x", sign, offset);
+  return 13;
+}
+
+
+int Disassembler::SubstituteBranchTargetField(Instruction* instr,
+                                              const char* format) {
+  ASSERT(strncmp(format, "BImm", 4) == 0);
+
+  int64_t offset = 0;
+  switch (format[5]) {
+    // BImmUncn - unconditional branch immediate.
+    case 'n': offset = instr->ImmUncondBranch(); break;
+    // BImmCond - conditional branch immediate.
+    case 'o': offset = instr->ImmCondBranch(); break;
+    // BImmCmpa - compare and branch immediate.
+    case 'm': offset = instr->ImmCmpBranch(); break;
+    // BImmTest - test and branch immediate.
+    case 'e': offset = instr->ImmTestBranch(); break;
+    default: UNIMPLEMENTED();
+  }
+  offset <<= kInstructionSizeLog2;
+  char sign = '+';
+  if (offset < 0) {
+    offset = -offset;
+    sign = '-';
+  }
+  AppendToOutput("#%c0x%" PRIx64, sign, offset);
+  return 8;
+}
+
+
+int Disassembler::SubstituteExtendField(Instruction* instr,
+                                        const char* format) {
+  ASSERT(strncmp(format, "Ext", 3) == 0);
+  ASSERT(instr->ExtendMode() <= 7);
+  USE(format);
+
+  const char* extend_mode[] = { "uxtb", "uxth", "uxtw", "uxtx",
+                                "sxtb", "sxth", "sxtw", "sxtx" };
+
+  // If rd or rn is SP, uxtw on 32-bit registers and uxtx on 64-bit
+  // registers becomes lsl.
+  if (((instr->Rd() == kZeroRegCode) || (instr->Rn() == kZeroRegCode)) &&
+      (((instr->ExtendMode() == UXTW) && (instr->SixtyFourBits() == 0)) ||
+       (instr->ExtendMode() == UXTX))) {
+    if (instr->ImmExtendShift() > 0) {
+      AppendToOutput(", lsl #%d", instr->ImmExtendShift());
+    }
+  } else {
+    AppendToOutput(", %s", extend_mode[instr->ExtendMode()]);
+    if (instr->ImmExtendShift() > 0) {
+      AppendToOutput(" #%d", instr->ImmExtendShift());
+    }
+  }
+  return 3;
+}
+
+
+int Disassembler::SubstituteLSRegOffsetField(Instruction* instr,
+                                             const char* format) {
+  ASSERT(strncmp(format, "Offsetreg", 9) == 0);
+  const char* extend_mode[] = { "undefined", "undefined", "uxtw", "lsl",
+                                "undefined", "undefined", "sxtw", "sxtx" };
+  USE(format);
+
+  unsigned shift = instr->ImmShiftLS();
+  Extend ext = static_cast<Extend>(instr->ExtendMode());
+  char reg_type = ((ext == UXTW) || (ext == SXTW)) ? 'w' : 'x';
+
+  unsigned rm = instr->Rm();
+  if (rm == kZeroRegCode) {
+    AppendToOutput("%czr", reg_type);
+  } else {
+    AppendToOutput("%c%d", reg_type, rm);
+  }
+
+  // Extend mode UXTX is an alias for shift mode LSL here.
+  if (!((ext == UXTX) && (shift == 0))) {
+    AppendToOutput(", %s", extend_mode[ext]);
+    if (shift != 0) {
+      AppendToOutput(" #%d", instr->SizeLS());
+    }
+  }
+  return 9;
+}
+
+
+int Disassembler::SubstitutePrefetchField(Instruction* instr,
+                                          const char* format) {
+  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);
+  return 6;
+}
+
+
+void Disassembler::ResetOutput() {
+  buffer_pos_ = 0;
+  buffer_[buffer_pos_] = 0;
+}
+
+
+void Disassembler::AppendToOutput(const char* format, ...) {
+  va_list args;
+  va_start(args, format);
+  buffer_pos_ += vsnprintf(&buffer_[buffer_pos_], buffer_size_, format, args);
+  va_end(args);
+}
+
+
+void PrintDisassembler::ProcessOutput(Instruction* instr) {
+  fprintf(stream_, "0x%016" PRIx64 "  %08" PRIx32 "\t\t%s\n",
+          reinterpret_cast<uint64_t>(instr),
+          instr->InstructionBits(),
+          GetOutput());
+}
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/disasm-a64.h b/disas/libvixl/src/a64/disasm-a64.h
new file mode 100644
index 0000000..857a5ac
--- /dev/null
+++ b/disas/libvixl/src/a64/disasm-a64.h
@@ -0,0 +1,109 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_DISASM_A64_H
+#define VIXL_A64_DISASM_A64_H
+
+#include "globals.h"
+#include "utils.h"
+#include "instructions-a64.h"
+#include "decoder-a64.h"
+
+namespace vixl {
+
+class Disassembler: public DecoderVisitor {
+ public:
+  Disassembler();
+  Disassembler(char* text_buffer, int buffer_size);
+  virtual ~Disassembler();
+  char* GetOutput();
+
+  // Declare all Visitor functions.
+  #define DECLARE(A)  void Visit##A(Instruction* instr);
+  VISITOR_LIST(DECLARE)
+  #undef DECLARE
+
+ protected:
+  virtual void ProcessOutput(Instruction* instr);
+
+ private:
+  void Format(Instruction* instr, const char* mnemonic, const char* format);
+  void Substitute(Instruction* instr, const char* string);
+  int SubstituteField(Instruction* instr, const char* format);
+  int SubstituteRegisterField(Instruction* instr, const char* format);
+  int SubstituteImmediateField(Instruction* instr, const char* format);
+  int SubstituteLiteralField(Instruction* instr, const char* format);
+  int SubstituteBitfieldImmediateField(Instruction* instr, const char* format);
+  int SubstituteShiftField(Instruction* instr, const char* format);
+  int SubstituteExtendField(Instruction* instr, const char* format);
+  int SubstituteConditionField(Instruction* instr, const char* format);
+  int SubstitutePCRelAddressField(Instruction* instr, const char* format);
+  int SubstituteBranchTargetField(Instruction* instr, const char* format);
+  int SubstituteLSRegOffsetField(Instruction* instr, const char* format);
+  int SubstitutePrefetchField(Instruction* instr, const char* format);
+
+  inline bool RdIsZROrSP(Instruction* instr) const {
+    return (instr->Rd() == kZeroRegCode);
+  }
+
+  inline bool RnIsZROrSP(Instruction* instr) const {
+    return (instr->Rn() == kZeroRegCode);
+  }
+
+  inline bool RmIsZROrSP(Instruction* instr) const {
+    return (instr->Rm() == kZeroRegCode);
+  }
+
+  inline bool RaIsZROrSP(Instruction* instr) const {
+    return (instr->Ra() == kZeroRegCode);
+  }
+
+  bool IsMovzMovnImm(unsigned reg_size, uint64_t value);
+
+  void ResetOutput();
+  void AppendToOutput(const char* string, ...);
+
+  char* buffer_;
+  uint32_t buffer_pos_;
+  uint32_t buffer_size_;
+  bool own_buffer_;
+};
+
+
+class PrintDisassembler: public Disassembler {
+ public:
+  explicit PrintDisassembler(FILE* stream) : stream_(stream) { }
+  ~PrintDisassembler() { }
+
+ protected:
+  virtual void ProcessOutput(Instruction* instr);
+
+ private:
+  FILE *stream_;
+};
+}  // namespace vixl
+
+#endif  // VIXL_A64_DISASM_A64_H
diff --git a/disas/libvixl/src/a64/instructions-a64.cc b/disas/libvixl/src/a64/instructions-a64.cc
new file mode 100644
index 0000000..e87fa3a
--- /dev/null
+++ b/disas/libvixl/src/a64/instructions-a64.cc
@@ -0,0 +1,238 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "a64/instructions-a64.h"
+#include "a64/assembler-a64.h"
+
+namespace vixl {
+
+
+static uint64_t RotateRight(uint64_t value,
+                            unsigned int rotate,
+                            unsigned int width) {
+  ASSERT(width <= 64);
+  rotate &= 63;
+  return ((value & ((1UL << rotate) - 1UL)) << (width - rotate)) |
+         (value >> rotate);
+}
+
+
+static uint64_t RepeatBitsAcrossReg(unsigned reg_size,
+                                    uint64_t value,
+                                    unsigned width) {
+  ASSERT((width == 2) || (width == 4) || (width == 8) || (width == 16) ||
+         (width == 32));
+  ASSERT((reg_size == kWRegSize) || (reg_size == kXRegSize));
+  uint64_t result = value & ((1UL << width) - 1UL);
+  for (unsigned i = width; i < reg_size; i *= 2) {
+    result |= (result << i);
+  }
+  return result;
+}
+
+
+// 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.
+uint64_t Instruction::ImmLogical() {
+  unsigned reg_size = SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t n = BitN();
+  int64_t imm_s = ImmSetBits();
+  int64_t imm_r = ImmRotate();
+
+  // An integer is constructed from the n, imm_s and imm_r bits according to
+  // the following table:
+  //
+  //  N   imms    immr    size        S             R
+  //  1  ssssss  rrrrrr    64    UInt(ssssss)  UInt(rrrrrr)
+  //  0  0sssss  xrrrrr    32    UInt(sssss)   UInt(rrrrr)
+  //  0  10ssss  xxrrrr    16    UInt(ssss)    UInt(rrrr)
+  //  0  110sss  xxxrrr     8    UInt(sss)     UInt(rrr)
+  //  0  1110ss  xxxxrr     4    UInt(ss)      UInt(rr)
+  //  0  11110s  xxxxxr     2    UInt(s)       UInt(r)
+  // (s bits must not be all set)
+  //
+  // A pattern is constructed of size bits, where the least significant S+1
+  // bits are set. The pattern is rotated right by R, and repeated across a
+  // 32 or 64-bit value, depending on destination register width.
+  //
+
+  if (n == 1) {
+    if (imm_s == 0x3F) {
+      return 0;
+    }
+    uint64_t bits = (1UL << (imm_s + 1)) - 1;
+    return RotateRight(bits, imm_r, 64);
+  } else {
+    if ((imm_s >> 1) == 0x1F) {
+      return 0;
+    }
+    for (int width = 0x20; width >= 0x2; width >>= 1) {
+      if ((imm_s & width) == 0) {
+        int mask = width - 1;
+        if ((imm_s & mask) == mask) {
+          return 0;
+        }
+        uint64_t bits = (1UL << ((imm_s & mask) + 1)) - 1;
+        return RepeatBitsAcrossReg(reg_size,
+                                   RotateRight(bits, imm_r & mask, width),
+                                   width);
+      }
+    }
+  }
+  UNREACHABLE();
+  return 0;
+}
+
+
+float Instruction::ImmFP32() {
+  //  ImmFP: abcdefgh (8 bits)
+  // Single: aBbb.bbbc.defg.h000.0000.0000.0000.0000 (32 bits)
+  // where B is b ^ 1
+  uint32_t bits = ImmFP();
+  uint32_t bit7 = (bits >> 7) & 0x1;
+  uint32_t bit6 = (bits >> 6) & 0x1;
+  uint32_t bit5_to_0 = bits & 0x3f;
+  uint32_t result = (bit7 << 31) | ((32 - bit6) << 25) | (bit5_to_0 << 19);
+
+  return rawbits_to_float(result);
+}
+
+
+double Instruction::ImmFP64() {
+  //  ImmFP: abcdefgh (8 bits)
+  // Double: aBbb.bbbb.bbcd.efgh.0000.0000.0000.0000
+  //         0000.0000.0000.0000.0000.0000.0000.0000 (64 bits)
+  // where B is b ^ 1
+  uint32_t bits = ImmFP();
+  uint64_t bit7 = (bits >> 7) & 0x1;
+  uint64_t bit6 = (bits >> 6) & 0x1;
+  uint64_t bit5_to_0 = bits & 0x3f;
+  uint64_t result = (bit7 << 63) | ((256 - bit6) << 54) | (bit5_to_0 << 48);
+
+  return rawbits_to_double(result);
+}
+
+
+LSDataSize CalcLSPairDataSize(LoadStorePairOp op) {
+  switch (op) {
+    case STP_x:
+    case LDP_x:
+    case STP_d:
+    case LDP_d: return LSDoubleWord;
+    default: return LSWord;
+  }
+}
+
+
+Instruction* Instruction::ImmPCOffsetTarget() {
+  ptrdiff_t offset;
+  if (IsPCRelAddressing()) {
+    // PC-relative addressing. Only ADR is supported.
+    offset = ImmPCRel();
+  } else {
+    // All PC-relative branches.
+    ASSERT(BranchType() != UnknownBranchType);
+    // Relative branch offsets are instruction-size-aligned.
+    offset = ImmBranch() << kInstructionSizeLog2;
+  }
+  return this + offset;
+}
+
+
+inline int Instruction::ImmBranch() const {
+  switch (BranchType()) {
+    case CondBranchType: return ImmCondBranch();
+    case UncondBranchType: return ImmUncondBranch();
+    case CompareBranchType: return ImmCmpBranch();
+    case TestBranchType: return ImmTestBranch();
+    default: UNREACHABLE();
+  }
+  return 0;
+}
+
+
+void Instruction::SetImmPCOffsetTarget(Instruction* target) {
+  if (IsPCRelAddressing()) {
+    SetPCRelImmTarget(target);
+  } else {
+    SetBranchImmTarget(target);
+  }
+}
+
+
+void Instruction::SetPCRelImmTarget(Instruction* target) {
+  // ADRP is not supported, so 'this' must point to an ADR instruction.
+  ASSERT(Mask(PCRelAddressingMask) == ADR);
+
+  Instr imm = Assembler::ImmPCRelAddress(target - this);
+
+  SetInstructionBits(Mask(~ImmPCRel_mask) | imm);
+}
+
+
+void Instruction::SetBranchImmTarget(Instruction* target) {
+  ASSERT(((target - this) & 3) == 0);
+  Instr branch_imm = 0;
+  uint32_t imm_mask = 0;
+  int offset = (target - this) >> kInstructionSizeLog2;
+  switch (BranchType()) {
+    case CondBranchType: {
+      branch_imm = Assembler::ImmCondBranch(offset);
+      imm_mask = ImmCondBranch_mask;
+      break;
+    }
+    case UncondBranchType: {
+      branch_imm = Assembler::ImmUncondBranch(offset);
+      imm_mask = ImmUncondBranch_mask;
+      break;
+    }
+    case CompareBranchType: {
+      branch_imm = Assembler::ImmCmpBranch(offset);
+      imm_mask = ImmCmpBranch_mask;
+      break;
+    }
+    case TestBranchType: {
+      branch_imm = Assembler::ImmTestBranch(offset);
+      imm_mask = ImmTestBranch_mask;
+      break;
+    }
+    default: UNREACHABLE();
+  }
+  SetInstructionBits(Mask(~imm_mask) | branch_imm);
+}
+
+
+void Instruction::SetImmLLiteral(Instruction* source) {
+  ASSERT(((source - this) & 3) == 0);
+  int offset = (source - this) >> kLiteralEntrySizeLog2;
+  Instr imm = Assembler::ImmLLiteral(offset);
+  Instr mask = ImmLLiteral_mask;
+
+  SetInstructionBits(Mask(~mask) | imm);
+}
+}  // namespace vixl
+
diff --git a/disas/libvixl/src/a64/instructions-a64.h b/disas/libvixl/src/a64/instructions-a64.h
new file mode 100644
index 0000000..0f31fcd
--- /dev/null
+++ b/disas/libvixl/src/a64/instructions-a64.h
@@ -0,0 +1,344 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_INSTRUCTIONS_A64_H_
+#define VIXL_A64_INSTRUCTIONS_A64_H_
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/constants-a64.h"
+
+namespace vixl {
+// ISA constants. --------------------------------------------------------------
+
+typedef uint32_t Instr;
+const unsigned kInstructionSize = 4;
+const unsigned kInstructionSizeLog2 = 2;
+const unsigned kLiteralEntrySize = 4;
+const unsigned kLiteralEntrySizeLog2 = 2;
+const unsigned kMaxLoadLiteralRange = 1 * MBytes;
+
+const unsigned kWRegSize = 32;
+const unsigned kWRegSizeLog2 = 5;
+const unsigned kWRegSizeInBytes = kWRegSize / 8;
+const unsigned kXRegSize = 64;
+const unsigned kXRegSizeLog2 = 6;
+const unsigned kXRegSizeInBytes = kXRegSize / 8;
+const unsigned kSRegSize = 32;
+const unsigned kSRegSizeLog2 = 5;
+const unsigned kSRegSizeInBytes = kSRegSize / 8;
+const unsigned kDRegSize = 64;
+const unsigned kDRegSizeLog2 = 6;
+const unsigned kDRegSizeInBytes = kDRegSize / 8;
+const int64_t kWRegMask = 0x00000000ffffffffL;
+const int64_t kXRegMask = 0xffffffffffffffffL;
+const int64_t kSRegMask = 0x00000000ffffffffL;
+const int64_t kDRegMask = 0xffffffffffffffffL;
+const int64_t kXSignMask = 0x1L << 63;
+const int64_t kWSignMask = 0x1L << 31;
+const int64_t kByteMask = 0xffL;
+const int64_t kHalfWordMask = 0xffffL;
+const int64_t kWordMask = 0xffffffffL;
+const uint64_t kXMaxUInt = 0xffffffffffffffffUL;
+const uint64_t kWMaxUInt = 0xffffffffUL;
+const int64_t kXMaxInt = 0x7fffffffffffffffL;
+const int64_t kXMinInt = 0x8000000000000000L;
+const int32_t kWMaxInt = 0x7fffffff;
+const int32_t kWMinInt = 0x80000000;
+const unsigned kLinkRegCode = 30;
+const unsigned kZeroRegCode = 31;
+const unsigned kSPRegInternalCode = 63;
+const unsigned kRegCodeMask = 0x1f;
+
+// AArch64 floating-point specifics. These match IEEE-754.
+const unsigned kDoubleMantissaBits = 52;
+const unsigned kDoubleExponentBits = 11;
+const unsigned kFloatMantissaBits = 23;
+const unsigned kFloatExponentBits = 8;
+
+const float kFP32PositiveInfinity = rawbits_to_float(0x7f800000);
+const float kFP32NegativeInfinity = rawbits_to_float(0xff800000);
+const double kFP64PositiveInfinity = rawbits_to_double(0x7ff0000000000000UL);
+const double kFP64NegativeInfinity = rawbits_to_double(0xfff0000000000000UL);
+
+// This value is a signalling NaN as both a double and as a float (taking the
+// least-significant word).
+static const double kFP64SignallingNaN = rawbits_to_double(0x7ff000007f800001);
+static const float kFP32SignallingNaN = rawbits_to_float(0x7f800001);
+
+// A similar value, but as a quiet NaN.
+static const double kFP64QuietNaN = rawbits_to_double(0x7ff800007fc00001);
+static const float kFP32QuietNaN = rawbits_to_float(0x7fc00001);
+
+enum LSDataSize {
+  LSByte        = 0,
+  LSHalfword    = 1,
+  LSWord        = 2,
+  LSDoubleWord  = 3
+};
+
+LSDataSize CalcLSPairDataSize(LoadStorePairOp op);
+
+enum ImmBranchType {
+  UnknownBranchType = 0,
+  CondBranchType    = 1,
+  UncondBranchType  = 2,
+  CompareBranchType = 3,
+  TestBranchType    = 4
+};
+
+enum AddrMode {
+  Offset,
+  PreIndex,
+  PostIndex
+};
+
+enum FPRounding {
+  // The first four values are encodable directly by FPCR<RMode>.
+  FPTieEven = 0x0,
+  FPPositiveInfinity = 0x1,
+  FPNegativeInfinity = 0x2,
+  FPZero = 0x3,
+
+  // The final rounding mode is only available when explicitly specified by the
+  // instruction (such as with fcvta). It cannot be set in FPCR.
+  FPTieAway
+};
+
+enum Reg31Mode {
+  Reg31IsStackPointer,
+  Reg31IsZeroRegister
+};
+
+// Instructions. ---------------------------------------------------------------
+
+class Instruction {
+ public:
+  inline Instr InstructionBits() const {
+    return *(reinterpret_cast<const Instr*>(this));
+  }
+
+  inline void SetInstructionBits(Instr new_instr) {
+    *(reinterpret_cast<Instr*>(this)) = new_instr;
+  }
+
+  inline int Bit(int pos) const {
+    return (InstructionBits() >> pos) & 1;
+  }
+
+  inline 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 bits = *(reinterpret_cast<const int32_t*>(this));
+    return signed_bitextract_32(msb, lsb, bits);
+  }
+
+  inline 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); }
+  INSTRUCTION_FIELDS_LIST(DEFINE_GETTER)
+  #undef DEFINE_GETTER
+
+  // ImmPCRel is a compound field (not present in INSTRUCTION_FIELDS_LIST),
+  // formed from ImmPCRelLo and ImmPCRelHi.
+  int ImmPCRel() const {
+    int const offset = ((ImmPCRelHi() << ImmPCRelLo_width) | ImmPCRelLo());
+    int const width = ImmPCRelLo_width + ImmPCRelHi_width;
+    return signed_bitextract_32(width-1, 0, offset);
+  }
+
+  uint64_t ImmLogical();
+  float ImmFP32();
+  double ImmFP64();
+
+  inline LSDataSize SizeLSPair() const {
+    return CalcLSPairDataSize(
+             static_cast<LoadStorePairOp>(Mask(LoadStorePairMask)));
+  }
+
+  // Helpers.
+  inline bool IsCondBranchImm() const {
+    return Mask(ConditionalBranchFMask) == ConditionalBranchFixed;
+  }
+
+  inline bool IsUncondBranchImm() const {
+    return Mask(UnconditionalBranchFMask) == UnconditionalBranchFixed;
+  }
+
+  inline bool IsCompareBranch() const {
+    return Mask(CompareBranchFMask) == CompareBranchFixed;
+  }
+
+  inline bool IsTestBranch() const {
+    return Mask(TestBranchFMask) == TestBranchFixed;
+  }
+
+  inline bool IsPCRelAddressing() const {
+    return Mask(PCRelAddressingFMask) == PCRelAddressingFixed;
+  }
+
+  inline bool IsLogicalImmediate() const {
+    return Mask(LogicalImmediateFMask) == LogicalImmediateFixed;
+  }
+
+  inline bool IsAddSubImmediate() const {
+    return Mask(AddSubImmediateFMask) == AddSubImmediateFixed;
+  }
+
+  inline bool IsAddSubExtended() const {
+    return Mask(AddSubExtendedFMask) == AddSubExtendedFixed;
+  }
+
+  inline bool IsLoadOrStore() const {
+    return Mask(LoadStoreAnyFMask) == LoadStoreAnyFixed;
+  }
+
+  inline 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 {
+    // 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.
+    //  Logical (immediate) when not setting the flags.
+    // Otherwise, r31 is the zero register.
+    if (IsAddSubImmediate() || IsAddSubExtended()) {
+      if (Mask(AddSubSetFlagsBit)) {
+        return Reg31IsZeroRegister;
+      } else {
+        return Reg31IsStackPointer;
+      }
+    }
+    if (IsLogicalImmediate()) {
+      // Of the logical (immediate) instructions, only ANDS (and its aliases)
+      // can set the flags. The others can all write into sp.
+      // Note that some logical operations are not available to
+      // immediate-operand instructions, so we have to combine two masks here.
+      if (Mask(LogicalImmediateMask & LogicalOpMask) == ANDS) {
+        return Reg31IsZeroRegister;
+      } else {
+        return Reg31IsStackPointer;
+      }
+    }
+    return Reg31IsZeroRegister;
+  }
+
+  // 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 {
+    // The following instructions use sp or wsp as Rn:
+    //  All loads and stores.
+    //  Add/sub (immediate).
+    //  Add/sub (extended).
+    // Otherwise, r31 is the zero register.
+    if (IsLoadOrStore() || IsAddSubImmediate() || IsAddSubExtended()) {
+      return Reg31IsStackPointer;
+    }
+    return Reg31IsZeroRegister;
+  }
+
+  inline ImmBranchType BranchType() const {
+    if (IsCondBranchImm()) {
+      return CondBranchType;
+    } else if (IsUncondBranchImm()) {
+      return UncondBranchType;
+    } else if (IsCompareBranch()) {
+      return CompareBranchType;
+    } else if (IsTestBranch()) {
+      return TestBranchType;
+    } else {
+      return UnknownBranchType;
+    }
+  }
+
+  // Find the target of this instruction. 'this' may be a branch or a
+  // PC-relative addressing instruction.
+  Instruction* ImmPCOffsetTarget();
+
+  // Patch a PC-relative offset to refer to 'target'. 'this' may be a branch or
+  // a PC-relative addressing instruction.
+  void SetImmPCOffsetTarget(Instruction* target);
+  // Patch a literal load instruction to load from 'source'.
+  void SetImmLLiteral(Instruction* source);
+
+  inline uint8_t* LiteralAddress() {
+    int offset = ImmLLiteral() << kLiteralEntrySizeLog2;
+    return reinterpret_cast<uint8_t*>(this) + offset;
+  }
+
+  inline uint32_t Literal32() {
+    uint32_t literal;
+    memcpy(&literal, LiteralAddress(), sizeof(literal));
+
+    return literal;
+  }
+
+  inline uint64_t Literal64() {
+    uint64_t literal;
+    memcpy(&literal, LiteralAddress(), sizeof(literal));
+
+    return literal;
+  }
+
+  inline float LiteralFP32() {
+    return rawbits_to_float(Literal32());
+  }
+
+  inline double LiteralFP64() {
+    return rawbits_to_double(Literal64());
+  }
+
+  inline Instruction* NextInstruction() {
+    return this + kInstructionSize;
+  }
+
+  inline Instruction* InstructionAtOffset(int64_t offset) {
+    ASSERT(IsWordAligned(this + offset));
+    return this + offset;
+  }
+
+  template<typename T> static inline Instruction* Cast(T src) {
+    return reinterpret_cast<Instruction*>(src);
+  }
+
+ private:
+  inline int ImmBranch() const;
+
+  void SetPCRelImmTarget(Instruction* target);
+  void SetBranchImmTarget(Instruction* target);
+};
+}  // namespace vixl
+
+#endif  // VIXL_A64_INSTRUCTIONS_A64_H_
diff --git a/disas/libvixl/src/a64/instrument-a64.cc b/disas/libvixl/src/a64/instrument-a64.cc
new file mode 100644
index 0000000..507410d
--- /dev/null
+++ b/disas/libvixl/src/a64/instrument-a64.cc
@@ -0,0 +1,638 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "a64/instrument-a64.h"
+
+namespace vixl {
+
+Counter::Counter(const char* name, CounterType type)
+    : count_(0), enabled_(false), type_(type) {
+  ASSERT(name != NULL);
+  strncpy(name_, name, kCounterNameMaxLength);
+}
+
+
+void Counter::Enable() {
+  enabled_ = true;
+}
+
+
+void Counter::Disable() {
+  enabled_ = false;
+}
+
+
+bool Counter::IsEnabled() {
+  return enabled_;
+}
+
+
+void Counter::Increment() {
+  if (enabled_) {
+    count_++;
+  }
+}
+
+
+uint64_t Counter::count() {
+  uint64_t result = count_;
+  if (type_ == Gauge) {
+    // If the counter is a Gauge, reset the count after reading.
+    count_ = 0;
+  }
+  return result;
+}
+
+
+const char* Counter::name() {
+  return name_;
+}
+
+
+CounterType Counter::type() {
+  return type_;
+}
+
+
+typedef struct {
+  const char* name;
+  CounterType type;
+} CounterDescriptor;
+
+
+static const CounterDescriptor kCounterList[] = {
+  {"Instruction", Cumulative},
+
+  {"Move Immediate", Gauge},
+  {"Add/Sub DP", Gauge},
+  {"Logical DP", Gauge},
+  {"Other Int DP", Gauge},
+  {"FP DP", Gauge},
+
+  {"Conditional Select", Gauge},
+  {"Conditional Compare", Gauge},
+
+  {"Unconditional Branch", Gauge},
+  {"Compare and Branch", Gauge},
+  {"Test and Branch", Gauge},
+  {"Conditional Branch", Gauge},
+
+  {"Load Integer", Gauge},
+  {"Load FP", Gauge},
+  {"Load Pair", Gauge},
+  {"Load Literal", Gauge},
+
+  {"Store Integer", Gauge},
+  {"Store FP", Gauge},
+  {"Store Pair", Gauge},
+
+  {"PC Addressing", Gauge},
+  {"Other", Gauge},
+};
+
+
+Instrument::Instrument(const char* datafile, uint64_t sample_period)
+    : output_stream_(stdout), sample_period_(sample_period) {
+
+  // Set up the output stream. If datafile is non-NULL, use that file. If it
+  // can't be opened, or datafile is NULL, use stdout.
+  if (datafile != NULL) {
+    output_stream_ = fopen(datafile, "w");
+    if (output_stream_ == NULL) {
+      printf("Can't open output file %s. Using stdout.\n", datafile);
+      output_stream_ = stdout;
+    }
+  }
+
+  static const int num_counters =
+    sizeof(kCounterList) / sizeof(CounterDescriptor);
+
+  // Dump an instrumentation description comment at the top of the file.
+  fprintf(output_stream_, "# counters=%d\n", num_counters);
+  fprintf(output_stream_, "# sample_period=%" PRIu64 "\n", sample_period_);
+
+  // Construct Counter objects from counter description array.
+  for (int i = 0; i < num_counters; i++) {
+    Counter* counter = new Counter(kCounterList[i].name, kCounterList[i].type);
+    counters_.push_back(counter);
+  }
+
+  DumpCounterNames();
+}
+
+
+Instrument::~Instrument() {
+  // Dump any remaining instruction data to the output file.
+  DumpCounters();
+
+  // Free all the counter objects.
+  std::list<Counter*>::iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    free(*it);
+  }
+
+  if (output_stream_ != stdout) {
+    fclose(output_stream_);
+  }
+}
+
+
+void Instrument::Update() {
+  // Increment the instruction counter, and dump all counters if a sample period
+  // has elapsed.
+  static Counter* counter = GetCounter("Instruction");
+  ASSERT(counter->type() == Cumulative);
+  counter->Increment();
+
+  if (counter->IsEnabled() && (counter->count() % sample_period_) == 0) {
+    DumpCounters();
+  }
+}
+
+
+void Instrument::DumpCounters() {
+  // Iterate through the counter objects, dumping their values to the output
+  // stream.
+  std::list<Counter*>::const_iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    fprintf(output_stream_, "%" PRIu64 ",", (*it)->count());
+  }
+  fprintf(output_stream_, "\n");
+  fflush(output_stream_);
+}
+
+
+void Instrument::DumpCounterNames() {
+  // Iterate through the counter objects, dumping the counter names to the
+  // output stream.
+  std::list<Counter*>::const_iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    fprintf(output_stream_, "%s,", (*it)->name());
+  }
+  fprintf(output_stream_, "\n");
+  fflush(output_stream_);
+}
+
+
+void Instrument::HandleInstrumentationEvent(unsigned event) {
+  switch (event) {
+    case InstrumentStateEnable: Enable(); break;
+    case InstrumentStateDisable: Disable(); break;
+    default: DumpEventMarker(event);
+  }
+}
+
+
+void Instrument::DumpEventMarker(unsigned marker) {
+  // Dumpan event marker to the output stream as a specially formatted comment
+  // line.
+  static Counter* counter = GetCounter("Instruction");
+
+  fprintf(output_stream_, "# %c%c @ %" PRId64 "\n", marker & 0xff,
+          (marker >> 8) & 0xff, counter->count());
+}
+
+
+Counter* Instrument::GetCounter(const char* name) {
+  // Get a Counter object by name from the counter list.
+  std::list<Counter*>::const_iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    if (strcmp((*it)->name(), name) == 0) {
+      return *it;
+    }
+  }
+
+  // A Counter by that name does not exist: print an error message to stderr
+  // and the output file, and exit.
+  static const char* error_message =
+    "# Error: Unknown counter \"%s\". Exiting.\n";
+  fprintf(stderr, error_message, name);
+  fprintf(output_stream_, error_message, name);
+  exit(1);
+}
+
+
+void Instrument::Enable() {
+  std::list<Counter*>::iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    (*it)->Enable();
+  }
+}
+
+
+void Instrument::Disable() {
+  std::list<Counter*>::iterator it;
+  for (it = counters_.begin(); it != counters_.end(); it++) {
+    (*it)->Disable();
+  }
+}
+
+
+void Instrument::VisitPCRelAddressing(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("PC Addressing");
+  counter->Increment();
+}
+
+
+void Instrument::VisitAddSubImmediate(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Add/Sub DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitLogicalImmediate(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Logical DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitMoveWideImmediate(Instruction* instr) {
+  Update();
+  static Counter* counter = GetCounter("Move Immediate");
+
+  if (instr->IsMovn() && (instr->Rd() == kZeroRegCode)) {
+    unsigned imm = instr->ImmMoveWide();
+    HandleInstrumentationEvent(imm);
+  } else {
+    counter->Increment();
+  }
+}
+
+
+void Instrument::VisitBitfield(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other Int DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitExtract(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other Int DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitUnconditionalBranch(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Unconditional Branch");
+  counter->Increment();
+}
+
+
+void Instrument::VisitUnconditionalBranchToRegister(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Unconditional Branch");
+  counter->Increment();
+}
+
+
+void Instrument::VisitCompareBranch(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Compare and Branch");
+  counter->Increment();
+}
+
+
+void Instrument::VisitTestBranch(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Test and Branch");
+  counter->Increment();
+}
+
+
+void Instrument::VisitConditionalBranch(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Branch");
+  counter->Increment();
+}
+
+
+void Instrument::VisitSystem(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other");
+  counter->Increment();
+}
+
+
+void Instrument::VisitException(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other");
+  counter->Increment();
+}
+
+
+void Instrument::InstrumentLoadStorePair(Instruction* instr) {
+  static Counter* load_pair_counter = GetCounter("Load Pair");
+  static Counter* store_pair_counter = GetCounter("Store Pair");
+
+  if (instr->Mask(LoadStorePairLBit) != 0) {
+    load_pair_counter->Increment();
+  } else {
+    store_pair_counter->Increment();
+  }
+}
+
+
+void Instrument::VisitLoadStorePairPostIndex(Instruction* instr) {
+  USE(instr);
+  Update();
+  InstrumentLoadStorePair(instr);
+}
+
+
+void Instrument::VisitLoadStorePairOffset(Instruction* instr) {
+  USE(instr);
+  Update();
+  InstrumentLoadStorePair(instr);
+}
+
+
+void Instrument::VisitLoadStorePairPreIndex(Instruction* instr) {
+  USE(instr);
+  Update();
+  InstrumentLoadStorePair(instr);
+}
+
+
+void Instrument::VisitLoadStorePairNonTemporal(Instruction* instr) {
+  USE(instr);
+  Update();
+  InstrumentLoadStorePair(instr);
+}
+
+
+void Instrument::VisitLoadLiteral(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Load Literal");
+  counter->Increment();
+}
+
+
+void Instrument::InstrumentLoadStore(Instruction* instr) {
+  static Counter* load_int_counter = GetCounter("Load Integer");
+  static Counter* store_int_counter = GetCounter("Store Integer");
+  static Counter* load_fp_counter = GetCounter("Load FP");
+  static Counter* store_fp_counter = GetCounter("Store FP");
+
+  switch (instr->Mask(LoadStoreOpMask)) {
+    case STRB_w:    // Fall through.
+    case STRH_w:    // Fall through.
+    case STR_w:     // Fall through.
+    case STR_x:     store_int_counter->Increment(); break;
+    case STR_s:     // Fall through.
+    case STR_d:     store_fp_counter->Increment(); break;
+    case LDRB_w:    // Fall through.
+    case LDRH_w:    // Fall through.
+    case LDR_w:     // Fall through.
+    case LDR_x:     // Fall through.
+    case LDRSB_x:   // Fall through.
+    case LDRSH_x:   // Fall through.
+    case LDRSW_x:   // Fall through.
+    case LDRSB_w:   // Fall through.
+    case LDRSH_w:   load_int_counter->Increment(); break;
+    case LDR_s:     // Fall through.
+    case LDR_d:     load_fp_counter->Increment(); break;
+  }
+}
+
+
+void Instrument::VisitLoadStoreUnscaledOffset(Instruction* instr) {
+  Update();
+  InstrumentLoadStore(instr);
+}
+
+
+void Instrument::VisitLoadStorePostIndex(Instruction* instr) {
+  USE(instr);
+  Update();
+  InstrumentLoadStore(instr);
+}
+
+
+void Instrument::VisitLoadStorePreIndex(Instruction* instr) {
+  Update();
+  InstrumentLoadStore(instr);
+}
+
+
+void Instrument::VisitLoadStoreRegisterOffset(Instruction* instr) {
+  Update();
+  InstrumentLoadStore(instr);
+}
+
+
+void Instrument::VisitLoadStoreUnsignedOffset(Instruction* instr) {
+  Update();
+  InstrumentLoadStore(instr);
+}
+
+
+void Instrument::VisitLogicalShifted(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Logical DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitAddSubShifted(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Add/Sub DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitAddSubExtended(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Add/Sub DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitAddSubWithCarry(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Add/Sub DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitConditionalCompareRegister(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Compare");
+  counter->Increment();
+}
+
+
+void Instrument::VisitConditionalCompareImmediate(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Compare");
+  counter->Increment();
+}
+
+
+void Instrument::VisitConditionalSelect(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Select");
+  counter->Increment();
+}
+
+
+void Instrument::VisitDataProcessing1Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other Int DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitDataProcessing2Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other Int DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitDataProcessing3Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other Int DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPCompare(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPConditionalCompare(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Compare");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPConditionalSelect(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Conditional Select");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPImmediate(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPDataProcessing1Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPDataProcessing2Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPDataProcessing3Source(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPIntegerConvert(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitFPFixedPointConvert(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("FP DP");
+  counter->Increment();
+}
+
+
+void Instrument::VisitUnallocated(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other");
+  counter->Increment();
+}
+
+
+void Instrument::VisitUnimplemented(Instruction* instr) {
+  USE(instr);
+  Update();
+  static Counter* counter = GetCounter("Other");
+  counter->Increment();
+}
+
+
+}  // namespace v8::internal
diff --git a/disas/libvixl/src/a64/instrument-a64.h b/disas/libvixl/src/a64/instrument-a64.h
new file mode 100644
index 0000000..bee965b
--- /dev/null
+++ b/disas/libvixl/src/a64/instrument-a64.h
@@ -0,0 +1,108 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_INSTRUMENT_A64_H_
+#define VIXL_A64_INSTRUMENT_A64_H_
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/decoder-a64.h"
+#include "a64/constants-a64.h"
+#include "a64/instrument-a64.h"
+
+namespace vixl {
+
+const int kCounterNameMaxLength = 256;
+const uint64_t kDefaultInstrumentationSamplingPeriod = 1 << 22;
+
+
+enum InstrumentState {
+  InstrumentStateDisable = 0,
+  InstrumentStateEnable = 1
+};
+
+
+enum CounterType {
+  Gauge = 0,      // Gauge counters reset themselves after reading.
+  Cumulative = 1  // Cumulative counters keep their value after reading.
+};
+
+
+class Counter {
+ public:
+  Counter(const char* name, CounterType type = Gauge);
+  ~Counter();
+
+  void Increment();
+  void Enable();
+  void Disable();
+  bool IsEnabled();
+  uint64_t count();
+  const char* name();
+  CounterType type();
+
+ private:
+  char name_[kCounterNameMaxLength];
+  uint64_t count_;
+  bool enabled_;
+  CounterType type_;
+};
+
+
+class Instrument: public DecoderVisitor {
+ public:
+  explicit Instrument(const char* datafile = NULL,
+    uint64_t sample_period = kDefaultInstrumentationSamplingPeriod);
+  ~Instrument();
+
+  void Enable();
+  void Disable();
+
+  // Declare all Visitor functions.
+  #define DECLARE(A) void Visit##A(Instruction* instr);
+  VISITOR_LIST(DECLARE)
+  #undef DECLARE
+
+ private:
+  void Update();
+  void DumpCounters();
+  void DumpCounterNames();
+  void DumpEventMarker(unsigned marker);
+  void HandleInstrumentationEvent(unsigned event);
+  Counter* GetCounter(const char* name);
+
+  void InstrumentLoadStore(Instruction* instr);
+  void InstrumentLoadStorePair(Instruction* instr);
+
+  std::list<Counter*> counters_;
+
+  FILE *output_stream_;
+  uint64_t sample_period_;
+};
+
+}  // namespace vixl
+
+#endif  // VIXL_A64_INSTRUMENT_A64_H_
diff --git a/disas/libvixl/src/a64/macro-assembler-a64.cc b/disas/libvixl/src/a64/macro-assembler-a64.cc
new file mode 100644
index 0000000..39a925c
--- /dev/null
+++ b/disas/libvixl/src/a64/macro-assembler-a64.cc
@@ -0,0 +1,1108 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "a64/macro-assembler-a64.h"
+namespace vixl {
+
+void MacroAssembler::And(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, (S == SetFlags) ? ANDS : AND);
+}
+
+
+void MacroAssembler::Tst(const Register& rn,
+                         const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  And(AppropriateZeroRegFor(rn), rn, operand, SetFlags);
+}
+
+
+void MacroAssembler::Bic(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, (S == SetFlags) ? BICS : BIC);
+}
+
+
+void MacroAssembler::Orr(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, ORR);
+}
+
+
+void MacroAssembler::Orn(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, ORN);
+}
+
+
+void MacroAssembler::Eor(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, EOR);
+}
+
+
+void MacroAssembler::Eon(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  LogicalMacro(rd, rn, operand, EON);
+}
+
+
+void MacroAssembler::LogicalMacro(const Register& rd,
+                                  const Register& rn,
+                                  const Operand& operand,
+                                  LogicalOp op) {
+  if (operand.IsImmediate()) {
+    int64_t immediate = operand.immediate();
+    unsigned reg_size = rd.size();
+    ASSERT(rd.Is64Bits() || is_uint32(immediate));
+
+    // If the operation is NOT, invert the operation and immediate.
+    if ((op & NOT) == NOT) {
+      op = static_cast<LogicalOp>(op & ~NOT);
+      immediate = ~immediate;
+      if (rd.Is32Bits()) {
+        immediate &= kWRegMask;
+      }
+    }
+
+    // Special cases for all set or all clear immediates.
+    if (immediate == 0) {
+      switch (op) {
+        case AND:
+          Mov(rd, 0);
+          return;
+        case ORR:  // Fall through.
+        case EOR:
+          Mov(rd, rn);
+          return;
+        case ANDS:  // Fall through.
+        case BICS:
+          break;
+        default:
+          UNREACHABLE();
+      }
+    } else if ((rd.Is64Bits() && (immediate == -1L)) ||
+               (rd.Is32Bits() && (immediate == 0xffffffffL))) {
+      switch (op) {
+        case AND:
+          Mov(rd, rn);
+          return;
+        case ORR:
+          Mov(rd, immediate);
+          return;
+        case EOR:
+          Mvn(rd, rn);
+          return;
+        case ANDS:  // Fall through.
+        case BICS:
+          break;
+        default:
+          UNREACHABLE();
+      }
+    }
+
+    unsigned n, imm_s, imm_r;
+    if (IsImmLogical(immediate, reg_size, &n, &imm_s, &imm_r)) {
+      // Immediate can be encoded in the instruction.
+      LogicalImmediate(rd, rn, n, imm_s, imm_r, op);
+    } else {
+      // Immediate can't be encoded: synthesize using move immediate.
+      Register temp = AppropriateTempFor(rn);
+      Mov(temp, immediate);
+      if (rd.Is(sp)) {
+        // If rd is the stack pointer we cannot use it as the destination
+        // register so we use the temp register as an intermediate again.
+        Logical(temp, rn, Operand(temp), op);
+        Mov(sp, temp);
+      } else {
+        Logical(rd, rn, Operand(temp), op);
+      }
+    }
+  } else if (operand.IsExtendedRegister()) {
+    ASSERT(operand.reg().size() <= rd.size());
+    // Add/sub extended supports shift <= 4. We want to support exactly the
+    // same modes here.
+    ASSERT(operand.shift_amount() <= 4);
+    ASSERT(operand.reg().Is64Bits() ||
+           ((operand.extend() != UXTX) && (operand.extend() != SXTX)));
+    Register temp = AppropriateTempFor(rn, operand.reg());
+    EmitExtendShift(temp, operand.reg(), operand.extend(),
+                    operand.shift_amount());
+    Logical(rd, rn, Operand(temp), op);
+  } else {
+    // The operand can be encoded in the instruction.
+    ASSERT(operand.IsShiftedRegister());
+    Logical(rd, rn, operand, op);
+  }
+}
+
+
+void MacroAssembler::Mov(const Register& rd, const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  if (operand.IsImmediate()) {
+    // Call the macro assembler for generic immediates.
+    Mov(rd, operand.immediate());
+  } else if (operand.IsShiftedRegister() && (operand.shift_amount() != 0)) {
+    // Emit a shift instruction if moving a shifted register. This operation
+    // could also be achieved using an orr instruction (like orn used by Mvn),
+    // but using a shift instruction makes the disassembly clearer.
+    EmitShift(rd, operand.reg(), operand.shift(), operand.shift_amount());
+  } else if (operand.IsExtendedRegister()) {
+    // Emit an extend instruction if moving an extended register. This handles
+    // extend with post-shift operations, too.
+    EmitExtendShift(rd, operand.reg(), operand.extend(),
+                    operand.shift_amount());
+  } else {
+    // Otherwise, emit a register move only if the registers are distinct, or
+    // if they are not X registers. Note that mov(w0, w0) is not a no-op
+    // because it clears the top word of x0.
+    // If the sp is an operand, add #0 is emitted, otherwise, orr #0.
+    if (!rd.Is(operand.reg()) || !rd.Is64Bits()) {
+      mov(rd, operand.reg());
+    }
+  }
+}
+
+
+void MacroAssembler::Mvn(const Register& rd, const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  if (operand.IsImmediate()) {
+    // Call the macro assembler for generic immediates.
+    Mvn(rd, operand.immediate());
+  } else if (operand.IsExtendedRegister()) {
+    // Emit two instructions for the extend case. This differs from Mov, as
+    // the extend and invert can't be achieved in one instruction.
+    Register temp = AppropriateTempFor(rd, operand.reg());
+    EmitExtendShift(temp, operand.reg(), operand.extend(),
+                    operand.shift_amount());
+    mvn(rd, Operand(temp));
+  } else {
+    // Otherwise, register and shifted register cases can be handled by the
+    // assembler directly, using orn.
+    mvn(rd, operand);
+  }
+}
+
+
+void MacroAssembler::Mov(const Register& rd, uint64_t imm) {
+  ASSERT(allow_macro_instructions_);
+  ASSERT(is_uint32(imm) || is_int32(imm) || rd.Is64Bits());
+
+  // Immediates on Aarch64 can be produced using an initial value, and zero to
+  // three move keep operations.
+  //
+  // Initial values can be generated with:
+  //  1. 64-bit move zero (movz).
+  //  2. 32-bit move negative (movn).
+  //  3. 64-bit move negative.
+  //  4. 32-bit orr immediate.
+  //  5. 64-bit orr immediate.
+  // Move-keep may then be used to modify each of the 16-bit nybbles.
+  //
+  // The code below supports all five initial value generators, and
+  // applying move-keep operations to move-zero initial values only.
+
+  unsigned reg_size = rd.size();
+  unsigned n, imm_s, imm_r;
+  if (IsImmMovz(imm, reg_size) && !rd.IsSP()) {
+    // Immediate can be represented in a move zero instruction.
+    movz(rd, imm);
+  } else if (IsImmMovn(imm, reg_size) && !rd.IsSP()) {
+    // Immediate can be represented in a move negative instruction. Movn can't
+    // write to the stack pointer.
+    movn(rd, rd.Is64Bits() ? ~imm : (~imm & kWRegMask));
+  } else if (IsImmLogical(imm, reg_size, &n, &imm_s, &imm_r)) {
+    // Immediate can be represented in a logical orr instruction.
+    ASSERT(!rd.IsZero());
+    LogicalImmediate(rd, AppropriateZeroRegFor(rd), n, imm_s, imm_r, ORR);
+  } else {
+    // Generic immediate case. Imm will be represented by
+    //   [imm3, imm2, imm1, imm0], where each imm is 16 bits.
+    // A move-zero is generated for the first non-zero immX, and a move-keep
+    // for subsequent non-zero immX.
+
+    // Use a temporary register when moving to the stack pointer.
+    Register temp = rd.IsSP() ? AppropriateTempFor(rd) : rd;
+
+    ASSERT((reg_size % 16) == 0);
+    bool first_mov_done = false;
+    for (unsigned i = 0; i < (temp.size() / 16); i++) {
+      uint64_t imm16 = (imm >> (16 * i)) & 0xffffL;
+      if (imm16 != 0) {
+        if (!first_mov_done) {
+          // Move the first non-zero 16-bit chunk into the destination register.
+          movz(temp, imm16, 16 * i);
+          first_mov_done = true;
+        } else {
+          // Construct a wider constant.
+          movk(temp, imm16, 16 * i);
+        }
+      }
+    }
+
+    if (rd.IsSP()) {
+      mov(rd, temp);
+    }
+
+    ASSERT(first_mov_done);
+  }
+}
+
+
+// The movz instruction can generate immediates containing an arbitrary 16-bit
+// value, with remaining bits set, eg. 0x00001234, 0x0000123400000000.
+bool MacroAssembler::IsImmMovz(uint64_t imm, unsigned reg_size) {
+  if (reg_size == kXRegSize) {
+    if (((imm & 0xffffffffffff0000UL) == 0UL) ||
+        ((imm & 0xffffffff0000ffffUL) == 0UL) ||
+        ((imm & 0xffff0000ffffffffUL) == 0UL) ||
+        ((imm & 0x0000ffffffffffffUL) == 0UL)) {
+      return true;
+    }
+  } else {
+    ASSERT(reg_size == kWRegSize);
+    imm &= kWRegMask;
+    if (((imm & 0xffff0000) == 0) ||
+        ((imm & 0x0000ffff) == 0)) {
+      return true;
+    }
+  }
+  return false;
+}
+
+
+// The movn instruction can generate immediates containing an arbitrary 16-bit
+// value, with remaining bits set, eg. 0xffff1234, 0xffff1234ffffffff.
+bool MacroAssembler::IsImmMovn(uint64_t imm, unsigned reg_size) {
+  return IsImmMovz(~imm, reg_size);
+}
+
+
+void MacroAssembler::Ccmp(const Register& rn,
+                          const Operand& operand,
+                          StatusFlags nzcv,
+                          Condition cond) {
+  ASSERT(allow_macro_instructions_);
+  ConditionalCompareMacro(rn, operand, nzcv, cond, CCMP);
+}
+
+
+void MacroAssembler::Ccmn(const Register& rn,
+                          const Operand& operand,
+                          StatusFlags nzcv,
+                          Condition cond) {
+  ASSERT(allow_macro_instructions_);
+  ConditionalCompareMacro(rn, operand, nzcv, cond, CCMN);
+}
+
+
+void MacroAssembler::ConditionalCompareMacro(const Register& rn,
+                                             const Operand& operand,
+                                             StatusFlags nzcv,
+                                             Condition cond,
+                                             ConditionalCompareOp op) {
+  ASSERT((cond != al) && (cond != nv));
+  if ((operand.IsShiftedRegister() && (operand.shift_amount() == 0)) ||
+      (operand.IsImmediate() && IsImmConditionalCompare(operand.immediate()))) {
+    // The immediate can be encoded in the instruction, or the operand is an
+    // unshifted register: call the assembler.
+    ConditionalCompare(rn, operand, nzcv, cond, op);
+  } else {
+    // The operand isn't directly supported by the instruction: perform the
+    // operation on a temporary register.
+    Register temp(NoReg);
+    if (operand.IsImmediate()) {
+      temp = AppropriateTempFor(rn);
+      Mov(temp, operand.immediate());
+    } else if (operand.IsShiftedRegister()) {
+      ASSERT(operand.shift() != ROR);
+      ASSERT(is_uintn(rn.size() == kXRegSize ? kXRegSizeLog2 : kWRegSizeLog2,
+                      operand.shift_amount()));
+      temp = AppropriateTempFor(rn, operand.reg());
+      EmitShift(temp, operand.reg(), operand.shift(), operand.shift_amount());
+    } else {
+      ASSERT(operand.IsExtendedRegister());
+      ASSERT(operand.reg().size() <= rn.size());
+      // Add/sub extended support a shift <= 4. We want to support exactly the
+      // same modes.
+      ASSERT(operand.shift_amount() <= 4);
+      ASSERT(operand.reg().Is64Bits() ||
+             ((operand.extend() != UXTX) && (operand.extend() != SXTX)));
+      temp = AppropriateTempFor(rn, operand.reg());
+      EmitExtendShift(temp, operand.reg(), operand.extend(),
+                    operand.shift_amount());
+    }
+    ConditionalCompare(rn, Operand(temp), nzcv, cond, op);
+  }
+}
+
+
+void MacroAssembler::Add(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  if (operand.IsImmediate() && (operand.immediate() < 0)) {
+    AddSubMacro(rd, rn, -operand.immediate(), S, SUB);
+  } else {
+    AddSubMacro(rd, rn, operand, S, ADD);
+  }
+}
+
+
+void MacroAssembler::Sub(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  if (operand.IsImmediate() && (operand.immediate() < 0)) {
+    AddSubMacro(rd, rn, -operand.immediate(), S, ADD);
+  } else {
+    AddSubMacro(rd, rn, operand, S, SUB);
+  }
+}
+
+
+void MacroAssembler::Cmn(const Register& rn, const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  Add(AppropriateZeroRegFor(rn), rn, operand, SetFlags);
+}
+
+
+void MacroAssembler::Cmp(const Register& rn, const Operand& operand) {
+  ASSERT(allow_macro_instructions_);
+  Sub(AppropriateZeroRegFor(rn), rn, operand, SetFlags);
+}
+
+
+void MacroAssembler::Neg(const Register& rd,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  if (operand.IsImmediate()) {
+    Mov(rd, -operand.immediate());
+  } else {
+    Sub(rd, AppropriateZeroRegFor(rd), operand, S);
+  }
+}
+
+
+void MacroAssembler::AddSubMacro(const Register& rd,
+                                 const Register& rn,
+                                 const Operand& operand,
+                                 FlagsUpdate S,
+                                 AddSubOp op) {
+  if ((operand.IsImmediate() && !IsImmAddSub(operand.immediate())) ||
+      (rn.IsZero() && !operand.IsShiftedRegister())                ||
+      (operand.IsShiftedRegister() && (operand.shift() == ROR))) {
+    Register temp = AppropriateTempFor(rn);
+    Mov(temp, operand);
+    AddSub(rd, rn, temp, S, op);
+  } else {
+    AddSub(rd, rn, operand, S, op);
+  }
+}
+
+
+void MacroAssembler::Adc(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  AddSubWithCarryMacro(rd, rn, operand, S, ADC);
+}
+
+
+void MacroAssembler::Sbc(const Register& rd,
+                         const Register& rn,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  AddSubWithCarryMacro(rd, rn, operand, S, SBC);
+}
+
+
+void MacroAssembler::Ngc(const Register& rd,
+                         const Operand& operand,
+                         FlagsUpdate S) {
+  ASSERT(allow_macro_instructions_);
+  Register zr = AppropriateZeroRegFor(rd);
+  Sbc(rd, zr, operand, S);
+}
+
+
+void MacroAssembler::AddSubWithCarryMacro(const Register& rd,
+                                          const Register& rn,
+                                          const Operand& operand,
+                                          FlagsUpdate S,
+                                          AddSubWithCarryOp op) {
+  ASSERT(rd.size() == rn.size());
+
+  if (operand.IsImmediate() ||
+      (operand.IsShiftedRegister() && (operand.shift() == ROR))) {
+    // Add/sub with carry (immediate or ROR shifted register.)
+    Register temp = AppropriateTempFor(rn);
+    Mov(temp, operand);
+    AddSubWithCarry(rd, rn, Operand(temp), S, op);
+  } else if (operand.IsShiftedRegister() && (operand.shift_amount() != 0)) {
+    // Add/sub with carry (shifted register).
+    ASSERT(operand.reg().size() == rd.size());
+    ASSERT(operand.shift() != ROR);
+    ASSERT(is_uintn(rd.size() == kXRegSize ? kXRegSizeLog2 : kWRegSizeLog2,
+                    operand.shift_amount()));
+    Register temp = AppropriateTempFor(rn, operand.reg());
+    EmitShift(temp, operand.reg(), operand.shift(), operand.shift_amount());
+    AddSubWithCarry(rd, rn, Operand(temp), S, op);
+  } else if (operand.IsExtendedRegister()) {
+    // Add/sub with carry (extended register).
+    ASSERT(operand.reg().size() <= rd.size());
+    // Add/sub extended supports a shift <= 4. We want to support exactly the
+    // same modes.
+    ASSERT(operand.shift_amount() <= 4);
+    ASSERT(operand.reg().Is64Bits() ||
+           ((operand.extend() != UXTX) && (operand.extend() != SXTX)));
+    Register temp = AppropriateTempFor(rn, operand.reg());
+    EmitExtendShift(temp, operand.reg(), operand.extend(),
+                    operand.shift_amount());
+    AddSubWithCarry(rd, rn, Operand(temp), S, op);
+  } else {
+    // The addressing mode is directly supported by the instruction.
+    AddSubWithCarry(rd, rn, operand, S, op);
+  }
+}
+
+
+#define DEFINE_FUNCTION(FN, REGTYPE, REG, OP)                         \
+void MacroAssembler::FN(const REGTYPE REG, const MemOperand& addr) {  \
+  LoadStoreMacro(REG, addr, OP);                                      \
+}
+LS_MACRO_LIST(DEFINE_FUNCTION)
+#undef DEFINE_FUNCTION
+
+void MacroAssembler::LoadStoreMacro(const CPURegister& rt,
+                                    const MemOperand& addr,
+                                    LoadStoreOp op) {
+  int64_t offset = addr.offset();
+  LSDataSize size = CalcLSDataSize(op);
+
+  // Check if an immediate offset fits in the immediate field of the
+  // appropriate instruction. If not, emit two instructions to perform
+  // the operation.
+  if (addr.IsImmediateOffset() && !IsImmLSScaled(offset, size) &&
+      !IsImmLSUnscaled(offset)) {
+    // Immediate offset that can't be encoded using unsigned or unscaled
+    // addressing modes.
+    Register temp = AppropriateTempFor(addr.base());
+    Mov(temp, addr.offset());
+    LoadStore(rt, MemOperand(addr.base(), temp), op);
+  } else if (addr.IsPostIndex() && !IsImmLSUnscaled(offset)) {
+    // Post-index beyond unscaled addressing range.
+    LoadStore(rt, MemOperand(addr.base()), op);
+    Add(addr.base(), addr.base(), Operand(offset));
+  } else if (addr.IsPreIndex() && !IsImmLSUnscaled(offset)) {
+    // Pre-index beyond unscaled addressing range.
+    Add(addr.base(), addr.base(), Operand(offset));
+    LoadStore(rt, MemOperand(addr.base()), op);
+  } else {
+    // Encodable in one load/store instruction.
+    LoadStore(rt, addr, op);
+  }
+}
+
+
+void MacroAssembler::Push(const CPURegister& src0, const CPURegister& src1,
+                          const CPURegister& src2, const CPURegister& src3) {
+  ASSERT(allow_macro_instructions_);
+  ASSERT(AreSameSizeAndType(src0, src1, src2, src3));
+  ASSERT(src0.IsValid());
+
+  int count = 1 + src1.IsValid() + src2.IsValid() + src3.IsValid();
+  int size = src0.SizeInBytes();
+
+  PrepareForPush(count, size);
+  PushHelper(count, size, src0, src1, src2, src3);
+}
+
+
+void MacroAssembler::Pop(const CPURegister& dst0, const CPURegister& dst1,
+                         const CPURegister& dst2, const CPURegister& dst3) {
+  // It is not valid to pop into the same register more than once in one
+  // instruction, not even into the zero register.
+  ASSERT(allow_macro_instructions_);
+  ASSERT(!AreAliased(dst0, dst1, dst2, dst3));
+  ASSERT(AreSameSizeAndType(dst0, dst1, dst2, dst3));
+  ASSERT(dst0.IsValid());
+
+  int count = 1 + dst1.IsValid() + dst2.IsValid() + dst3.IsValid();
+  int size = dst0.SizeInBytes();
+
+  PrepareForPop(count, size);
+  PopHelper(count, size, dst0, dst1, dst2, dst3);
+}
+
+
+void MacroAssembler::PushCPURegList(CPURegList registers) {
+  int size = registers.RegisterSizeInBytes();
+
+  PrepareForPush(registers.Count(), size);
+  // Push up to four registers at a time because if the current stack pointer is
+  // sp and reg_size is 32, registers must be pushed in blocks of four in order
+  // to maintain the 16-byte alignment for sp.
+  ASSERT(allow_macro_instructions_);
+  while (!registers.IsEmpty()) {
+    int count_before = registers.Count();
+    const CPURegister& src0 = registers.PopHighestIndex();
+    const CPURegister& src1 = registers.PopHighestIndex();
+    const CPURegister& src2 = registers.PopHighestIndex();
+    const CPURegister& src3 = registers.PopHighestIndex();
+    int count = count_before - registers.Count();
+    PushHelper(count, size, src0, src1, src2, src3);
+  }
+}
+
+
+void MacroAssembler::PopCPURegList(CPURegList registers) {
+  int size = registers.RegisterSizeInBytes();
+
+  PrepareForPop(registers.Count(), size);
+  // Pop up to four registers at a time because if the current stack pointer is
+  // sp and reg_size is 32, registers must be pushed in blocks of four in order
+  // to maintain the 16-byte alignment for sp.
+  ASSERT(allow_macro_instructions_);
+  while (!registers.IsEmpty()) {
+    int count_before = registers.Count();
+    const CPURegister& dst0 = registers.PopLowestIndex();
+    const CPURegister& dst1 = registers.PopLowestIndex();
+    const CPURegister& dst2 = registers.PopLowestIndex();
+    const CPURegister& dst3 = registers.PopLowestIndex();
+    int count = count_before - registers.Count();
+    PopHelper(count, size, dst0, dst1, dst2, dst3);
+  }
+}
+
+
+void MacroAssembler::PushMultipleTimes(int count, Register src) {
+  ASSERT(allow_macro_instructions_);
+  int size = src.SizeInBytes();
+
+  PrepareForPush(count, size);
+  // Push up to four registers at a time if possible because if the current
+  // stack pointer is sp and the register size is 32, registers must be pushed
+  // in blocks of four in order to maintain the 16-byte alignment for sp.
+  while (count >= 4) {
+    PushHelper(4, size, src, src, src, src);
+    count -= 4;
+  }
+  if (count >= 2) {
+    PushHelper(2, size, src, src, NoReg, NoReg);
+    count -= 2;
+  }
+  if (count == 1) {
+    PushHelper(1, size, src, NoReg, NoReg, NoReg);
+    count -= 1;
+  }
+  ASSERT(count == 0);
+}
+
+
+void MacroAssembler::PushHelper(int count, int size,
+                                const CPURegister& src0,
+                                const CPURegister& src1,
+                                const CPURegister& src2,
+                                const CPURegister& src3) {
+  // Ensure that we don't unintentionally modify scratch or debug registers.
+  InstructionAccurateScope scope(this);
+
+  ASSERT(AreSameSizeAndType(src0, src1, src2, src3));
+  ASSERT(size == src0.SizeInBytes());
+
+  // When pushing multiple registers, the store order is chosen such that
+  // Push(a, b) is equivalent to Push(a) followed by Push(b).
+  switch (count) {
+    case 1:
+      ASSERT(src1.IsNone() && src2.IsNone() && src3.IsNone());
+      str(src0, MemOperand(StackPointer(), -1 * size, PreIndex));
+      break;
+    case 2:
+      ASSERT(src2.IsNone() && src3.IsNone());
+      stp(src1, src0, MemOperand(StackPointer(), -2 * size, PreIndex));
+      break;
+    case 3:
+      ASSERT(src3.IsNone());
+      stp(src2, src1, MemOperand(StackPointer(), -3 * size, PreIndex));
+      str(src0, MemOperand(StackPointer(), 2 * size));
+      break;
+    case 4:
+      // Skip over 4 * size, then fill in the gap. This allows four W registers
+      // to be pushed using sp, whilst maintaining 16-byte alignment for sp at
+      // all times.
+      stp(src3, src2, MemOperand(StackPointer(), -4 * size, PreIndex));
+      stp(src1, src0, MemOperand(StackPointer(), 2 * size));
+      break;
+    default:
+      UNREACHABLE();
+  }
+}
+
+
+void MacroAssembler::PopHelper(int count, int size,
+                               const CPURegister& dst0,
+                               const CPURegister& dst1,
+                               const CPURegister& dst2,
+                               const CPURegister& dst3) {
+  // Ensure that we don't unintentionally modify scratch or debug registers.
+  InstructionAccurateScope scope(this);
+
+  ASSERT(AreSameSizeAndType(dst0, dst1, dst2, dst3));
+  ASSERT(size == dst0.SizeInBytes());
+
+  // When popping multiple registers, the load order is chosen such that
+  // Pop(a, b) is equivalent to Pop(a) followed by Pop(b).
+  switch (count) {
+    case 1:
+      ASSERT(dst1.IsNone() && dst2.IsNone() && dst3.IsNone());
+      ldr(dst0, MemOperand(StackPointer(), 1 * size, PostIndex));
+      break;
+    case 2:
+      ASSERT(dst2.IsNone() && dst3.IsNone());
+      ldp(dst0, dst1, MemOperand(StackPointer(), 2 * size, PostIndex));
+      break;
+    case 3:
+      ASSERT(dst3.IsNone());
+      ldr(dst2, MemOperand(StackPointer(), 2 * size));
+      ldp(dst0, dst1, MemOperand(StackPointer(), 3 * size, PostIndex));
+      break;
+    case 4:
+      // Load the higher addresses first, then load the lower addresses and skip
+      // the whole block in the second instruction. This allows four W registers
+      // to be popped using sp, whilst maintaining 16-byte alignment for sp at
+      // all times.
+      ldp(dst2, dst3, MemOperand(StackPointer(), 2 * size));
+      ldp(dst0, dst1, MemOperand(StackPointer(), 4 * size, PostIndex));
+      break;
+    default:
+      UNREACHABLE();
+  }
+}
+
+
+void MacroAssembler::PrepareForPush(int count, int size) {
+  if (sp.Is(StackPointer())) {
+    // If the current stack pointer is sp, then it must be aligned to 16 bytes
+    // on entry and the total size of the specified registers must also be a
+    // multiple of 16 bytes.
+    ASSERT((count * size) % 16 == 0);
+  } else {
+    // Even if the current stack pointer is not the system stack pointer (sp),
+    // the system stack pointer will still be modified in order to comply with
+    // ABI rules about accessing memory below the system stack pointer.
+    BumpSystemStackPointer(count * size);
+  }
+}
+
+
+void MacroAssembler::PrepareForPop(int count, int size) {
+  USE(count);
+  USE(size);
+  if (sp.Is(StackPointer())) {
+    // If the current stack pointer is sp, then it must be aligned to 16 bytes
+    // on entry and the total size of the specified registers must also be a
+    // multiple of 16 bytes.
+    ASSERT((count * size) % 16 == 0);
+  }
+}
+
+void MacroAssembler::Poke(const Register& src, const Operand& offset) {
+  ASSERT(allow_macro_instructions_);
+  if (offset.IsImmediate()) {
+    ASSERT(offset.immediate() >= 0);
+  }
+
+  Str(src, MemOperand(StackPointer(), offset));
+}
+
+
+void MacroAssembler::Peek(const Register& dst, const Operand& offset) {
+  ASSERT(allow_macro_instructions_);
+  if (offset.IsImmediate()) {
+    ASSERT(offset.immediate() >= 0);
+  }
+
+  Ldr(dst, MemOperand(StackPointer(), offset));
+}
+
+
+void MacroAssembler::Claim(const Operand& size) {
+  ASSERT(allow_macro_instructions_);
+  if (size.IsImmediate()) {
+    ASSERT(size.immediate() >= 0);
+    if (sp.Is(StackPointer())) {
+      ASSERT((size.immediate() % 16) == 0);
+    }
+  }
+
+  if (!sp.Is(StackPointer())) {
+    BumpSystemStackPointer(size);
+  }
+
+  Sub(StackPointer(), StackPointer(), size);
+}
+
+
+void MacroAssembler::Drop(const Operand& size) {
+  ASSERT(allow_macro_instructions_);
+  if (size.IsImmediate()) {
+    ASSERT(size.immediate() >= 0);
+    if (sp.Is(StackPointer())) {
+      ASSERT((size.immediate() % 16) == 0);
+    }
+  }
+
+  Add(StackPointer(), StackPointer(), size);
+}
+
+
+void MacroAssembler::PushCalleeSavedRegisters() {
+  // Ensure that the macro-assembler doesn't use any scratch registers.
+  InstructionAccurateScope scope(this);
+
+  // This method must not be called unless the current stack pointer is sp.
+  ASSERT(sp.Is(StackPointer()));
+
+  MemOperand tos(sp, -2 * kXRegSizeInBytes, PreIndex);
+
+  stp(d14, d15, tos);
+  stp(d12, d13, tos);
+  stp(d10, d11, tos);
+  stp(d8, d9, tos);
+
+  stp(x29, x30, tos);
+  stp(x27, x28, tos);
+  stp(x25, x26, tos);
+  stp(x23, x24, tos);
+  stp(x21, x22, tos);
+  stp(x19, x20, tos);
+}
+
+
+void MacroAssembler::PopCalleeSavedRegisters() {
+  // Ensure that the macro-assembler doesn't use any scratch registers.
+  InstructionAccurateScope scope(this);
+
+  // This method must not be called unless the current stack pointer is sp.
+  ASSERT(sp.Is(StackPointer()));
+
+  MemOperand tos(sp, 2 * kXRegSizeInBytes, PostIndex);
+
+  ldp(x19, x20, tos);
+  ldp(x21, x22, tos);
+  ldp(x23, x24, tos);
+  ldp(x25, x26, tos);
+  ldp(x27, x28, tos);
+  ldp(x29, x30, tos);
+
+  ldp(d8, d9, tos);
+  ldp(d10, d11, tos);
+  ldp(d12, d13, tos);
+  ldp(d14, d15, tos);
+}
+
+void MacroAssembler::BumpSystemStackPointer(const Operand& space) {
+  ASSERT(!sp.Is(StackPointer()));
+  // TODO: Several callers rely on this not using scratch registers, so we use
+  // the assembler directly here. However, this means that large immediate
+  // values of 'space' cannot be handled.
+  InstructionAccurateScope scope(this);
+  sub(sp, StackPointer(), space);
+}
+
+
+// This is the main Printf implementation. All callee-saved registers are
+// preserved, but NZCV and the caller-saved registers may be clobbered.
+void MacroAssembler::PrintfNoPreserve(const char * format,
+                                      const CPURegister& arg0,
+                                      const CPURegister& arg1,
+                                      const CPURegister& arg2,
+                                      const CPURegister& arg3) {
+  // We cannot handle a caller-saved stack pointer. It doesn't make much sense
+  // in most cases anyway, so this restriction shouldn't be too serious.
+  ASSERT(!kCallerSaved.IncludesAliasOf(StackPointer()));
+
+  // We cannot print Tmp0() or Tmp1() as they're used internally by the macro
+  // assembler. We cannot print the stack pointer because it is typically used
+  // to preserve caller-saved registers (using other Printf variants which
+  // depend on this helper).
+  ASSERT(!AreAliased(Tmp0(), Tmp1(), StackPointer(), arg0));
+  ASSERT(!AreAliased(Tmp0(), Tmp1(), StackPointer(), arg1));
+  ASSERT(!AreAliased(Tmp0(), Tmp1(), StackPointer(), arg2));
+  ASSERT(!AreAliased(Tmp0(), Tmp1(), StackPointer(), arg3));
+
+  static const int kMaxArgCount = 4;
+  // Assume that we have the maximum number of arguments until we know
+  // otherwise.
+  int arg_count = kMaxArgCount;
+
+  // The provided arguments.
+  CPURegister args[kMaxArgCount] = {arg0, arg1, arg2, arg3};
+
+  // The PCS registers where the arguments need to end up.
+  CPURegister pcs[kMaxArgCount];
+
+  // Promote FP arguments to doubles, and integer arguments to X registers.
+  // Note that FP and integer arguments cannot be mixed, but we'll check
+  // AreSameSizeAndType once we've processed these promotions.
+  for (int i = 0; i < kMaxArgCount; i++) {
+    if (args[i].IsRegister()) {
+      // Note that we use x1 onwards, because x0 will hold the format string.
+      pcs[i] = Register::XRegFromCode(i + 1);
+      // For simplicity, we handle all integer arguments as X registers. An X
+      // register argument takes the same space as a W register argument in the
+      // PCS anyway. The only limitation is that we must explicitly clear the
+      // top word for W register arguments as the callee will expect it to be
+      // clear.
+      if (!args[i].Is64Bits()) {
+        const Register& as_x = args[i].X();
+        And(as_x, as_x, 0x00000000ffffffff);
+        args[i] = as_x;
+      }
+    } else if (args[i].IsFPRegister()) {
+      pcs[i] = FPRegister::DRegFromCode(i);
+      // C and C++ varargs functions (such as printf) implicitly promote float
+      // arguments to doubles.
+      if (!args[i].Is64Bits()) {
+        FPRegister s(args[i]);
+        const FPRegister& as_d = args[i].D();
+        Fcvt(as_d, s);
+        args[i] = as_d;
+      }
+    } else {
+      // This is the first empty (NoCPUReg) argument, so use it to set the
+      // argument count and bail out.
+      arg_count = i;
+      break;
+    }
+  }
+  ASSERT((arg_count >= 0) && (arg_count <= kMaxArgCount));
+  // Check that every remaining argument is NoCPUReg.
+  for (int i = arg_count; i < kMaxArgCount; i++) {
+    ASSERT(args[i].IsNone());
+  }
+  ASSERT((arg_count == 0) || AreSameSizeAndType(args[0], args[1],
+                                                args[2], args[3],
+                                                pcs[0], pcs[1],
+                                                pcs[2], pcs[3]));
+
+  // Move the arguments into the appropriate PCS registers.
+  //
+  // Arranging an arbitrary list of registers into x1-x4 (or d0-d3) is
+  // surprisingly complicated.
+  //
+  //  * For even numbers of registers, we push the arguments and then pop them
+  //    into their final registers. This maintains 16-byte stack alignment in
+  //    case sp is the stack pointer, since we're only handling X or D registers
+  //    at this point.
+  //
+  //  * For odd numbers of registers, we push and pop all but one register in
+  //    the same way, but the left-over register is moved directly, since we
+  //    can always safely move one register without clobbering any source.
+  if (arg_count >= 4) {
+    Push(args[3], args[2], args[1], args[0]);
+  } else if (arg_count >= 2) {
+    Push(args[1], args[0]);
+  }
+
+  if ((arg_count % 2) != 0) {
+    // Move the left-over register directly.
+    const CPURegister& leftover_arg = args[arg_count - 1];
+    const CPURegister& leftover_pcs = pcs[arg_count - 1];
+    if (leftover_arg.IsRegister()) {
+      Mov(Register(leftover_pcs), Register(leftover_arg));
+    } else {
+      Fmov(FPRegister(leftover_pcs), FPRegister(leftover_arg));
+    }
+  }
+
+  if (arg_count >= 4) {
+    Pop(pcs[0], pcs[1], pcs[2], pcs[3]);
+  } else if (arg_count >= 2) {
+    Pop(pcs[0], pcs[1]);
+  }
+
+  // Load the format string into x0, as per the procedure-call standard.
+  //
+  // To make the code as portable as possible, the format string is encoded
+  // directly in the instruction stream. It might be cleaner to encode it in a
+  // literal pool, but since Printf is usually used for debugging, it is
+  // beneficial for it to be minimally dependent on other features.
+  Label format_address;
+  Adr(x0, &format_address);
+
+  // Emit the format string directly in the instruction stream.
+  { BlockLiteralPoolScope scope(this);
+    Label after_data;
+    B(&after_data);
+    Bind(&format_address);
+    EmitStringData(format);
+    Unreachable();
+    Bind(&after_data);
+  }
+
+  // We don't pass any arguments on the stack, but we still need to align the C
+  // stack pointer to a 16-byte boundary for PCS compliance.
+  if (!sp.Is(StackPointer())) {
+    Bic(sp, StackPointer(), 0xf);
+  }
+
+  // Actually call printf. This part needs special handling for the simulator,
+  // since the system printf function will use a different instruction set and
+  // the procedure-call standard will not be compatible.
+#ifdef USE_SIMULATOR
+  { InstructionAccurateScope scope(this, kPrintfLength / kInstructionSize);
+    hlt(kPrintfOpcode);
+    dc32(pcs[0].type());
+  }
+#else
+  Mov(Tmp0(), reinterpret_cast<uintptr_t>(printf));
+  Blr(Tmp0());
+#endif
+}
+
+
+void MacroAssembler::Printf(const char * format,
+                            const CPURegister& arg0,
+                            const CPURegister& arg1,
+                            const CPURegister& arg2,
+                            const CPURegister& arg3) {
+  // Preserve all caller-saved registers as well as NZCV.
+  // If sp is the stack pointer, PushCPURegList asserts that the size of each
+  // list is a multiple of 16 bytes.
+  PushCPURegList(kCallerSaved);
+  PushCPURegList(kCallerSavedFP);
+  // Use Tmp0() as a scratch register. It is not accepted by Printf so it will
+  // never overlap an argument register.
+  Mrs(Tmp0(), NZCV);
+  Push(Tmp0(), xzr);
+
+  PrintfNoPreserve(format, arg0, arg1, arg2, arg3);
+
+  Pop(xzr, Tmp0());
+  Msr(NZCV, Tmp0());
+  PopCPURegList(kCallerSavedFP);
+  PopCPURegList(kCallerSaved);
+}
+
+void MacroAssembler::Trace(TraceParameters parameters, TraceCommand command) {
+  ASSERT(allow_macro_instructions_);
+
+#ifdef USE_SIMULATOR
+  // The arguments to the trace pseudo instruction need to be contiguous in
+  // memory, so make sure we don't try to emit a literal pool.
+  InstructionAccurateScope scope(this, kTraceLength / kInstructionSize);
+
+  Label start;
+  bind(&start);
+
+  // Refer to instructions-a64.h for a description of the marker and its
+  // arguments.
+  hlt(kTraceOpcode);
+
+  ASSERT(SizeOfCodeGeneratedSince(&start) == kTraceParamsOffset);
+  dc32(parameters);
+
+  ASSERT(SizeOfCodeGeneratedSince(&start) == kTraceCommandOffset);
+  dc32(command);
+#else
+  // Emit nothing on real hardware.
+  USE(parameters);
+  USE(command);
+#endif
+}
+
+
+void MacroAssembler::Log(TraceParameters parameters) {
+  ASSERT(allow_macro_instructions_);
+
+#ifdef USE_SIMULATOR
+  // The arguments to the log pseudo instruction need to be contiguous in
+  // memory, so make sure we don't try to emit a literal pool.
+  InstructionAccurateScope scope(this, kLogLength / kInstructionSize);
+
+  Label start;
+  bind(&start);
+
+  // Refer to instructions-a64.h for a description of the marker and its
+  // arguments.
+  hlt(kLogOpcode);
+
+  ASSERT(SizeOfCodeGeneratedSince(&start) == kLogParamsOffset);
+  dc32(parameters);
+#else
+  // Emit nothing on real hardware.
+  USE(parameters);
+#endif
+}
+
+
+void MacroAssembler::EnableInstrumentation() {
+  ASSERT(!isprint(InstrumentStateEnable));
+  InstructionAccurateScope scope(this, 1);
+  movn(xzr, InstrumentStateEnable);
+}
+
+
+void MacroAssembler::DisableInstrumentation() {
+  ASSERT(!isprint(InstrumentStateDisable));
+  InstructionAccurateScope scope(this, 1);
+  movn(xzr, InstrumentStateDisable);
+}
+
+
+void MacroAssembler::AnnotateInstrumentation(const char* marker_name) {
+  ASSERT(strlen(marker_name) == 2);
+
+  // We allow only printable characters in the marker names. Unprintable
+  // characters are reserved for controlling features of the instrumentation.
+  ASSERT(isprint(marker_name[0]) && isprint(marker_name[1]));
+
+  InstructionAccurateScope scope(this, 1);
+  movn(xzr, (marker_name[1] << 8) | marker_name[0]);
+}
+
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/macro-assembler-a64.h b/disas/libvixl/src/a64/macro-assembler-a64.h
new file mode 100644
index 0000000..f266063
--- /dev/null
+++ b/disas/libvixl/src/a64/macro-assembler-a64.h
@@ -0,0 +1,1175 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_MACRO_ASSEMBLER_A64_H_
+#define VIXL_A64_MACRO_ASSEMBLER_A64_H_
+
+#include "globals.h"
+#include "a64/assembler-a64.h"
+#include "a64/debugger-a64.h"
+
+
+#define LS_MACRO_LIST(V)                                      \
+  V(Ldrb, Register&, rt, LDRB_w)                              \
+  V(Strb, Register&, rt, STRB_w)                              \
+  V(Ldrsb, Register&, rt, rt.Is64Bits() ? LDRSB_x : LDRSB_w)  \
+  V(Ldrh, Register&, rt, LDRH_w)                              \
+  V(Strh, Register&, rt, STRH_w)                              \
+  V(Ldrsh, Register&, rt, rt.Is64Bits() ? LDRSH_x : LDRSH_w)  \
+  V(Ldr, CPURegister&, rt, LoadOpFor(rt))                     \
+  V(Str, CPURegister&, rt, StoreOpFor(rt))                    \
+  V(Ldrsw, Register&, rt, LDRSW_x)
+
+namespace vixl {
+
+class MacroAssembler : public Assembler {
+ public:
+  MacroAssembler(byte * buffer, unsigned buffer_size)
+      : Assembler(buffer, buffer_size),
+#ifdef DEBUG
+        allow_macro_instructions_(true),
+#endif
+        sp_(sp), tmp0_(ip0), tmp1_(ip1), fptmp0_(d31) {}
+
+  // Logical macros.
+  void And(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Bic(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Orr(const Register& rd,
+           const Register& rn,
+           const Operand& operand);
+  void Orn(const Register& rd,
+           const Register& rn,
+           const Operand& operand);
+  void Eor(const Register& rd,
+           const Register& rn,
+           const Operand& operand);
+  void Eon(const Register& rd,
+           const Register& rn,
+           const Operand& operand);
+  void Tst(const Register& rn, const Operand& operand);
+  void LogicalMacro(const Register& rd,
+                    const Register& rn,
+                    const Operand& operand,
+                    LogicalOp op);
+
+  // Add and sub macros.
+  void Add(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Sub(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Cmn(const Register& rn, const Operand& operand);
+  void Cmp(const Register& rn, const Operand& operand);
+  void Neg(const Register& rd,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void AddSubMacro(const Register& rd,
+                   const Register& rn,
+                   const Operand& operand,
+                   FlagsUpdate S,
+                   AddSubOp op);
+
+  // Add/sub with carry macros.
+  void Adc(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Sbc(const Register& rd,
+           const Register& rn,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void Ngc(const Register& rd,
+           const Operand& operand,
+           FlagsUpdate S = LeaveFlags);
+  void AddSubWithCarryMacro(const Register& rd,
+                            const Register& rn,
+                            const Operand& operand,
+                            FlagsUpdate S,
+                            AddSubWithCarryOp op);
+
+  // Move macros.
+  void Mov(const Register& rd, uint64_t imm);
+  void Mov(const Register& rd, const Operand& operand);
+  void Mvn(const Register& rd, uint64_t imm) {
+    Mov(rd, ~imm);
+  };
+  void Mvn(const Register& rd, const Operand& operand);
+  bool IsImmMovn(uint64_t imm, unsigned reg_size);
+  bool IsImmMovz(uint64_t imm, unsigned reg_size);
+
+  // Conditional compare macros.
+  void Ccmp(const Register& rn,
+            const Operand& operand,
+            StatusFlags nzcv,
+            Condition cond);
+  void Ccmn(const Register& rn,
+            const Operand& operand,
+            StatusFlags nzcv,
+            Condition cond);
+  void ConditionalCompareMacro(const Register& rn,
+                               const Operand& operand,
+                               StatusFlags nzcv,
+                               Condition cond,
+                               ConditionalCompareOp op);
+
+  // Load/store macros.
+#define DECLARE_FUNCTION(FN, REGTYPE, REG, OP) \
+  void FN(const REGTYPE REG, const MemOperand& addr);
+  LS_MACRO_LIST(DECLARE_FUNCTION)
+#undef DECLARE_FUNCTION
+
+  void LoadStoreMacro(const CPURegister& rt,
+                      const MemOperand& addr,
+                      LoadStoreOp op);
+
+  // Push or pop up to 4 registers of the same width to or from the stack,
+  // using the current stack pointer as set by SetStackPointer.
+  //
+  // If an argument register is 'NoReg', all further arguments are also assumed
+  // to be 'NoReg', and are thus not pushed or popped.
+  //
+  // Arguments are ordered such that "Push(a, b);" is functionally equivalent
+  // to "Push(a); Push(b);".
+  //
+  // It is valid to push the same register more than once, and there is no
+  // restriction on the order in which registers are specified.
+  //
+  // It is not valid to pop into the same register more than once in one
+  // operation, not even into the zero register.
+  //
+  // If the current stack pointer (as set by SetStackPointer) is sp, then it
+  // must be aligned to 16 bytes on entry and the total size of the specified
+  // registers must also be a multiple of 16 bytes.
+  //
+  // Even if the current stack pointer is not the system stack pointer (sp),
+  // Push (and derived methods) will still modify the system stack pointer in
+  // order to comply with ABI rules about accessing memory below the system
+  // stack pointer.
+  //
+  // Other than the registers passed into Pop, the stack pointer and (possibly)
+  // the system stack pointer, these methods do not modify any other registers.
+  // Scratch registers such as Tmp0() and Tmp1() are preserved.
+  void Push(const CPURegister& src0, const CPURegister& src1 = NoReg,
+            const CPURegister& src2 = NoReg, const CPURegister& src3 = NoReg);
+  void Pop(const CPURegister& dst0, const CPURegister& dst1 = NoReg,
+           const CPURegister& dst2 = NoReg, const CPURegister& dst3 = NoReg);
+
+  // Alternative forms of Push and Pop, taking a RegList or CPURegList that
+  // specifies the registers that are to be pushed or popped. Higher-numbered
+  // registers are associated with higher memory addresses (as in the A32 push
+  // and pop instructions).
+  //
+  // (Push|Pop)SizeRegList allow you to specify the register size as a
+  // parameter. Only kXRegSize, kWRegSize, kDRegSize and kSRegSize are
+  // supported.
+  //
+  // Otherwise, (Push|Pop)(CPU|X|W|D|S)RegList is preferred.
+  void PushCPURegList(CPURegList registers);
+  void PopCPURegList(CPURegList registers);
+
+  void PushSizeRegList(RegList registers, unsigned reg_size,
+      CPURegister::RegisterType type = CPURegister::kRegister) {
+    PushCPURegList(CPURegList(type, reg_size, registers));
+  }
+  void PopSizeRegList(RegList registers, unsigned reg_size,
+      CPURegister::RegisterType type = CPURegister::kRegister) {
+    PopCPURegList(CPURegList(type, reg_size, registers));
+  }
+  void PushXRegList(RegList regs) {
+    PushSizeRegList(regs, kXRegSize);
+  }
+  void PopXRegList(RegList regs) {
+    PopSizeRegList(regs, kXRegSize);
+  }
+  void PushWRegList(RegList regs) {
+    PushSizeRegList(regs, kWRegSize);
+  }
+  void PopWRegList(RegList regs) {
+    PopSizeRegList(regs, kWRegSize);
+  }
+  inline void PushDRegList(RegList regs) {
+    PushSizeRegList(regs, kDRegSize, CPURegister::kFPRegister);
+  }
+  inline void PopDRegList(RegList regs) {
+    PopSizeRegList(regs, kDRegSize, CPURegister::kFPRegister);
+  }
+  inline void PushSRegList(RegList regs) {
+    PushSizeRegList(regs, kSRegSize, CPURegister::kFPRegister);
+  }
+  inline void PopSRegList(RegList regs) {
+    PopSizeRegList(regs, kSRegSize, CPURegister::kFPRegister);
+  }
+
+  // Push the specified register 'count' times.
+  void PushMultipleTimes(int count, Register src);
+
+  // Poke 'src' onto the stack. The offset is in bytes.
+  //
+  // If the current stack pointer (as set by SetStackPointer) is sp, then sp
+  // must be aligned to 16 bytes.
+  void Poke(const Register& src, const Operand& offset);
+
+  // Peek at a value on the stack, and put it in 'dst'. The offset is in bytes.
+  //
+  // If the current stack pointer (as set by SetStackPointer) is sp, then sp
+  // must be aligned to 16 bytes.
+  void Peek(const Register& dst, const Operand& offset);
+
+  // Claim or drop stack space without actually accessing memory.
+  //
+  // If the current stack pointer (as set by SetStackPointer) is sp, then it
+  // must be aligned to 16 bytes and the size claimed or dropped must be a
+  // multiple of 16 bytes.
+  void Claim(const Operand& size);
+  void Drop(const Operand& size);
+
+  // Preserve the callee-saved registers (as defined by AAPCS64).
+  //
+  // Higher-numbered registers are pushed before lower-numbered registers, and
+  // thus get higher addresses.
+  // Floating-point registers are pushed before general-purpose registers, and
+  // thus get higher addresses.
+  //
+  // This method must not be called unless StackPointer() is sp, and it is
+  // aligned to 16 bytes.
+  void PushCalleeSavedRegisters();
+
+  // Restore the callee-saved registers (as defined by AAPCS64).
+  //
+  // Higher-numbered registers are popped after lower-numbered registers, and
+  // thus come from higher addresses.
+  // Floating-point registers are popped after general-purpose registers, and
+  // thus come from higher addresses.
+  //
+  // This method must not be called unless StackPointer() is sp, and it is
+  // aligned to 16 bytes.
+  void PopCalleeSavedRegisters();
+
+  // Remaining instructions are simple pass-through calls to the assembler.
+  void Adr(const Register& rd, Label* label) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    adr(rd, label);
+  }
+  void Asr(const Register& rd, const Register& rn, unsigned shift) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    asr(rd, rn, shift);
+  }
+  void Asr(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    asrv(rd, rn, rm);
+  }
+  void B(Label* label) {
+    b(label);
+  }
+  void B(Label* label, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT((cond != al) && (cond != nv));
+    b(label, cond);
+  }
+  void Bfi(const Register& rd,
+           const Register& rn,
+           unsigned lsb,
+           unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    bfi(rd, rn, lsb, width);
+  }
+  void Bfxil(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    bfxil(rd, rn, lsb, width);
+  }
+  void Bind(Label* label) {
+    ASSERT(allow_macro_instructions_);
+    bind(label);
+  }
+  void Bl(Label* label) {
+    ASSERT(allow_macro_instructions_);
+    bl(label);
+  }
+  void Blr(const Register& xn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!xn.IsZero());
+    blr(xn);
+  }
+  void Br(const Register& xn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!xn.IsZero());
+    br(xn);
+  }
+  void Brk(int code = 0) {
+    ASSERT(allow_macro_instructions_);
+    brk(code);
+  }
+  void Cbnz(const Register& rt, Label* label) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    cbnz(rt, label);
+  }
+  void Cbz(const Register& rt, Label* label) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    cbz(rt, label);
+  }
+  void Cinc(const Register& rd, const Register& rn, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    cinc(rd, rn, cond);
+  }
+  void Cinv(const Register& rd, const Register& rn, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    cinv(rd, rn, cond);
+  }
+  void Cls(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    cls(rd, rn);
+  }
+  void Clz(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    clz(rd, rn);
+  }
+  void Cneg(const Register& rd, const Register& rn, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    cneg(rd, rn, cond);
+  }
+  void Csel(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT((cond != al) && (cond != nv));
+    csel(rd, rn, rm, cond);
+  }
+  void Cset(const Register& rd, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    cset(rd, cond);
+  }
+  void Csetm(const Register& rd, Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    csetm(rd, cond);
+  }
+  void Csinc(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT((cond != al) && (cond != nv));
+    csinc(rd, rn, rm, cond);
+  }
+  void Csinv(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT((cond != al) && (cond != nv));
+    csinv(rd, rn, rm, cond);
+  }
+  void Csneg(const Register& rd,
+             const Register& rn,
+             const Register& rm,
+             Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT((cond != al) && (cond != nv));
+    csneg(rd, rn, rm, cond);
+  }
+  void Extr(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            unsigned lsb) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    extr(rd, rn, rm, lsb);
+  }
+  void Fabs(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    fabs(fd, fn);
+  }
+  void Fadd(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fadd(fd, fn, fm);
+  }
+  void Fccmp(const FPRegister& fn,
+             const FPRegister& fm,
+             StatusFlags nzcv,
+             Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT((cond != al) && (cond != nv));
+    fccmp(fn, fm, nzcv, cond);
+  }
+  void Fcmp(const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fcmp(fn, fm);
+  }
+  void Fcmp(const FPRegister& fn, double value) {
+    ASSERT(allow_macro_instructions_);
+    if (value != 0.0) {
+      FPRegister tmp = AppropriateTempFor(fn);
+      Fmov(tmp, value);
+      fcmp(fn, tmp);
+    } else {
+      fcmp(fn, value);
+    }
+  }
+  void Fcsel(const FPRegister& fd,
+             const FPRegister& fn,
+             const FPRegister& fm,
+             Condition cond) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT((cond != al) && (cond != nv));
+    fcsel(fd, fn, fm, cond);
+  }
+  void Fcvt(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    fcvt(fd, fn);
+  }
+  void Fcvtms(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtms(rd, fn);
+  }
+  void Fcvtmu(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtmu(rd, fn);
+  }
+  void Fcvtns(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtns(rd, fn);
+  }
+  void Fcvtnu(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtnu(rd, fn);
+  }
+  void Fcvtzs(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtzs(rd, fn);
+  }
+  void Fcvtzu(const Register& rd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fcvtzu(rd, fn);
+  }
+  void Fdiv(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fdiv(fd, fn, fm);
+  }
+  void Fmax(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fmax(fd, fn, fm);
+  }
+  void Fmin(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fmin(fd, fn, fm);
+  }
+  void Fmov(FPRegister fd, FPRegister fn) {
+    ASSERT(allow_macro_instructions_);
+    // Only emit an instruction if fd and fn are different, and they are both D
+    // registers. fmov(s0, s0) is not a no-op because it clears the top word of
+    // d0. Technically, fmov(d0, d0) is not a no-op either because it clears
+    // the top of q0, but FPRegister does not currently support Q registers.
+    if (!fd.Is(fn) || !fd.Is64Bits()) {
+      fmov(fd, fn);
+    }
+  }
+  void Fmov(FPRegister fd, Register rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rn.IsZero());
+    fmov(fd, rn);
+  }
+  void Fmov(FPRegister fd, double imm) {
+    ASSERT(allow_macro_instructions_);
+    fmov(fd, imm);
+  }
+  void Fmov(Register rd, FPRegister fn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    fmov(rd, fn);
+  }
+  void Fmul(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fmul(fd, fn, fm);
+  }
+  void Fmsub(const FPRegister& fd,
+             const FPRegister& fn,
+             const FPRegister& fm,
+             const FPRegister& fa) {
+    ASSERT(allow_macro_instructions_);
+    fmsub(fd, fn, fm, fa);
+  }
+  void Fneg(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    fneg(fd, fn);
+  }
+  void Frintn(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    frintn(fd, fn);
+  }
+  void Frintz(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    frintz(fd, fn);
+  }
+  void Fsqrt(const FPRegister& fd, const FPRegister& fn) {
+    ASSERT(allow_macro_instructions_);
+    fsqrt(fd, fn);
+  }
+  void Fsub(const FPRegister& fd, const FPRegister& fn, const FPRegister& fm) {
+    ASSERT(allow_macro_instructions_);
+    fsub(fd, fn, fm);
+  }
+  void Hint(SystemHint code) {
+    ASSERT(allow_macro_instructions_);
+    hint(code);
+  }
+  void Hlt(int code) {
+    ASSERT(allow_macro_instructions_);
+    hlt(code);
+  }
+  void Ldnp(const CPURegister& rt,
+            const CPURegister& rt2,
+            const MemOperand& src) {
+    ASSERT(allow_macro_instructions_);
+    ldnp(rt, rt2, src);
+  }
+  void Ldp(const CPURegister& rt,
+           const CPURegister& rt2,
+           const MemOperand& src) {
+    ASSERT(allow_macro_instructions_);
+    ldp(rt, rt2, src);
+  }
+  void Ldpsw(const Register& rt, const Register& rt2, const MemOperand& src) {
+    ASSERT(allow_macro_instructions_);
+    ldpsw(rt, rt2, src);
+  }
+  void Ldr(const FPRegister& ft, double imm) {
+    ASSERT(allow_macro_instructions_);
+    ldr(ft, imm);
+  }
+  void Ldr(const Register& rt, uint64_t imm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    ldr(rt, imm);
+  }
+  void Lsl(const Register& rd, const Register& rn, unsigned shift) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    lsl(rd, rn, shift);
+  }
+  void Lsl(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    lslv(rd, rn, rm);
+  }
+  void Lsr(const Register& rd, const Register& rn, unsigned shift) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    lsr(rd, rn, shift);
+  }
+  void Lsr(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    lsrv(rd, rn, rm);
+  }
+  void Madd(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    madd(rd, rn, rm, ra);
+  }
+  void Mneg(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    mneg(rd, rn, rm);
+  }
+  void Mov(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    mov(rd, rn);
+  }
+  void Mrs(const Register& rt, SystemRegister sysreg) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    mrs(rt, sysreg);
+  }
+  void Msr(SystemRegister sysreg, const Register& rt) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    msr(sysreg, rt);
+  }
+  void Msub(const Register& rd,
+            const Register& rn,
+            const Register& rm,
+            const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    msub(rd, rn, rm, ra);
+  }
+  void Mul(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    mul(rd, rn, rm);
+  }
+  void Nop() {
+    ASSERT(allow_macro_instructions_);
+    nop();
+  }
+  void Rbit(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    rbit(rd, rn);
+  }
+  void Ret(const Register& xn = lr) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!xn.IsZero());
+    ret(xn);
+  }
+  void Rev(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    rev(rd, rn);
+  }
+  void Rev16(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    rev16(rd, rn);
+  }
+  void Rev32(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    rev32(rd, rn);
+  }
+  void Ror(const Register& rd, const Register& rs, unsigned shift) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rs.IsZero());
+    ror(rd, rs, shift);
+  }
+  void Ror(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    rorv(rd, rn, rm);
+  }
+  void Sbfiz(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    sbfiz(rd, rn, lsb, width);
+  }
+  void Sbfx(const Register& rd,
+            const Register& rn,
+            unsigned lsb,
+            unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    sbfx(rd, rn, lsb, width);
+  }
+  void Scvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rn.IsZero());
+    scvtf(fd, rn, fbits);
+  }
+  void Sdiv(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    sdiv(rd, rn, rm);
+  }
+  void Smaddl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    smaddl(rd, rn, rm, ra);
+  }
+  void Smsubl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    smsubl(rd, rn, rm, ra);
+  }
+  void Smull(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    smull(rd, rn, rm);
+  }
+  void Smulh(const Register& xd, const Register& xn, const Register& xm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!xd.IsZero());
+    ASSERT(!xn.IsZero());
+    ASSERT(!xm.IsZero());
+    smulh(xd, xn, xm);
+  }
+  void Stnp(const CPURegister& rt,
+            const CPURegister& rt2,
+            const MemOperand& dst) {
+    ASSERT(allow_macro_instructions_);
+    stnp(rt, rt2, dst);
+  }
+  void Stp(const CPURegister& rt,
+           const CPURegister& rt2,
+           const MemOperand& dst) {
+    ASSERT(allow_macro_instructions_);
+    stp(rt, rt2, dst);
+  }
+  void Sxtb(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    sxtb(rd, rn);
+  }
+  void Sxth(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    sxth(rd, rn);
+  }
+  void Sxtw(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    sxtw(rd, rn);
+  }
+  void Tbnz(const Register& rt, unsigned bit_pos, Label* label) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    tbnz(rt, bit_pos, label);
+  }
+  void Tbz(const Register& rt, unsigned bit_pos, Label* label) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rt.IsZero());
+    tbz(rt, bit_pos, label);
+  }
+  void Ubfiz(const Register& rd,
+             const Register& rn,
+             unsigned lsb,
+             unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ubfiz(rd, rn, lsb, width);
+  }
+  void Ubfx(const Register& rd,
+            const Register& rn,
+            unsigned lsb,
+            unsigned width) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ubfx(rd, rn, lsb, width);
+  }
+  void Ucvtf(const FPRegister& fd, const Register& rn, unsigned fbits = 0) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rn.IsZero());
+    ucvtf(fd, rn, fbits);
+  }
+  void Udiv(const Register& rd, const Register& rn, const Register& rm) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    udiv(rd, rn, rm);
+  }
+  void Umaddl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    umaddl(rd, rn, rm, ra);
+  }
+  void Umsubl(const Register& rd,
+              const Register& rn,
+              const Register& rm,
+              const Register& ra) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    ASSERT(!rm.IsZero());
+    ASSERT(!ra.IsZero());
+    umsubl(rd, rn, rm, ra);
+  }
+  void Unreachable() {
+    ASSERT(allow_macro_instructions_);
+#ifdef USE_SIMULATOR
+    hlt(kUnreachableOpcode);
+#else
+    // Branch to 0 to generate a segfault.
+    // lr - kInstructionSize is the address of the offending instruction.
+    blr(xzr);
+#endif
+  }
+  void Uxtb(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    uxtb(rd, rn);
+  }
+  void Uxth(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    uxth(rd, rn);
+  }
+  void Uxtw(const Register& rd, const Register& rn) {
+    ASSERT(allow_macro_instructions_);
+    ASSERT(!rd.IsZero());
+    ASSERT(!rn.IsZero());
+    uxtw(rd, rn);
+  }
+
+  // Push the system stack pointer (sp) down to allow the same to be done to
+  // the current stack pointer (according to StackPointer()). This must be
+  // called _before_ accessing the memory.
+  //
+  // This is necessary when pushing or otherwise adding things to the stack, to
+  // satisfy the AAPCS64 constraint that the memory below the system stack
+  // pointer is not accessed.
+  //
+  // This method asserts that StackPointer() is not sp, since the call does
+  // not make sense in that context.
+  //
+  // TODO: This method can only accept values of 'space' that can be encoded in
+  // one instruction. Refer to the implementation for details.
+  void BumpSystemStackPointer(const Operand& space);
+
+#if DEBUG
+  void SetAllowMacroInstructions(bool value) {
+    allow_macro_instructions_ = value;
+  }
+
+  bool AllowMacroInstructions() const {
+    return allow_macro_instructions_;
+  }
+#endif
+
+  // Set the current stack pointer, but don't generate any code.
+  // Note that this does not directly affect LastStackPointer().
+  void SetStackPointer(const Register& stack_pointer) {
+    ASSERT(!AreAliased(stack_pointer, Tmp0(), Tmp1()));
+    sp_ = stack_pointer;
+  }
+
+  // Return the current stack pointer, as set by SetStackPointer.
+  const Register& StackPointer() const {
+    return sp_;
+  }
+
+  // Set the registers used internally by the MacroAssembler as scratch
+  // registers. These registers are used to implement behaviours which are not
+  // directly supported by A64, and where an intermediate result is required.
+  //
+  // Both tmp0 and tmp1 may be set to any X register except for xzr, sp,
+  // and StackPointer(). Also, they must not be the same register (though they
+  // may both be NoReg).
+  //
+  // It is valid to set either or both of these registers to NoReg if you don't
+  // want the MacroAssembler to use any scratch registers. In a debug build, the
+  // Assembler will assert that any registers it uses are valid. Be aware that
+  // this check is not present in release builds. If this is a problem, use the
+  // Assembler directly.
+  void SetScratchRegisters(const Register& tmp0, const Register& tmp1) {
+    ASSERT(!AreAliased(xzr, sp, tmp0, tmp1));
+    ASSERT(!AreAliased(StackPointer(), tmp0, tmp1));
+    tmp0_ = tmp0;
+    tmp1_ = tmp1;
+  }
+
+  const Register& Tmp0() const {
+    return tmp0_;
+  }
+
+  const Register& Tmp1() const {
+    return tmp1_;
+  }
+
+  void SetFPScratchRegister(const FPRegister& fptmp0) {
+    fptmp0_ = fptmp0;
+  }
+
+  const FPRegister& FPTmp0() const {
+    return fptmp0_;
+  }
+
+  const Register AppropriateTempFor(
+      const Register& target,
+      const CPURegister& forbidden = NoCPUReg) const {
+    Register candidate = forbidden.Is(Tmp0()) ? Tmp1() : Tmp0();
+    ASSERT(!candidate.Is(target));
+    return Register(candidate.code(), target.size());
+  }
+
+  const FPRegister AppropriateTempFor(
+      const FPRegister& target,
+      const CPURegister& forbidden = NoCPUReg) const {
+    USE(forbidden);
+    FPRegister candidate = FPTmp0();
+    ASSERT(!candidate.Is(forbidden));
+    ASSERT(!candidate.Is(target));
+    return FPRegister(candidate.code(), target.size());
+  }
+
+  // Like printf, but print at run-time from generated code.
+  //
+  // The caller must ensure that arguments for floating-point placeholders
+  // (such as %e, %f or %g) are FPRegisters, and that arguments for integer
+  // placeholders are Registers.
+  //
+  // A maximum of four arguments may be given to any single Printf call. The
+  // arguments must be of the same type, but they do not need to have the same
+  // size.
+  //
+  // The following registers cannot be printed:
+  //    Tmp0(), Tmp1(), StackPointer(), sp.
+  //
+  // This function automatically preserves caller-saved registers so that
+  // calling code can use Printf at any point without having to worry about
+  // corruption. The preservation mechanism generates a lot of code. If this is
+  // a problem, preserve the important registers manually and then call
+  // PrintfNoPreserve. Callee-saved registers are not used by Printf, and are
+  // implicitly preserved.
+  //
+  // This function assumes (and asserts) that the current stack pointer is
+  // callee-saved, not caller-saved. This is most likely the case anyway, as a
+  // caller-saved stack pointer doesn't make a lot of sense.
+  void Printf(const char * format,
+              const CPURegister& arg0 = NoCPUReg,
+              const CPURegister& arg1 = NoCPUReg,
+              const CPURegister& arg2 = NoCPUReg,
+              const CPURegister& arg3 = NoCPUReg);
+
+  // Like Printf, but don't preserve any caller-saved registers, not even 'lr'.
+  //
+  // The return code from the system printf call will be returned in x0.
+  void PrintfNoPreserve(const char * format,
+                        const CPURegister& arg0 = NoCPUReg,
+                        const CPURegister& arg1 = NoCPUReg,
+                        const CPURegister& arg2 = NoCPUReg,
+                        const CPURegister& arg3 = NoCPUReg);
+
+  // Trace control when running the debug simulator.
+  //
+  // For example:
+  //
+  // __ Trace(LOG_REGS, TRACE_ENABLE);
+  // Will add registers to the trace if it wasn't already the case.
+  //
+  // __ Trace(LOG_DISASM, TRACE_DISABLE);
+  // Will stop logging disassembly. It has no effect if the disassembly wasn't
+  // already being logged.
+  void Trace(TraceParameters parameters, TraceCommand command);
+
+  // Log the requested data independently of what is being traced.
+  //
+  // For example:
+  //
+  // __ Log(LOG_FLAGS)
+  // Will output the flags.
+  void Log(TraceParameters parameters);
+
+  // Enable or disable instrumentation when an Instrument visitor is attached to
+  // the simulator.
+  void EnableInstrumentation();
+  void DisableInstrumentation();
+
+  // Add a marker to the instrumentation data produced by an Instrument visitor.
+  // The name is a two character string that will be attached to the marker in
+  // the output data.
+  void AnnotateInstrumentation(const char* marker_name);
+
+ private:
+  // The actual Push and Pop implementations. These don't generate any code
+  // other than that required for the push or pop. This allows
+  // (Push|Pop)CPURegList to bundle together setup code for a large block of
+  // registers.
+  //
+  // Note that size is per register, and is specified in bytes.
+  void PushHelper(int count, int size,
+                  const CPURegister& src0, const CPURegister& src1,
+                  const CPURegister& src2, const CPURegister& src3);
+  void PopHelper(int count, int size,
+                 const CPURegister& dst0, const CPURegister& dst1,
+                 const CPURegister& dst2, const CPURegister& dst3);
+
+  // Perform necessary maintenance operations before a push or pop.
+  //
+  // Note that size is per register, and is specified in bytes.
+  void PrepareForPush(int count, int size);
+  void PrepareForPop(int count, int size);
+
+#if DEBUG
+  // Tell whether any of the macro instruction can be used. When false the
+  // MacroAssembler will assert if a method which can emit a variable number
+  // of instructions is called.
+  bool allow_macro_instructions_;
+#endif
+
+  // The register to use as a stack pointer for stack operations.
+  Register sp_;
+
+  // Scratch registers used internally by the MacroAssembler.
+  Register tmp0_;
+  Register tmp1_;
+  FPRegister fptmp0_;
+};
+
+
+// Use this scope when you need a one-to-one mapping between methods and
+// instructions. This scope prevents the MacroAssembler from being called and
+// literal pools from being emitted. It also asserts the number of instructions
+// emitted is what you specified when creating the scope.
+class InstructionAccurateScope {
+ public:
+  explicit InstructionAccurateScope(MacroAssembler* masm)
+      : masm_(masm), size_(0) {
+    masm_->BlockLiteralPool();
+#ifdef DEBUG
+    old_allow_macro_instructions_ = masm_->AllowMacroInstructions();
+    masm_->SetAllowMacroInstructions(false);
+#endif
+  }
+
+  InstructionAccurateScope(MacroAssembler* masm, int count)
+      : masm_(masm), size_(count * kInstructionSize) {
+    masm_->BlockLiteralPool();
+#ifdef DEBUG
+    masm_->bind(&start_);
+    old_allow_macro_instructions_ = masm_->AllowMacroInstructions();
+    masm_->SetAllowMacroInstructions(false);
+#endif
+  }
+
+  ~InstructionAccurateScope() {
+    masm_->ReleaseLiteralPool();
+#ifdef DEBUG
+    if (start_.IsBound()) {
+      ASSERT(masm_->SizeOfCodeGeneratedSince(&start_) == size_);
+    }
+    masm_->SetAllowMacroInstructions(old_allow_macro_instructions_);
+#endif
+  }
+
+ private:
+  MacroAssembler* masm_;
+  uint64_t size_;
+#ifdef DEBUG
+  Label start_;
+  bool old_allow_macro_instructions_;
+#endif
+};
+
+
+}  // namespace vixl
+
+#endif  // VIXL_A64_MACRO_ASSEMBLER_A64_H_
diff --git a/disas/libvixl/src/a64/simulator-a64.cc b/disas/libvixl/src/a64/simulator-a64.cc
new file mode 100644
index 0000000..f08e0ed
--- /dev/null
+++ b/disas/libvixl/src/a64/simulator-a64.cc
@@ -0,0 +1,2077 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include <math.h>
+#include "a64/simulator-a64.h"
+
+namespace vixl {
+
+const Instruction* Simulator::kEndOfSimAddress = NULL;
+
+void SimSystemRegister::SetBits(int msb, int lsb, uint32_t bits) {
+  int width = msb - lsb + 1;
+  ASSERT(is_uintn(width, bits) || is_intn(width, bits));
+
+  bits <<= lsb;
+  uint32_t mask = ((1 << width) - 1) << lsb;
+  ASSERT((mask & write_ignore_mask_) == 0);
+
+  value_ = (value_ & ~mask) | (bits & mask);
+}
+
+
+SimSystemRegister SimSystemRegister::DefaultValueFor(SystemRegister id) {
+  switch (id) {
+    case NZCV:
+      return SimSystemRegister(0x00000000, NZCVWriteIgnoreMask);
+    case FPCR:
+      return SimSystemRegister(0x00000000, FPCRWriteIgnoreMask);
+    default:
+      UNREACHABLE();
+      return SimSystemRegister();
+  }
+}
+
+
+Simulator::Simulator(Decoder* decoder, FILE* stream) {
+  // Ensure shift operations act as the simulator expects.
+  ASSERT((static_cast<int32_t>(-1) >> 1) == -1);
+  ASSERT((static_cast<uint32_t>(-1) >> 1) == 0x7FFFFFFF);
+
+  // Setup the decoder.
+  decoder_ = decoder;
+  decoder_->AppendVisitor(this);
+
+  ResetState();
+
+  // Allocate and setup the simulator stack.
+  stack_ = reinterpret_cast<byte*>(malloc(stack_size_));
+  stack_limit_ = stack_ + stack_protection_size_;
+  byte* tos = stack_ + stack_size_ - stack_protection_size_;
+  // The stack pointer must be 16 bytes aligned.
+  set_sp(reinterpret_cast<int64_t>(tos) & ~0xfUL);
+
+  stream_ = stream;
+  print_disasm_ = new PrintDisassembler(stream_);
+  coloured_trace_ = false;
+  disasm_trace_ = false;
+
+  // Set the sample period to 10, as the VIXL examples and tests are short.
+  instrumentation_ = new Instrument("vixl_stats.csv", 10);
+}
+
+
+void Simulator::ResetState() {
+  // Reset the system registers.
+  nzcv_ = SimSystemRegister::DefaultValueFor(NZCV);
+  fpcr_ = SimSystemRegister::DefaultValueFor(FPCR);
+
+  // Reset registers to 0.
+  pc_ = NULL;
+  pc_modified_ = false;
+  for (unsigned i = 0; i < kNumberOfRegisters; i++) {
+    set_xreg(i, 0xbadbeef);
+  }
+  for (unsigned i = 0; i < kNumberOfFPRegisters; i++) {
+    // Set FP registers to a value that is NaN in both 32-bit and 64-bit FP.
+    set_dreg_bits(i, 0x7ff000007f800001UL);
+  }
+  // Returning to address 0 exits the Simulator.
+  set_lr(reinterpret_cast<int64_t>(kEndOfSimAddress));
+}
+
+
+Simulator::~Simulator() {
+  free(stack_);
+  // The decoder may outlive the simulator.
+  decoder_->RemoveVisitor(print_disasm_);
+  delete print_disasm_;
+
+  decoder_->RemoveVisitor(instrumentation_);
+  delete instrumentation_;
+}
+
+
+void Simulator::Run() {
+  while (pc_ != kEndOfSimAddress) {
+    ExecuteInstruction();
+  }
+}
+
+
+void Simulator::RunFrom(Instruction* first) {
+  pc_ = first;
+  pc_modified_ = false;
+  Run();
+}
+
+
+const char* Simulator::xreg_names[] = {
+"x0",  "x1",  "x2",  "x3",  "x4",  "x5",  "x6",  "x7",
+"x8",  "x9",  "x10", "x11", "x12", "x13", "x14", "x15",
+"x16", "x17", "x18", "x19", "x20", "x21", "x22", "x23",
+"x24", "x25", "x26", "x27", "x28", "x29", "lr",  "xzr", "sp"};
+
+
+const char* Simulator::wreg_names[] = {
+"w0",  "w1",  "w2",  "w3",  "w4",  "w5",  "w6",  "w7",
+"w8",  "w9",  "w10", "w11", "w12", "w13", "w14", "w15",
+"w16", "w17", "w18", "w19", "w20", "w21", "w22", "w23",
+"w24", "w25", "w26", "w27", "w28", "w29", "w30", "wzr", "wsp"};
+
+const char* Simulator::sreg_names[] = {
+"s0",  "s1",  "s2",  "s3",  "s4",  "s5",  "s6",  "s7",
+"s8",  "s9",  "s10", "s11", "s12", "s13", "s14", "s15",
+"s16", "s17", "s18", "s19", "s20", "s21", "s22", "s23",
+"s24", "s25", "s26", "s27", "s28", "s29", "s30", "s31"};
+
+const char* Simulator::dreg_names[] = {
+"d0",  "d1",  "d2",  "d3",  "d4",  "d5",  "d6",  "d7",
+"d8",  "d9",  "d10", "d11", "d12", "d13", "d14", "d15",
+"d16", "d17", "d18", "d19", "d20", "d21", "d22", "d23",
+"d24", "d25", "d26", "d27", "d28", "d29", "d30", "d31"};
+
+const char* Simulator::vreg_names[] = {
+"v0",  "v1",  "v2",  "v3",  "v4",  "v5",  "v6",  "v7",
+"v8",  "v9",  "v10", "v11", "v12", "v13", "v14", "v15",
+"v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23",
+"v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31"};
+
+
+
+const char* Simulator::WRegNameForCode(unsigned code, Reg31Mode mode) {
+  ASSERT(code < kNumberOfRegisters);
+  // If the code represents the stack pointer, index the name after zr.
+  if ((code == kZeroRegCode) && (mode == Reg31IsStackPointer)) {
+    code = kZeroRegCode + 1;
+  }
+  return wreg_names[code];
+}
+
+
+const char* Simulator::XRegNameForCode(unsigned code, Reg31Mode mode) {
+  ASSERT(code < kNumberOfRegisters);
+  // If the code represents the stack pointer, index the name after zr.
+  if ((code == kZeroRegCode) && (mode == Reg31IsStackPointer)) {
+    code = kZeroRegCode + 1;
+  }
+  return xreg_names[code];
+}
+
+
+const char* Simulator::SRegNameForCode(unsigned code) {
+  ASSERT(code < kNumberOfFPRegisters);
+  return sreg_names[code];
+}
+
+
+const char* Simulator::DRegNameForCode(unsigned code) {
+  ASSERT(code < kNumberOfFPRegisters);
+  return dreg_names[code];
+}
+
+
+const char* Simulator::VRegNameForCode(unsigned code) {
+  ASSERT(code < kNumberOfFPRegisters);
+  return vreg_names[code];
+}
+
+
+// Helpers ---------------------------------------------------------------------
+int64_t Simulator::AddWithCarry(unsigned reg_size,
+                                bool set_flags,
+                                int64_t src1,
+                                int64_t src2,
+                                int64_t carry_in) {
+  ASSERT((carry_in == 0) || (carry_in == 1));
+  ASSERT((reg_size == kXRegSize) || (reg_size == kWRegSize));
+
+  uint64_t u1, u2;
+  int64_t result;
+  int64_t signed_sum = src1 + src2 + carry_in;
+
+  uint32_t N, Z, C, V;
+
+  if (reg_size == kWRegSize) {
+    u1 = static_cast<uint64_t>(src1) & kWRegMask;
+    u2 = static_cast<uint64_t>(src2) & kWRegMask;
+
+    result = signed_sum & kWRegMask;
+    // Compute the C flag by comparing the sum to the max unsigned integer.
+    C = ((kWMaxUInt - u1) < (u2 + carry_in)) ||
+        ((kWMaxUInt - u1 - carry_in) < u2);
+    // Overflow iff the sign bit is the same for the two inputs and different
+    // for the result.
+    int64_t s_src1 = src1 << (kXRegSize - kWRegSize);
+    int64_t s_src2 = src2 << (kXRegSize - kWRegSize);
+    int64_t s_result = result << (kXRegSize - kWRegSize);
+    V = ((s_src1 ^ s_src2) >= 0) && ((s_src1 ^ s_result) < 0);
+
+  } else {
+    u1 = static_cast<uint64_t>(src1);
+    u2 = static_cast<uint64_t>(src2);
+
+    result = signed_sum;
+    // Compute the C flag by comparing the sum to the max unsigned integer.
+    C = ((kXMaxUInt - u1) < (u2 + carry_in)) ||
+        ((kXMaxUInt - u1 - carry_in) < u2);
+    // Overflow iff the sign bit is the same for the two inputs and different
+    // for the result.
+    V = ((src1 ^ src2) >= 0) && ((src1 ^ result) < 0);
+  }
+
+  N = CalcNFlag(result, reg_size);
+  Z = CalcZFlag(result);
+
+  if (set_flags) {
+    nzcv().SetN(N);
+    nzcv().SetZ(Z);
+    nzcv().SetC(C);
+    nzcv().SetV(V);
+  }
+  return result;
+}
+
+
+int64_t Simulator::ShiftOperand(unsigned reg_size,
+                                int64_t value,
+                                Shift shift_type,
+                                unsigned amount) {
+  if (amount == 0) {
+    return value;
+  }
+  int64_t mask = reg_size == kXRegSize ? kXRegMask : kWRegMask;
+  switch (shift_type) {
+    case LSL:
+      return (value << amount) & mask;
+    case LSR:
+      return static_cast<uint64_t>(value) >> amount;
+    case ASR: {
+      // Shift used to restore the sign.
+      unsigned s_shift = kXRegSize - reg_size;
+      // Value with its sign restored.
+      int64_t s_value = (value << s_shift) >> s_shift;
+      return (s_value >> amount) & mask;
+    }
+    case ROR: {
+      if (reg_size == kWRegSize) {
+        value &= kWRegMask;
+      }
+      return (static_cast<uint64_t>(value) >> amount) |
+             ((value & ((1L << amount) - 1L)) << (reg_size - amount));
+    }
+    default:
+      UNIMPLEMENTED();
+      return 0;
+  }
+}
+
+
+int64_t Simulator::ExtendValue(unsigned reg_size,
+                               int64_t value,
+                               Extend extend_type,
+                               unsigned left_shift) {
+  switch (extend_type) {
+    case UXTB:
+      value &= kByteMask;
+      break;
+    case UXTH:
+      value &= kHalfWordMask;
+      break;
+    case UXTW:
+      value &= kWordMask;
+      break;
+    case SXTB:
+      value = (value << 56) >> 56;
+      break;
+    case SXTH:
+      value = (value << 48) >> 48;
+      break;
+    case SXTW:
+      value = (value << 32) >> 32;
+      break;
+    case UXTX:
+    case SXTX:
+      break;
+    default:
+      UNREACHABLE();
+  }
+  int64_t mask = (reg_size == kXRegSize) ? kXRegMask : kWRegMask;
+  return (value << left_shift) & mask;
+}
+
+
+void Simulator::FPCompare(double val0, double val1) {
+  AssertSupportedFPCR();
+
+  // TODO: This assumes that the C++ implementation handles comparisons in the
+  // way that we expect (as per AssertSupportedFPCR()).
+  if ((isnan(val0) != 0) || (isnan(val1) != 0)) {
+    nzcv().SetRawValue(FPUnorderedFlag);
+  } else if (val0 < val1) {
+    nzcv().SetRawValue(FPLessThanFlag);
+  } else if (val0 > val1) {
+    nzcv().SetRawValue(FPGreaterThanFlag);
+  } else if (val0 == val1) {
+    nzcv().SetRawValue(FPEqualFlag);
+  } else {
+    UNREACHABLE();
+  }
+}
+
+
+void Simulator::PrintSystemRegisters(bool print_all) {
+  static bool first_run = true;
+
+  // Define some colour codes to use for the register dump.
+  // TODO: Find a more elegant way of defining these.
+  char const * const clr_normal     = (coloured_trace_) ? ("\033[m") : ("");
+  char const * const clr_flag_name  = (coloured_trace_) ? ("\033[1;30m") : ("");
+  char const * const clr_flag_value = (coloured_trace_) ? ("\033[1;37m") : ("");
+
+  static SimSystemRegister last_nzcv;
+  if (print_all || first_run || (last_nzcv.RawValue() != nzcv().RawValue())) {
+    fprintf(stream_, "# %sFLAGS: %sN:%d Z:%d C:%d V:%d%s\n",
+            clr_flag_name,
+            clr_flag_value,
+            N(), Z(), C(), V(),
+            clr_normal);
+  }
+  last_nzcv = nzcv();
+
+  static SimSystemRegister last_fpcr;
+  if (print_all || first_run || (last_fpcr.RawValue() != fpcr().RawValue())) {
+    static const char * rmode[] = {
+      "0b00 (Round to Nearest)",
+      "0b01 (Round towards Plus Infinity)",
+      "0b10 (Round towards Minus Infinity)",
+      "0b11 (Round towards Zero)"
+    };
+    ASSERT(fpcr().RMode() <= (sizeof(rmode) / sizeof(rmode[0])));
+    fprintf(stream_, "# %sFPCR: %sAHP:%d DN:%d FZ:%d RMode:%s%s\n",
+            clr_flag_name,
+            clr_flag_value,
+            fpcr().AHP(), fpcr().DN(), fpcr().FZ(), rmode[fpcr().RMode()],
+            clr_normal);
+  }
+  last_fpcr = fpcr();
+
+  first_run = false;
+}
+
+
+void Simulator::PrintRegisters(bool print_all_regs) {
+  static bool first_run = true;
+  static int64_t last_regs[kNumberOfRegisters];
+
+  // Define some colour codes to use for the register dump.
+  // TODO: Find a more elegant way of defining these.
+  char const * const clr_normal    = (coloured_trace_) ? ("\033[m") : ("");
+  char const * const clr_reg_name  = (coloured_trace_) ? ("\033[1;34m") : ("");
+  char const * const clr_reg_value = (coloured_trace_) ? ("\033[1;36m") : ("");
+
+  for (unsigned i = 0; i < kNumberOfRegisters; i++) {
+    if (print_all_regs || first_run || (last_regs[i] != registers_[i].x)) {
+      fprintf(stream_,
+              "# %s%4s:%s 0x%016" PRIx64 "%s\n",
+              clr_reg_name,
+              XRegNameForCode(i, Reg31IsStackPointer),
+              clr_reg_value,
+              registers_[i].x,
+              clr_normal);
+    }
+    // Cache the new register value so the next run can detect any changes.
+    last_regs[i] = registers_[i].x;
+  }
+  first_run = false;
+}
+
+
+void Simulator::PrintFPRegisters(bool print_all_regs) {
+  static bool first_run = true;
+  static uint64_t last_regs[kNumberOfFPRegisters];
+
+  // Define some colour codes to use for the register dump.
+  // TODO: Find a more elegant way of defining these.
+  char const * const clr_normal    = (coloured_trace_) ? ("\033[m") : ("");
+  char const * const clr_reg_name  = (coloured_trace_) ? ("\033[1;33m") : ("");
+  char const * const clr_reg_value = (coloured_trace_) ? ("\033[1;35m") : ("");
+
+  // Print as many rows of registers as necessary, keeping each individual
+  // register in the same column each time (to make it easy to visually scan
+  // for changes).
+  for (unsigned i = 0; i < kNumberOfFPRegisters; i++) {
+    if (print_all_regs || first_run ||
+        (last_regs[i] != double_to_rawbits(fpregisters_[i].d))) {
+      fprintf(stream_,
+              "# %s %4s:%s 0x%016" PRIx64 "%s (%s%s:%s %g%s %s:%s %g%s)\n",
+              clr_reg_name,
+              VRegNameForCode(i),
+              clr_reg_value,
+              double_to_rawbits(fpregisters_[i].d),
+              clr_normal,
+              clr_reg_name,
+              DRegNameForCode(i),
+              clr_reg_value,
+              fpregisters_[i].d,
+              clr_reg_name,
+              SRegNameForCode(i),
+              clr_reg_value,
+              fpregisters_[i].s,
+              clr_normal);
+    }
+    // Cache the new register value so the next run can detect any changes.
+    last_regs[i] = double_to_rawbits(fpregisters_[i].d);
+  }
+  first_run = false;
+}
+
+
+void Simulator::PrintProcessorState() {
+  PrintSystemRegisters();
+  PrintRegisters();
+  PrintFPRegisters();
+}
+
+
+// Visitors---------------------------------------------------------------------
+
+void Simulator::VisitUnimplemented(Instruction* instr) {
+  printf("Unimplemented instruction at 0x%p: 0x%08" PRIx32 "\n",
+         reinterpret_cast<void*>(instr), instr->InstructionBits());
+  UNIMPLEMENTED();
+}
+
+
+void Simulator::VisitUnallocated(Instruction* instr) {
+  printf("Unallocated instruction at 0x%p: 0x%08" PRIx32 "\n",
+         reinterpret_cast<void*>(instr), instr->InstructionBits());
+  UNIMPLEMENTED();
+}
+
+
+void Simulator::VisitPCRelAddressing(Instruction* instr) {
+  switch (instr->Mask(PCRelAddressingMask)) {
+    case ADR:
+      set_reg(kXRegSize,
+              instr->Rd(),
+              reinterpret_cast<int64_t>(instr->ImmPCOffsetTarget()));
+      break;
+    case ADRP:  // Not implemented in the assembler.
+      UNIMPLEMENTED();
+      break;
+    default:
+      UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitUnconditionalBranch(Instruction* instr) {
+  switch (instr->Mask(UnconditionalBranchMask)) {
+    case BL:
+      set_lr(reinterpret_cast<int64_t>(instr->NextInstruction()));
+      // Fall through.
+    case B:
+      set_pc(instr->ImmPCOffsetTarget());
+      break;
+    default: UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitConditionalBranch(Instruction* instr) {
+  ASSERT(instr->Mask(ConditionalBranchMask) == B_cond);
+  if (ConditionPassed(static_cast<Condition>(instr->ConditionBranch()))) {
+    set_pc(instr->ImmPCOffsetTarget());
+  }
+}
+
+
+void Simulator::VisitUnconditionalBranchToRegister(Instruction* instr) {
+  Instruction* target = Instruction::Cast(xreg(instr->Rn()));
+
+  switch (instr->Mask(UnconditionalBranchToRegisterMask)) {
+    case BLR:
+      set_lr(reinterpret_cast<int64_t>(instr->NextInstruction()));
+      // Fall through.
+    case BR:
+    case RET: set_pc(target); break;
+    default: UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitTestBranch(Instruction* instr) {
+  unsigned bit_pos = (instr->ImmTestBranchBit5() << 5) |
+                     instr->ImmTestBranchBit40();
+  bool take_branch = ((xreg(instr->Rt()) & (1UL << bit_pos)) == 0);
+  switch (instr->Mask(TestBranchMask)) {
+    case TBZ: break;
+    case TBNZ: take_branch = !take_branch; break;
+    default: UNIMPLEMENTED();
+  }
+  if (take_branch) {
+    set_pc(instr->ImmPCOffsetTarget());
+  }
+}
+
+
+void Simulator::VisitCompareBranch(Instruction* instr) {
+  unsigned rt = instr->Rt();
+  bool take_branch = false;
+  switch (instr->Mask(CompareBranchMask)) {
+    case CBZ_w: take_branch = (wreg(rt) == 0); break;
+    case CBZ_x: take_branch = (xreg(rt) == 0); break;
+    case CBNZ_w: take_branch = (wreg(rt) != 0); break;
+    case CBNZ_x: take_branch = (xreg(rt) != 0); break;
+    default: UNIMPLEMENTED();
+  }
+  if (take_branch) {
+    set_pc(instr->ImmPCOffsetTarget());
+  }
+}
+
+
+void Simulator::AddSubHelper(Instruction* instr, int64_t op2) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  bool set_flags = instr->FlagsUpdate();
+  int64_t new_val = 0;
+  Instr operation = instr->Mask(AddSubOpMask);
+
+  switch (operation) {
+    case ADD:
+    case ADDS: {
+      new_val = AddWithCarry(reg_size,
+                             set_flags,
+                             reg(reg_size, instr->Rn(), instr->RnMode()),
+                             op2);
+      break;
+    }
+    case SUB:
+    case SUBS: {
+      new_val = AddWithCarry(reg_size,
+                             set_flags,
+                             reg(reg_size, instr->Rn(), instr->RnMode()),
+                             ~op2,
+                             1);
+      break;
+    }
+    default: UNREACHABLE();
+  }
+
+  set_reg(reg_size, instr->Rd(), new_val, instr->RdMode());
+}
+
+
+void Simulator::VisitAddSubShifted(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t op2 = ShiftOperand(reg_size,
+                             reg(reg_size, instr->Rm()),
+                             static_cast<Shift>(instr->ShiftDP()),
+                             instr->ImmDPShift());
+  AddSubHelper(instr, op2);
+}
+
+
+void Simulator::VisitAddSubImmediate(Instruction* instr) {
+  int64_t op2 = instr->ImmAddSub() << ((instr->ShiftAddSub() == 1) ? 12 : 0);
+  AddSubHelper(instr, op2);
+}
+
+
+void Simulator::VisitAddSubExtended(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t op2 = ExtendValue(reg_size,
+                            reg(reg_size, instr->Rm()),
+                            static_cast<Extend>(instr->ExtendMode()),
+                            instr->ImmExtendShift());
+  AddSubHelper(instr, op2);
+}
+
+
+void Simulator::VisitAddSubWithCarry(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t op2 = reg(reg_size, instr->Rm());
+  int64_t new_val;
+
+  if ((instr->Mask(AddSubOpMask) == SUB) || instr->Mask(AddSubOpMask) == SUBS) {
+    op2 = ~op2;
+  }
+
+  new_val = AddWithCarry(reg_size,
+                         instr->FlagsUpdate(),
+                         reg(reg_size, instr->Rn()),
+                         op2,
+                         C());
+
+  set_reg(reg_size, instr->Rd(), new_val);
+}
+
+
+void Simulator::VisitLogicalShifted(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  Shift shift_type = static_cast<Shift>(instr->ShiftDP());
+  unsigned shift_amount = instr->ImmDPShift();
+  int64_t op2 = ShiftOperand(reg_size, reg(reg_size, instr->Rm()), shift_type,
+                             shift_amount);
+  if (instr->Mask(NOT) == NOT) {
+    op2 = ~op2;
+  }
+  LogicalHelper(instr, op2);
+}
+
+
+void Simulator::VisitLogicalImmediate(Instruction* instr) {
+  LogicalHelper(instr, instr->ImmLogical());
+}
+
+
+void Simulator::LogicalHelper(Instruction* instr, int64_t op2) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t op1 = reg(reg_size, instr->Rn());
+  int64_t result = 0;
+  bool update_flags = false;
+
+  // Switch on the logical operation, stripping out the NOT bit, as it has a
+  // different meaning for logical immediate instructions.
+  switch (instr->Mask(LogicalOpMask & ~NOT)) {
+    case ANDS: update_flags = true;  // Fall through.
+    case AND: result = op1 & op2; break;
+    case ORR: result = op1 | op2; break;
+    case EOR: result = op1 ^ op2; break;
+    default:
+      UNIMPLEMENTED();
+  }
+
+  if (update_flags) {
+    nzcv().SetN(CalcNFlag(result, reg_size));
+    nzcv().SetZ(CalcZFlag(result));
+    nzcv().SetC(0);
+    nzcv().SetV(0);
+  }
+
+  set_reg(reg_size, instr->Rd(), result, instr->RdMode());
+}
+
+
+void Simulator::VisitConditionalCompareRegister(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  ConditionalCompareHelper(instr, reg(reg_size, instr->Rm()));
+}
+
+
+void Simulator::VisitConditionalCompareImmediate(Instruction* instr) {
+  ConditionalCompareHelper(instr, instr->ImmCondCmp());
+}
+
+
+void Simulator::ConditionalCompareHelper(Instruction* instr, int64_t op2) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t op1 = reg(reg_size, instr->Rn());
+
+  if (ConditionPassed(static_cast<Condition>(instr->Condition()))) {
+    // If the condition passes, set the status flags to the result of comparing
+    // the operands.
+    if (instr->Mask(ConditionalCompareMask) == CCMP) {
+      AddWithCarry(reg_size, true, op1, ~op2, 1);
+    } else {
+      ASSERT(instr->Mask(ConditionalCompareMask) == CCMN);
+      AddWithCarry(reg_size, true, op1, op2, 0);
+    }
+  } else {
+    // If the condition fails, set the status flags to the nzcv immediate.
+    nzcv().SetFlags(instr->Nzcv());
+  }
+}
+
+
+void Simulator::VisitLoadStoreUnsignedOffset(Instruction* instr) {
+  int offset = instr->ImmLSUnsigned() << instr->SizeLS();
+  LoadStoreHelper(instr, offset, Offset);
+}
+
+
+void Simulator::VisitLoadStoreUnscaledOffset(Instruction* instr) {
+  LoadStoreHelper(instr, instr->ImmLS(), Offset);
+}
+
+
+void Simulator::VisitLoadStorePreIndex(Instruction* instr) {
+  LoadStoreHelper(instr, instr->ImmLS(), PreIndex);
+}
+
+
+void Simulator::VisitLoadStorePostIndex(Instruction* instr) {
+  LoadStoreHelper(instr, instr->ImmLS(), PostIndex);
+}
+
+
+void Simulator::VisitLoadStoreRegisterOffset(Instruction* instr) {
+  Extend ext = static_cast<Extend>(instr->ExtendMode());
+  ASSERT((ext == UXTW) || (ext == UXTX) || (ext == SXTW) || (ext == SXTX));
+  unsigned shift_amount = instr->ImmShiftLS() * instr->SizeLS();
+
+  int64_t offset = ExtendValue(kXRegSize, xreg(instr->Rm()), ext,
+                               shift_amount);
+  LoadStoreHelper(instr, offset, Offset);
+}
+
+
+void Simulator::LoadStoreHelper(Instruction* instr,
+                                int64_t offset,
+                                AddrMode addrmode) {
+  unsigned srcdst = instr->Rt();
+  uint8_t* address = AddressModeHelper(instr->Rn(), offset, addrmode);
+  int num_bytes = 1 << instr->SizeLS();
+
+  LoadStoreOp op = static_cast<LoadStoreOp>(instr->Mask(LoadStoreOpMask));
+  switch (op) {
+    case LDRB_w:
+    case LDRH_w:
+    case LDR_w:
+    case LDR_x: set_xreg(srcdst, MemoryRead(address, num_bytes)); break;
+    case STRB_w:
+    case STRH_w:
+    case STR_w:
+    case STR_x: MemoryWrite(address, xreg(srcdst), num_bytes); break;
+    case LDRSB_w: {
+      set_wreg(srcdst, ExtendValue(kWRegSize, MemoryRead8(address), SXTB));
+      break;
+    }
+    case LDRSB_x: {
+      set_xreg(srcdst, ExtendValue(kXRegSize, MemoryRead8(address), SXTB));
+      break;
+    }
+    case LDRSH_w: {
+      set_wreg(srcdst, ExtendValue(kWRegSize, MemoryRead16(address), SXTH));
+      break;
+    }
+    case LDRSH_x: {
+      set_xreg(srcdst, ExtendValue(kXRegSize, MemoryRead16(address), SXTH));
+      break;
+    }
+    case LDRSW_x: {
+      set_xreg(srcdst, ExtendValue(kXRegSize, MemoryRead32(address), SXTW));
+      break;
+    }
+    case LDR_s: set_sreg(srcdst, MemoryReadFP32(address)); break;
+    case LDR_d: set_dreg(srcdst, MemoryReadFP64(address)); break;
+    case STR_s: MemoryWriteFP32(address, sreg(srcdst)); break;
+    case STR_d: MemoryWriteFP64(address, dreg(srcdst)); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitLoadStorePairOffset(Instruction* instr) {
+  LoadStorePairHelper(instr, Offset);
+}
+
+
+void Simulator::VisitLoadStorePairPreIndex(Instruction* instr) {
+  LoadStorePairHelper(instr, PreIndex);
+}
+
+
+void Simulator::VisitLoadStorePairPostIndex(Instruction* instr) {
+  LoadStorePairHelper(instr, PostIndex);
+}
+
+
+void Simulator::VisitLoadStorePairNonTemporal(Instruction* instr) {
+  LoadStorePairHelper(instr, Offset);
+}
+
+
+void Simulator::LoadStorePairHelper(Instruction* instr,
+                                    AddrMode addrmode) {
+  unsigned rt = instr->Rt();
+  unsigned rt2 = instr->Rt2();
+  int offset = instr->ImmLSPair() << instr->SizeLSPair();
+  uint8_t* address = AddressModeHelper(instr->Rn(), offset, addrmode);
+
+  LoadStorePairOp op =
+    static_cast<LoadStorePairOp>(instr->Mask(LoadStorePairMask));
+
+  // 'rt' and 'rt2' can only be aliased for stores.
+  ASSERT(((op & LoadStorePairLBit) == 0) || (rt != rt2));
+
+  switch (op) {
+    case LDP_w: {
+      set_wreg(rt, MemoryRead32(address));
+      set_wreg(rt2, MemoryRead32(address + kWRegSizeInBytes));
+      break;
+    }
+    case LDP_s: {
+      set_sreg(rt, MemoryReadFP32(address));
+      set_sreg(rt2, MemoryReadFP32(address + kSRegSizeInBytes));
+      break;
+    }
+    case LDP_x: {
+      set_xreg(rt, MemoryRead64(address));
+      set_xreg(rt2, MemoryRead64(address + kXRegSizeInBytes));
+      break;
+    }
+    case LDP_d: {
+      set_dreg(rt, MemoryReadFP64(address));
+      set_dreg(rt2, MemoryReadFP64(address + kDRegSizeInBytes));
+      break;
+    }
+    case LDPSW_x: {
+      set_xreg(rt, ExtendValue(kXRegSize, MemoryRead32(address), SXTW));
+      set_xreg(rt2, ExtendValue(kXRegSize,
+               MemoryRead32(address + kWRegSizeInBytes), SXTW));
+      break;
+    }
+    case STP_w: {
+      MemoryWrite32(address, wreg(rt));
+      MemoryWrite32(address + kWRegSizeInBytes, wreg(rt2));
+      break;
+    }
+    case STP_s: {
+      MemoryWriteFP32(address, sreg(rt));
+      MemoryWriteFP32(address + kSRegSizeInBytes, sreg(rt2));
+      break;
+    }
+    case STP_x: {
+      MemoryWrite64(address, xreg(rt));
+      MemoryWrite64(address + kXRegSizeInBytes, xreg(rt2));
+      break;
+    }
+    case STP_d: {
+      MemoryWriteFP64(address, dreg(rt));
+      MemoryWriteFP64(address + kDRegSizeInBytes, dreg(rt2));
+      break;
+    }
+    default: UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitLoadLiteral(Instruction* instr) {
+  uint8_t* address = instr->LiteralAddress();
+  unsigned rt = instr->Rt();
+
+  switch (instr->Mask(LoadLiteralMask)) {
+    case LDR_w_lit: set_wreg(rt, MemoryRead32(address));  break;
+    case LDR_x_lit: set_xreg(rt, MemoryRead64(address));  break;
+    case LDR_s_lit: set_sreg(rt, MemoryReadFP32(address));  break;
+    case LDR_d_lit: set_dreg(rt, MemoryReadFP64(address));  break;
+    default: UNREACHABLE();
+  }
+}
+
+
+uint8_t* Simulator::AddressModeHelper(unsigned addr_reg,
+                                      int64_t offset,
+                                      AddrMode addrmode) {
+  uint64_t address = xreg(addr_reg, Reg31IsStackPointer);
+  ASSERT((sizeof(uintptr_t) == kXRegSizeInBytes) ||
+         (address < 0x100000000UL));
+  if ((addr_reg == 31) && ((address % 16) != 0)) {
+    // When the base register is SP the stack pointer is required to be
+    // quadword aligned prior to the address calculation and write-backs.
+    // Misalignment will cause a stack alignment fault.
+    ALIGNMENT_EXCEPTION();
+  }
+  if ((addrmode == PreIndex) || (addrmode == PostIndex)) {
+    ASSERT(offset != 0);
+    set_xreg(addr_reg, address + offset, Reg31IsStackPointer);
+  }
+
+  if ((addrmode == Offset) || (addrmode == PreIndex)) {
+    address += offset;
+  }
+
+  return reinterpret_cast<uint8_t*>(address);
+}
+
+
+uint64_t Simulator::MemoryRead(const uint8_t* address, unsigned num_bytes) {
+  ASSERT(address != NULL);
+  ASSERT((num_bytes > 0) && (num_bytes <= sizeof(uint64_t)));
+  uint64_t read = 0;
+  memcpy(&read, address, num_bytes);
+  return read;
+}
+
+
+uint8_t Simulator::MemoryRead8(uint8_t* address) {
+  return MemoryRead(address, sizeof(uint8_t));
+}
+
+
+uint16_t Simulator::MemoryRead16(uint8_t* address) {
+  return MemoryRead(address, sizeof(uint16_t));
+}
+
+
+uint32_t Simulator::MemoryRead32(uint8_t* address) {
+  return MemoryRead(address, sizeof(uint32_t));
+}
+
+
+float Simulator::MemoryReadFP32(uint8_t* address) {
+  return rawbits_to_float(MemoryRead32(address));
+}
+
+
+uint64_t Simulator::MemoryRead64(uint8_t* address) {
+  return MemoryRead(address, sizeof(uint64_t));
+}
+
+
+double Simulator::MemoryReadFP64(uint8_t* address) {
+  return rawbits_to_double(MemoryRead64(address));
+}
+
+
+void Simulator::MemoryWrite(uint8_t* address,
+                            uint64_t value,
+                            unsigned num_bytes) {
+  ASSERT(address != NULL);
+  ASSERT((num_bytes > 0) && (num_bytes <= sizeof(uint64_t)));
+  memcpy(address, &value, num_bytes);
+}
+
+
+void Simulator::MemoryWrite32(uint8_t* address, uint32_t value) {
+  MemoryWrite(address, value, sizeof(uint32_t));
+}
+
+
+void Simulator::MemoryWriteFP32(uint8_t* address, float value) {
+  MemoryWrite32(address, float_to_rawbits(value));
+}
+
+
+void Simulator::MemoryWrite64(uint8_t* address, uint64_t value) {
+  MemoryWrite(address, value, sizeof(uint64_t));
+}
+
+
+void Simulator::MemoryWriteFP64(uint8_t* address, double value) {
+  MemoryWrite64(address, double_to_rawbits(value));
+}
+
+
+void Simulator::VisitMoveWideImmediate(Instruction* instr) {
+  MoveWideImmediateOp mov_op =
+    static_cast<MoveWideImmediateOp>(instr->Mask(MoveWideImmediateMask));
+  int64_t new_xn_val = 0;
+
+  bool is_64_bits = instr->SixtyFourBits() == 1;
+  // Shift is limited for W operations.
+  ASSERT(is_64_bits || (instr->ShiftMoveWide() < 2));
+
+  // Get the shifted immediate.
+  int64_t shift = instr->ShiftMoveWide() * 16;
+  int64_t shifted_imm16 = instr->ImmMoveWide() << shift;
+
+  // Compute the new value.
+  switch (mov_op) {
+    case MOVN_w:
+    case MOVN_x: {
+        new_xn_val = ~shifted_imm16;
+        if (!is_64_bits) new_xn_val &= kWRegMask;
+      break;
+    }
+    case MOVK_w:
+    case MOVK_x: {
+        unsigned reg_code = instr->Rd();
+        int64_t prev_xn_val = is_64_bits ? xreg(reg_code)
+                                         : wreg(reg_code);
+        new_xn_val = (prev_xn_val & ~(0xffffL << shift)) | shifted_imm16;
+      break;
+    }
+    case MOVZ_w:
+    case MOVZ_x: {
+        new_xn_val = shifted_imm16;
+      break;
+    }
+    default:
+      UNREACHABLE();
+  }
+
+  // Update the destination register.
+  set_xreg(instr->Rd(), new_xn_val);
+}
+
+
+void Simulator::VisitConditionalSelect(Instruction* instr) {
+  uint64_t new_val = xreg(instr->Rn());
+
+  if (ConditionFailed(static_cast<Condition>(instr->Condition()))) {
+    new_val = xreg(instr->Rm());
+    switch (instr->Mask(ConditionalSelectMask)) {
+      case CSEL_w:
+      case CSEL_x: break;
+      case CSINC_w:
+      case CSINC_x: new_val++; break;
+      case CSINV_w:
+      case CSINV_x: new_val = ~new_val; break;
+      case CSNEG_w:
+      case CSNEG_x: new_val = -new_val; break;
+      default: UNIMPLEMENTED();
+    }
+  }
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  set_reg(reg_size, instr->Rd(), new_val);
+}
+
+
+void Simulator::VisitDataProcessing1Source(Instruction* instr) {
+  unsigned dst = instr->Rd();
+  unsigned src = instr->Rn();
+
+  switch (instr->Mask(DataProcessing1SourceMask)) {
+    case RBIT_w: set_wreg(dst, ReverseBits(wreg(src), kWRegSize)); break;
+    case RBIT_x: set_xreg(dst, ReverseBits(xreg(src), kXRegSize)); break;
+    case REV16_w: set_wreg(dst, ReverseBytes(wreg(src), Reverse16)); break;
+    case REV16_x: set_xreg(dst, ReverseBytes(xreg(src), Reverse16)); break;
+    case REV_w: set_wreg(dst, ReverseBytes(wreg(src), Reverse32)); break;
+    case REV32_x: set_xreg(dst, ReverseBytes(xreg(src), Reverse32)); break;
+    case REV_x: set_xreg(dst, ReverseBytes(xreg(src), Reverse64)); break;
+    case CLZ_w: set_wreg(dst, CountLeadingZeros(wreg(src), kWRegSize)); break;
+    case CLZ_x: set_xreg(dst, CountLeadingZeros(xreg(src), kXRegSize)); break;
+    case CLS_w: {
+      set_wreg(dst, CountLeadingSignBits(wreg(src), kWRegSize));
+      break;
+    }
+    case CLS_x: {
+      set_xreg(dst, CountLeadingSignBits(xreg(src), kXRegSize));
+      break;
+    }
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+uint64_t Simulator::ReverseBits(uint64_t value, unsigned num_bits) {
+  ASSERT((num_bits == kWRegSize) || (num_bits == kXRegSize));
+  uint64_t result = 0;
+  for (unsigned i = 0; i < num_bits; i++) {
+    result = (result << 1) | (value & 1);
+    value >>= 1;
+  }
+  return result;
+}
+
+
+uint64_t Simulator::ReverseBytes(uint64_t value, ReverseByteMode mode) {
+  // Split the 64-bit value into an 8-bit array, where b[0] is the least
+  // significant byte, and b[7] is the most significant.
+  uint8_t bytes[8];
+  uint64_t mask = 0xff00000000000000UL;
+  for (int i = 7; i >= 0; i--) {
+    bytes[i] = (value & mask) >> (i * 8);
+    mask >>= 8;
+  }
+
+  // Permutation tables for REV instructions.
+  //  permute_table[Reverse16] is used by REV16_x, REV16_w
+  //  permute_table[Reverse32] is used by REV32_x, REV_w
+  //  permute_table[Reverse64] is used by REV_x
+  ASSERT((Reverse16 == 0) && (Reverse32 == 1) && (Reverse64 == 2));
+  static const uint8_t permute_table[3][8] = { {6, 7, 4, 5, 2, 3, 0, 1},
+                                               {4, 5, 6, 7, 0, 1, 2, 3},
+                                               {0, 1, 2, 3, 4, 5, 6, 7} };
+  uint64_t result = 0;
+  for (int i = 0; i < 8; i++) {
+    result <<= 8;
+    result |= bytes[permute_table[mode][i]];
+  }
+  return result;
+}
+
+
+void Simulator::VisitDataProcessing2Source(Instruction* instr) {
+  Shift shift_op = NO_SHIFT;
+  int64_t result = 0;
+  switch (instr->Mask(DataProcessing2SourceMask)) {
+    case SDIV_w: result = wreg(instr->Rn()) / wreg(instr->Rm()); break;
+    case SDIV_x: result = xreg(instr->Rn()) / xreg(instr->Rm()); break;
+    case UDIV_w: {
+      uint32_t rn = static_cast<uint32_t>(wreg(instr->Rn()));
+      uint32_t rm = static_cast<uint32_t>(wreg(instr->Rm()));
+      result = rn / rm;
+      break;
+    }
+    case UDIV_x: {
+      uint64_t rn = static_cast<uint64_t>(xreg(instr->Rn()));
+      uint64_t rm = static_cast<uint64_t>(xreg(instr->Rm()));
+      result = rn / rm;
+      break;
+    }
+    case LSLV_w:
+    case LSLV_x: shift_op = LSL; break;
+    case LSRV_w:
+    case LSRV_x: shift_op = LSR; break;
+    case ASRV_w:
+    case ASRV_x: shift_op = ASR; break;
+    case RORV_w:
+    case RORV_x: shift_op = ROR; break;
+    default: UNIMPLEMENTED();
+  }
+
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  if (shift_op != NO_SHIFT) {
+    // Shift distance encoded in the least-significant five/six bits of the
+    // register.
+    int mask = (instr->SixtyFourBits() == 1) ? 0x3f : 0x1f;
+    unsigned shift = wreg(instr->Rm()) & mask;
+    result = ShiftOperand(reg_size, reg(reg_size, instr->Rn()), shift_op,
+                          shift);
+  }
+  set_reg(reg_size, instr->Rd(), result);
+}
+
+
+// The algorithm used is adapted from the one described in section 8.2 of
+//   Hacker's Delight, by Henry S. Warren, Jr.
+// It assumes that a right shift on a signed integer is an arithmetic shift.
+static int64_t MultiplyHighSigned(int64_t u, int64_t v) {
+  uint64_t u0, v0, w0;
+  int64_t u1, v1, w1, w2, t;
+
+  u0 = u & 0xffffffffL;
+  u1 = u >> 32;
+  v0 = v & 0xffffffffL;
+  v1 = v >> 32;
+
+  w0 = u0 * v0;
+  t = u1 * v0 + (w0 >> 32);
+  w1 = t & 0xffffffffL;
+  w2 = t >> 32;
+  w1 = u0 * v1 + w1;
+
+  return u1 * v1 + w2 + (w1 >> 32);
+}
+
+
+void Simulator::VisitDataProcessing3Source(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+
+  int64_t result = 0;
+  uint64_t rn;
+  uint64_t rm;
+  switch (instr->Mask(DataProcessing3SourceMask)) {
+    case MADD_w:
+    case MADD_x:
+      result = xreg(instr->Ra()) + (xreg(instr->Rn()) * xreg(instr->Rm()));
+      break;
+    case MSUB_w:
+    case MSUB_x:
+      result = xreg(instr->Ra()) - (xreg(instr->Rn()) * xreg(instr->Rm()));
+      break;
+    case SMADDL_x:
+      result = xreg(instr->Ra()) + (wreg(instr->Rn()) * wreg(instr->Rm()));
+      break;
+    case SMSUBL_x:
+      result = xreg(instr->Ra()) - (wreg(instr->Rn()) * wreg(instr->Rm()));
+      break;
+    case UMADDL_x:
+      rn = static_cast<uint32_t>(wreg(instr->Rn()));
+      rm = static_cast<uint32_t>(wreg(instr->Rm()));
+      result = xreg(instr->Ra()) + (rn * rm);
+      break;
+    case UMSUBL_x:
+      rn = static_cast<uint32_t>(wreg(instr->Rn()));
+      rm = static_cast<uint32_t>(wreg(instr->Rm()));
+      result = xreg(instr->Ra()) - (rn * rm);
+      break;
+    case SMULH_x:
+      result = MultiplyHighSigned(xreg(instr->Rn()), xreg(instr->Rm()));
+      break;
+    default: UNIMPLEMENTED();
+  }
+  set_reg(reg_size, instr->Rd(), result);
+}
+
+
+void Simulator::VisitBitfield(Instruction* instr) {
+  unsigned reg_size = instr->SixtyFourBits() ? kXRegSize : kWRegSize;
+  int64_t reg_mask = instr->SixtyFourBits() ? kXRegMask : kWRegMask;
+  int64_t R = instr->ImmR();
+  int64_t S = instr->ImmS();
+  int64_t diff = S - R;
+  int64_t mask;
+  if (diff >= 0) {
+    mask = diff < reg_size - 1 ? (1L << (diff + 1)) - 1
+                               : reg_mask;
+  } else {
+    mask = ((1L << (S + 1)) - 1);
+    mask = (static_cast<uint64_t>(mask) >> R) | (mask << (reg_size - R));
+    diff += reg_size;
+  }
+
+  // inzero indicates if the extracted bitfield is inserted into the
+  // destination register value or in zero.
+  // If extend is true, extend the sign of the extracted bitfield.
+  bool inzero = false;
+  bool extend = false;
+  switch (instr->Mask(BitfieldMask)) {
+    case BFM_x:
+    case BFM_w:
+      break;
+    case SBFM_x:
+    case SBFM_w:
+      inzero = true;
+      extend = true;
+      break;
+    case UBFM_x:
+    case UBFM_w:
+      inzero = true;
+      break;
+    default:
+      UNIMPLEMENTED();
+  }
+
+  int64_t dst = inzero ? 0 : reg(reg_size, instr->Rd());
+  int64_t src = reg(reg_size, instr->Rn());
+  // Rotate source bitfield into place.
+  int64_t result = (static_cast<uint64_t>(src) >> R) | (src << (reg_size - R));
+  // Determine the sign extension.
+  int64_t topbits = ((1L << (reg_size - diff - 1)) - 1) << (diff + 1);
+  int64_t signbits = extend && ((src >> S) & 1) ? topbits : 0;
+
+  // Merge sign extension, dest/zero and bitfield.
+  result = signbits | (result & mask) | (dst & ~mask);
+
+  set_reg(reg_size, instr->Rd(), result);
+}
+
+
+void Simulator::VisitExtract(Instruction* instr) {
+  unsigned lsb = instr->ImmS();
+  unsigned reg_size = (instr->SixtyFourBits() == 1) ? kXRegSize
+                                                    : kWRegSize;
+  set_reg(reg_size,
+          instr->Rd(),
+          (static_cast<uint64_t>(reg(reg_size, instr->Rm())) >> lsb) |
+          (reg(reg_size, instr->Rn()) << (reg_size - lsb)));
+}
+
+
+void Simulator::VisitFPImmediate(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned dest = instr->Rd();
+  switch (instr->Mask(FPImmediateMask)) {
+    case FMOV_s_imm: set_sreg(dest, instr->ImmFP32()); break;
+    case FMOV_d_imm: set_dreg(dest, instr->ImmFP64()); break;
+    default: UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitFPIntegerConvert(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned dst = instr->Rd();
+  unsigned src = instr->Rn();
+
+  FPRounding round = RMode();
+
+  switch (instr->Mask(FPIntegerConvertMask)) {
+    case FCVTMS_ws:
+      set_wreg(dst, FPToInt32(sreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMS_xs:
+      set_xreg(dst, FPToInt64(sreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMS_wd:
+      set_wreg(dst, FPToInt32(dreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMS_xd:
+      set_xreg(dst, FPToInt64(dreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMU_ws:
+      set_wreg(dst, FPToUInt32(sreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMU_xs:
+      set_xreg(dst, FPToUInt64(sreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMU_wd:
+      set_wreg(dst, FPToUInt32(dreg(src), FPNegativeInfinity));
+      break;
+    case FCVTMU_xd:
+      set_xreg(dst, FPToUInt64(dreg(src), FPNegativeInfinity));
+      break;
+    case FCVTNS_ws: set_wreg(dst, FPToInt32(sreg(src), FPTieEven)); break;
+    case FCVTNS_xs: set_xreg(dst, FPToInt64(sreg(src), FPTieEven)); break;
+    case FCVTNS_wd: set_wreg(dst, FPToInt32(dreg(src), FPTieEven)); break;
+    case FCVTNS_xd: set_xreg(dst, FPToInt64(dreg(src), FPTieEven)); break;
+    case FCVTNU_ws: set_wreg(dst, FPToUInt32(sreg(src), FPTieEven)); break;
+    case FCVTNU_xs: set_xreg(dst, FPToUInt64(sreg(src), FPTieEven)); break;
+    case FCVTNU_wd: set_wreg(dst, FPToUInt32(dreg(src), FPTieEven)); break;
+    case FCVTNU_xd: set_xreg(dst, FPToUInt64(dreg(src), FPTieEven)); break;
+    case FCVTZS_ws: set_wreg(dst, FPToInt32(sreg(src), FPZero)); break;
+    case FCVTZS_xs: set_xreg(dst, FPToInt64(sreg(src), FPZero)); break;
+    case FCVTZS_wd: set_wreg(dst, FPToInt32(dreg(src), FPZero)); break;
+    case FCVTZS_xd: set_xreg(dst, FPToInt64(dreg(src), FPZero)); break;
+    case FCVTZU_ws: set_wreg(dst, FPToUInt32(sreg(src), FPZero)); break;
+    case FCVTZU_xs: set_xreg(dst, FPToUInt64(sreg(src), FPZero)); break;
+    case FCVTZU_wd: set_wreg(dst, FPToUInt32(dreg(src), FPZero)); break;
+    case FCVTZU_xd: set_xreg(dst, FPToUInt64(dreg(src), FPZero)); break;
+    case FMOV_ws: set_wreg(dst, sreg_bits(src)); break;
+    case FMOV_xd: set_xreg(dst, dreg_bits(src)); break;
+    case FMOV_sw: set_sreg_bits(dst, wreg(src)); break;
+    case FMOV_dx: set_dreg_bits(dst, xreg(src)); break;
+
+    // A 32-bit input can be handled in the same way as a 64-bit input, since
+    // the sign- or zero-extension will not affect the conversion.
+    case SCVTF_dx: set_dreg(dst, FixedToDouble(xreg(src), 0, round)); break;
+    case SCVTF_dw: set_dreg(dst, FixedToDouble(wreg(src), 0, round)); break;
+    case UCVTF_dx: set_dreg(dst, UFixedToDouble(xreg(src), 0, round)); break;
+    case UCVTF_dw: {
+      set_dreg(dst, UFixedToDouble(static_cast<uint32_t>(wreg(src)), 0, round));
+      break;
+    }
+    case SCVTF_sx: set_sreg(dst, FixedToFloat(xreg(src), 0, round)); break;
+    case SCVTF_sw: set_sreg(dst, FixedToFloat(wreg(src), 0, round)); break;
+    case UCVTF_sx: set_sreg(dst, UFixedToFloat(xreg(src), 0, round)); break;
+    case UCVTF_sw: {
+      set_sreg(dst, UFixedToFloat(static_cast<uint32_t>(wreg(src)), 0, round));
+      break;
+    }
+
+    default: UNREACHABLE();
+  }
+}
+
+
+void Simulator::VisitFPFixedPointConvert(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned dst = instr->Rd();
+  unsigned src = instr->Rn();
+  int fbits = 64 - instr->FPScale();
+
+  FPRounding round = RMode();
+
+  switch (instr->Mask(FPFixedPointConvertMask)) {
+    // A 32-bit input can be handled in the same way as a 64-bit input, since
+    // the sign- or zero-extension will not affect the conversion.
+    case SCVTF_dx_fixed:
+      set_dreg(dst, FixedToDouble(xreg(src), fbits, round));
+      break;
+    case SCVTF_dw_fixed:
+      set_dreg(dst, FixedToDouble(wreg(src), fbits, round));
+      break;
+    case UCVTF_dx_fixed:
+      set_dreg(dst, UFixedToDouble(xreg(src), fbits, round));
+      break;
+    case UCVTF_dw_fixed: {
+      set_dreg(dst,
+               UFixedToDouble(static_cast<uint32_t>(wreg(src)), fbits, round));
+      break;
+    }
+    case SCVTF_sx_fixed:
+      set_sreg(dst, FixedToFloat(xreg(src), fbits, round));
+      break;
+    case SCVTF_sw_fixed:
+      set_sreg(dst, FixedToFloat(wreg(src), fbits, round));
+      break;
+    case UCVTF_sx_fixed:
+      set_sreg(dst, UFixedToFloat(xreg(src), fbits, round));
+      break;
+    case UCVTF_sw_fixed: {
+      set_sreg(dst,
+               UFixedToFloat(static_cast<uint32_t>(wreg(src)), fbits, round));
+      break;
+    }
+    default: UNREACHABLE();
+  }
+}
+
+
+int32_t Simulator::FPToInt32(double value, FPRounding rmode) {
+  value = FPRoundInt(value, rmode);
+  if (value >= kWMaxInt) {
+    return kWMaxInt;
+  } else if (value < kWMinInt) {
+    return kWMinInt;
+  }
+  return isnan(value) ? 0 : static_cast<int32_t>(value);
+}
+
+
+int64_t Simulator::FPToInt64(double value, FPRounding rmode) {
+  value = FPRoundInt(value, rmode);
+  if (value >= kXMaxInt) {
+    return kXMaxInt;
+  } else if (value < kXMinInt) {
+    return kXMinInt;
+  }
+  return isnan(value) ? 0 : static_cast<int64_t>(value);
+}
+
+
+uint32_t Simulator::FPToUInt32(double value, FPRounding rmode) {
+  value = FPRoundInt(value, rmode);
+  if (value >= kWMaxUInt) {
+    return kWMaxUInt;
+  } else if (value < 0.0) {
+    return 0;
+  }
+  return isnan(value) ? 0 : static_cast<uint32_t>(value);
+}
+
+
+uint64_t Simulator::FPToUInt64(double value, FPRounding rmode) {
+  value = FPRoundInt(value, rmode);
+  if (value >= kXMaxUInt) {
+    return kXMaxUInt;
+  } else if (value < 0.0) {
+    return 0;
+  }
+  return isnan(value) ? 0 : static_cast<uint64_t>(value);
+}
+
+
+void Simulator::VisitFPCompare(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned reg_size = instr->FPType() == FP32 ? kSRegSize : kDRegSize;
+  double fn_val = fpreg(reg_size, instr->Rn());
+
+  switch (instr->Mask(FPCompareMask)) {
+    case FCMP_s:
+    case FCMP_d: FPCompare(fn_val, fpreg(reg_size, instr->Rm())); break;
+    case FCMP_s_zero:
+    case FCMP_d_zero: FPCompare(fn_val, 0.0); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitFPConditionalCompare(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  switch (instr->Mask(FPConditionalCompareMask)) {
+    case FCCMP_s:
+    case FCCMP_d: {
+      if (ConditionPassed(static_cast<Condition>(instr->Condition()))) {
+        // If the condition passes, set the status flags to the result of
+        // comparing the operands.
+        unsigned reg_size = instr->FPType() == FP32 ? kSRegSize : kDRegSize;
+        FPCompare(fpreg(reg_size, instr->Rn()), fpreg(reg_size, instr->Rm()));
+      } else {
+        // If the condition fails, set the status flags to the nzcv immediate.
+        nzcv().SetFlags(instr->Nzcv());
+      }
+      break;
+    }
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitFPConditionalSelect(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned reg_size = instr->FPType() == FP32 ? kSRegSize : kDRegSize;
+
+  double selected_val;
+  if (ConditionPassed(static_cast<Condition>(instr->Condition()))) {
+    selected_val = fpreg(reg_size, instr->Rn());
+  } else {
+    selected_val = fpreg(reg_size, instr->Rm());
+  }
+
+  switch (instr->Mask(FPConditionalSelectMask)) {
+    case FCSEL_s:
+    case FCSEL_d: set_fpreg(reg_size, instr->Rd(), selected_val); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitFPDataProcessing1Source(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned fd = instr->Rd();
+  unsigned fn = instr->Rn();
+
+  switch (instr->Mask(FPDataProcessing1SourceMask)) {
+    case FMOV_s: set_sreg(fd, sreg(fn)); break;
+    case FMOV_d: set_dreg(fd, dreg(fn)); break;
+    case FABS_s: set_sreg(fd, fabs(sreg(fn))); break;
+    case FABS_d: set_dreg(fd, fabs(dreg(fn))); break;
+    case FNEG_s: set_sreg(fd, -sreg(fn)); break;
+    case FNEG_d: set_dreg(fd, -dreg(fn)); break;
+    case FSQRT_s: set_sreg(fd, sqrt(sreg(fn))); break;
+    case FSQRT_d: set_dreg(fd, sqrt(dreg(fn))); break;
+    case FRINTN_s: set_sreg(fd, FPRoundInt(sreg(fn), FPTieEven)); break;
+    case FRINTN_d: set_dreg(fd, FPRoundInt(dreg(fn), FPTieEven)); break;
+    case FRINTZ_s: set_sreg(fd, FPRoundInt(sreg(fn), FPZero)); break;
+    case FRINTZ_d: set_dreg(fd, FPRoundInt(dreg(fn), FPZero)); break;
+    case FCVT_ds: set_dreg(fd, FPToDouble(sreg(fn))); break;
+    case FCVT_sd: set_sreg(fd, FPToFloat(dreg(fn), FPTieEven)); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+// Assemble the specified IEEE-754 components into the target type and apply
+// appropriate rounding.
+//  sign:     0 = positive, 1 = negative
+//  exponent: Unbiased IEEE-754 exponent.
+//  mantissa: The mantissa of the input. The top bit (which is not encoded for
+//            normal IEEE-754 values) must not be omitted. This bit has the
+//            value 'pow(2, exponent)'.
+//
+// The input value is assumed to be a normalized value. That is, the input may
+// not be infinity or NaN. If the source value is subnormal, it must be
+// normalized before calling this function such that the highest set bit in the
+// mantissa has the value 'pow(2, exponent)'.
+//
+// Callers should use FPRoundToFloat or FPRoundToDouble directly, rather than
+// calling a templated FPRound.
+template <class T, int ebits, int mbits>
+static T FPRound(int64_t sign, int64_t exponent, uint64_t mantissa,
+                 FPRounding round_mode) {
+  ASSERT((sign == 0) || (sign == 1));
+
+  // Only the FPTieEven rounding mode is implemented.
+  ASSERT(round_mode == FPTieEven);
+  USE(round_mode);
+
+  // Rounding can promote subnormals to normals, and normals to infinities. For
+  // example, a double with exponent 127 (FLT_MAX_EXP) would appear to be
+  // encodable as a float, but rounding based on the low-order mantissa bits
+  // could make it overflow. With ties-to-even rounding, this value would become
+  // an infinity.
+
+  // ---- Rounding Method ----
+  //
+  // The exponent is irrelevant in the rounding operation, so we treat the
+  // lowest-order bit that will fit into the result ('onebit') as having
+  // the value '1'. Similarly, the highest-order bit that won't fit into
+  // the result ('halfbit') has the value '0.5'. The 'point' sits between
+  // 'onebit' and 'halfbit':
+  //
+  //            These bits fit into the result.
+  //               |---------------------|
+  //  mantissa = 0bxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
+  //                                     ||
+  //                                    / |
+  //                                   /  halfbit
+  //                               onebit
+  //
+  // For subnormal outputs, the range of representable bits is smaller and
+  // the position of onebit and halfbit depends on the exponent of the
+  // input, but the method is otherwise similar.
+  //
+  //   onebit(frac)
+  //     |
+  //     | halfbit(frac)          halfbit(adjusted)
+  //     | /                      /
+  //     | |                      |
+  //  0b00.0 (exact)      -> 0b00.0 (exact)                    -> 0b00
+  //  0b00.0...           -> 0b00.0...                         -> 0b00
+  //  0b00.1 (exact)      -> 0b00.0111..111                    -> 0b00
+  //  0b00.1...           -> 0b00.1...                         -> 0b01
+  //  0b01.0 (exact)      -> 0b01.0 (exact)                    -> 0b01
+  //  0b01.0...           -> 0b01.0...                         -> 0b01
+  //  0b01.1 (exact)      -> 0b01.1 (exact)                    -> 0b10
+  //  0b01.1...           -> 0b01.1...                         -> 0b10
+  //  0b10.0 (exact)      -> 0b10.0 (exact)                    -> 0b10
+  //  0b10.0...           -> 0b10.0...                         -> 0b10
+  //  0b10.1 (exact)      -> 0b10.0111..111                    -> 0b10
+  //  0b10.1...           -> 0b10.1...                         -> 0b11
+  //  0b11.0 (exact)      -> 0b11.0 (exact)                    -> 0b11
+  //  ...                   /             |                      /   |
+  //                       /              |                     /    |
+  //                                                           /     |
+  // adjusted = frac - (halfbit(mantissa) & ~onebit(frac));   /      |
+  //
+  //                   mantissa = (mantissa >> shift) + halfbit(adjusted);
+
+  static const int mantissa_offset = 0;
+  static const int exponent_offset = mantissa_offset + mbits;
+  static const int sign_offset = exponent_offset + ebits;
+  ASSERT(sign_offset == (sizeof(T) * 8 - 1));
+
+  // Bail out early for zero inputs.
+  if (mantissa == 0) {
+    return sign << sign_offset;
+  }
+
+  // If all bits in the exponent are set, the value is infinite or NaN.
+  // This is true for all binary IEEE-754 formats.
+  static const int infinite_exponent = (1 << ebits) - 1;
+  static const int max_normal_exponent = infinite_exponent - 1;
+
+  // Apply the exponent bias to encode it for the result. Doing this early makes
+  // it easy to detect values that will be infinite or subnormal.
+  exponent += max_normal_exponent >> 1;
+
+  if (exponent > max_normal_exponent) {
+    // Overflow: The input is too large for the result type to represent. The
+    // FPTieEven rounding mode handles overflows using infinities.
+    exponent = infinite_exponent;
+    mantissa = 0;
+    return (sign << sign_offset) |
+           (exponent << exponent_offset) |
+           (mantissa << mantissa_offset);
+  }
+
+  // Calculate the shift required to move the top mantissa bit to the proper
+  // place in the destination type.
+  const int highest_significant_bit = 63 - CountLeadingZeros(mantissa, 64);
+  int shift = highest_significant_bit - mbits;
+
+  if (exponent <= 0) {
+    // The output will be subnormal (before rounding).
+
+    // For subnormal outputs, the shift must be adjusted by the exponent. The +1
+    // is necessary because the exponent of a subnormal value (encoded as 0) is
+    // the same as the exponent of the smallest normal value (encoded as 1).
+    shift += -exponent + 1;
+
+    // Handle inputs that would produce a zero output.
+    //
+    // Shifts higher than highest_significant_bit+1 will always produce a zero
+    // result. A shift of exactly highest_significant_bit+1 might produce a
+    // non-zero result after rounding.
+    if (shift > (highest_significant_bit + 1)) {
+      // The result will always be +/-0.0.
+      return sign << sign_offset;
+    }
+
+    // Properly encode the exponent for a subnormal output.
+    exponent = 0;
+  } else {
+    // Clear the topmost mantissa bit, since this is not encoded in IEEE-754
+    // normal values.
+    mantissa &= ~(1UL << highest_significant_bit);
+  }
+
+  if (shift > 0) {
+    // We have to shift the mantissa to the right. Some precision is lost, so we
+    // need to apply rounding.
+    uint64_t onebit_mantissa = (mantissa >> (shift)) & 1;
+    uint64_t halfbit_mantissa = (mantissa >> (shift-1)) & 1;
+    uint64_t adjusted = mantissa - (halfbit_mantissa & ~onebit_mantissa);
+    T halfbit_adjusted = (adjusted >> (shift-1)) & 1;
+
+    T result = (sign << sign_offset) |
+               (exponent << exponent_offset) |
+               ((mantissa >> shift) << mantissa_offset);
+
+    // A very large mantissa can overflow during rounding. If this happens, the
+    // exponent should be incremented and the mantissa set to 1.0 (encoded as
+    // 0). Applying halfbit_adjusted after assembling the float has the nice
+    // side-effect that this case is handled for free.
+    //
+    // This also handles cases where a very large finite value overflows to
+    // infinity, or where a very large subnormal value overflows to become
+    // normal.
+    return result + halfbit_adjusted;
+  } else {
+    // We have to shift the mantissa to the left (or not at all). The input
+    // mantissa is exactly representable in the output mantissa, so apply no
+    // rounding correction.
+    return (sign << sign_offset) |
+           (exponent << exponent_offset) |
+           ((mantissa << -shift) << mantissa_offset);
+  }
+}
+
+
+// See FPRound for a description of this function.
+static inline double FPRoundToDouble(int64_t sign, int64_t exponent,
+                                     uint64_t mantissa, FPRounding round_mode) {
+  int64_t bits =
+      FPRound<int64_t, kDoubleExponentBits, kDoubleMantissaBits>(sign,
+                                                                 exponent,
+                                                                 mantissa,
+                                                                 round_mode);
+  return rawbits_to_double(bits);
+}
+
+
+// See FPRound for a description of this function.
+static inline float FPRoundToFloat(int64_t sign, int64_t exponent,
+                                   uint64_t mantissa, FPRounding round_mode) {
+  int32_t bits =
+      FPRound<int32_t, kFloatExponentBits, kFloatMantissaBits>(sign,
+                                                               exponent,
+                                                               mantissa,
+                                                               round_mode);
+  return rawbits_to_float(bits);
+}
+
+
+double Simulator::FixedToDouble(int64_t src, int fbits, FPRounding round) {
+  if (src >= 0) {
+    return UFixedToDouble(src, fbits, round);
+  } else {
+    // This works for all negative values, including INT64_MIN.
+    return -UFixedToDouble(-src, fbits, round);
+  }
+}
+
+
+double Simulator::UFixedToDouble(uint64_t src, int fbits, FPRounding round) {
+  // An input of 0 is a special case because the result is effectively
+  // subnormal: The exponent is encoded as 0 and there is no implicit 1 bit.
+  if (src == 0) {
+    return 0.0;
+  }
+
+  // Calculate the exponent. The highest significant bit will have the value
+  // 2^exponent.
+  const int highest_significant_bit = 63 - CountLeadingZeros(src, 64);
+  const int64_t exponent = highest_significant_bit - fbits;
+
+  return FPRoundToDouble(0, exponent, src, round);
+}
+
+
+float Simulator::FixedToFloat(int64_t src, int fbits, FPRounding round) {
+  if (src >= 0) {
+    return UFixedToFloat(src, fbits, round);
+  } else {
+    // This works for all negative values, including INT64_MIN.
+    return -UFixedToFloat(-src, fbits, round);
+  }
+}
+
+
+float Simulator::UFixedToFloat(uint64_t src, int fbits, FPRounding round) {
+  // An input of 0 is a special case because the result is effectively
+  // subnormal: The exponent is encoded as 0 and there is no implicit 1 bit.
+  if (src == 0) {
+    return 0.0f;
+  }
+
+  // Calculate the exponent. The highest significant bit will have the value
+  // 2^exponent.
+  const int highest_significant_bit = 63 - CountLeadingZeros(src, 64);
+  const int32_t exponent = highest_significant_bit - fbits;
+
+  return FPRoundToFloat(0, exponent, src, round);
+}
+
+
+double Simulator::FPRoundInt(double value, FPRounding round_mode) {
+  if ((value == 0.0) || (value == kFP64PositiveInfinity) ||
+      (value == kFP64NegativeInfinity) || isnan(value)) {
+    return value;
+  }
+
+  double int_result = floor(value);
+  double error = value - int_result;
+  switch (round_mode) {
+    case FPTieEven: {
+      // If the error is greater than 0.5, or is equal to 0.5 and the integer
+      // result is odd, round up.
+      if ((error > 0.5) ||
+          ((error == 0.5) && (fmod(int_result, 2) != 0))) {
+        int_result++;
+      }
+      break;
+    }
+    case FPZero: {
+      // If value>0 then we take floor(value)
+      // otherwise, ceil(value).
+      if (value < 0) {
+         int_result = ceil(value);
+      }
+      break;
+    }
+    case FPNegativeInfinity: {
+      // We always use floor(value).
+      break;
+    }
+    default: UNIMPLEMENTED();
+  }
+  return int_result;
+}
+
+
+double Simulator::FPToDouble(float value) {
+  switch (fpclassify(value)) {
+    case FP_NAN: {
+      // Convert NaNs as the processor would, assuming that FPCR.DN (default
+      // NaN) is not set:
+      //  - The sign is propagated.
+      //  - The payload (mantissa) is transferred entirely, except that the top
+      //    bit is forced to '1', making the result a quiet NaN. The unused
+      //    (low-order) payload bits are set to 0.
+      uint32_t raw = float_to_rawbits(value);
+
+      uint64_t sign = raw >> 31;
+      uint64_t exponent = (1 << 11) - 1;
+      uint64_t payload = unsigned_bitextract_64(21, 0, raw);
+      payload <<= (52 - 23);  // The unused low-order bits should be 0.
+      payload |= (1L << 51);  // Force a quiet NaN.
+
+      return rawbits_to_double((sign << 63) | (exponent << 52) | payload);
+    }
+
+    case FP_ZERO:
+    case FP_NORMAL:
+    case FP_SUBNORMAL:
+    case FP_INFINITE: {
+      // All other inputs are preserved in a standard cast, because every value
+      // representable using an IEEE-754 float is also representable using an
+      // IEEE-754 double.
+      return static_cast<double>(value);
+    }
+  }
+
+  UNREACHABLE();
+  return static_cast<double>(value);
+}
+
+
+float Simulator::FPToFloat(double value, FPRounding round_mode) {
+  // Only the FPTieEven rounding mode is implemented.
+  ASSERT(round_mode == FPTieEven);
+  USE(round_mode);
+
+  switch (fpclassify(value)) {
+    case FP_NAN: {
+      // Convert NaNs as the processor would, assuming that FPCR.DN (default
+      // NaN) is not set:
+      //  - The sign is propagated.
+      //  - The payload (mantissa) is transferred as much as possible, except
+      //    that the top bit is forced to '1', making the result a quiet NaN.
+      uint64_t raw = double_to_rawbits(value);
+
+      uint32_t sign = raw >> 63;
+      uint32_t exponent = (1 << 8) - 1;
+      uint32_t payload = unsigned_bitextract_64(50, 52 - 23, raw);
+      payload |= (1 << 22);   // Force a quiet NaN.
+
+      return rawbits_to_float((sign << 31) | (exponent << 23) | payload);
+    }
+
+    case FP_ZERO:
+    case FP_INFINITE: {
+      // In a C++ cast, any value representable in the target type will be
+      // unchanged. This is always the case for +/-0.0 and infinities.
+      return static_cast<float>(value);
+    }
+
+    case FP_NORMAL:
+    case FP_SUBNORMAL: {
+      // Convert double-to-float as the processor would, assuming that FPCR.FZ
+      // (flush-to-zero) is not set.
+      uint64_t raw = double_to_rawbits(value);
+      // Extract the IEEE-754 double components.
+      uint32_t sign = raw >> 63;
+      // Extract the exponent and remove the IEEE-754 encoding bias.
+      int32_t exponent = unsigned_bitextract_64(62, 52, raw) - 1023;
+      // Extract the mantissa and add the implicit '1' bit.
+      uint64_t mantissa = unsigned_bitextract_64(51, 0, raw);
+      if (fpclassify(value) == FP_NORMAL) {
+        mantissa |= (1UL << 52);
+      }
+      return FPRoundToFloat(sign, exponent, mantissa, round_mode);
+    }
+  }
+
+  UNREACHABLE();
+  return value;
+}
+
+
+void Simulator::VisitFPDataProcessing2Source(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned fd = instr->Rd();
+  unsigned fn = instr->Rn();
+  unsigned fm = instr->Rm();
+
+  switch (instr->Mask(FPDataProcessing2SourceMask)) {
+    case FADD_s: set_sreg(fd, sreg(fn) + sreg(fm)); break;
+    case FADD_d: set_dreg(fd, dreg(fn) + dreg(fm)); break;
+    case FSUB_s: set_sreg(fd, sreg(fn) - sreg(fm)); break;
+    case FSUB_d: set_dreg(fd, dreg(fn) - dreg(fm)); break;
+    case FMUL_s: set_sreg(fd, sreg(fn) * sreg(fm)); break;
+    case FMUL_d: set_dreg(fd, dreg(fn) * dreg(fm)); break;
+    case FDIV_s: set_sreg(fd, sreg(fn) / sreg(fm)); break;
+    case FDIV_d: set_dreg(fd, dreg(fn) / dreg(fm)); break;
+    case FMAX_s: set_sreg(fd, FPMax(sreg(fn), sreg(fm))); break;
+    case FMAX_d: set_dreg(fd, FPMax(dreg(fn), dreg(fm))); break;
+    case FMIN_s: set_sreg(fd, FPMin(sreg(fn), sreg(fm))); break;
+    case FMIN_d: set_dreg(fd, FPMin(dreg(fn), dreg(fm))); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitFPDataProcessing3Source(Instruction* instr) {
+  AssertSupportedFPCR();
+
+  unsigned fd = instr->Rd();
+  unsigned fn = instr->Rn();
+  unsigned fm = instr->Rm();
+  unsigned fa = instr->Ra();
+
+  // Note: The FMSUB implementation here is not precisely the same as the
+  // instruction definition. In full implementation rounding of results would
+  // occur once at the end, here rounding will occur after the first multiply
+  // and then after the subsequent addition.  A full implementation here would
+  // be possible but would require an effort isn't immediately justified given
+  // the small differences we expect to see in most cases.
+
+  switch (instr->Mask(FPDataProcessing3SourceMask)) {
+    case FMSUB_s: set_sreg(fd, sreg(fa) + (-sreg(fn))*sreg(fm)); break;
+    case FMSUB_d: set_dreg(fd, dreg(fa) + (-dreg(fn))*dreg(fm)); break;
+    default: UNIMPLEMENTED();
+  }
+}
+
+
+double Simulator::FPMax(double a, double b) {
+  if (isnan(a)) {
+    return a;
+  } else if (isnan(b)) {
+    return b;
+  }
+
+  if ((a == 0.0) && (b == 0.0) &&
+      (copysign(1.0, a) != copysign(1.0, b))) {
+    // a and b are zero, and the sign differs: return +0.0.
+    return 0.0;
+  } else {
+    return (a > b) ? a : b;
+  }
+}
+
+
+double Simulator::FPMin(double a, double b) {
+  if (isnan(a)) {
+    return a;
+  } else if (isnan(b)) {
+    return b;
+  }
+
+  if ((a == 0.0) && (b == 0.0) &&
+      (copysign(1.0, a) != copysign(1.0, b))) {
+    // a and b are zero, and the sign differs: return -0.0.
+    return -0.0;
+  } else {
+    return (a < b) ? a : b;
+  }
+}
+
+
+void Simulator::VisitSystem(Instruction* instr) {
+  // Some system instructions hijack their Op and Cp fields to represent a
+  // range of immediates instead of indicating a different instruction. This
+  // makes the decoding tricky.
+  if (instr->Mask(SystemSysRegFMask) == SystemSysRegFixed) {
+    switch (instr->Mask(SystemSysRegMask)) {
+      case MRS: {
+        switch (instr->ImmSystemRegister()) {
+          case NZCV: set_xreg(instr->Rt(), nzcv().RawValue()); break;
+          case FPCR: set_xreg(instr->Rt(), fpcr().RawValue()); break;
+          default: UNIMPLEMENTED();
+        }
+        break;
+      }
+      case MSR: {
+        switch (instr->ImmSystemRegister()) {
+          case NZCV: nzcv().SetRawValue(xreg(instr->Rt())); break;
+          case FPCR: fpcr().SetRawValue(xreg(instr->Rt())); break;
+          default: UNIMPLEMENTED();
+        }
+        break;
+      }
+    }
+  } else if (instr->Mask(SystemHintFMask) == SystemHintFixed) {
+    ASSERT(instr->Mask(SystemHintMask) == HINT);
+    switch (instr->ImmHint()) {
+      case NOP: break;
+      default: UNIMPLEMENTED();
+    }
+  } else {
+    UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::VisitException(Instruction* instr) {
+  switch (instr->Mask(ExceptionMask)) {
+    case BRK: HostBreakpoint(); break;
+    case HLT:
+      // The Printf pseudo instruction is so useful, we include it in the
+      // default simulator.
+      if (instr->ImmException() == kPrintfOpcode) {
+        DoPrintf(instr);
+      } else {
+        HostBreakpoint();
+      }
+      break;
+    default:
+      UNIMPLEMENTED();
+  }
+}
+
+
+void Simulator::DoPrintf(Instruction* instr) {
+  ASSERT((instr->Mask(ExceptionMask) == HLT) &&
+         (instr->ImmException() == kPrintfOpcode));
+
+  // Read the argument encoded inline in the instruction stream.
+  uint32_t type;
+  ASSERT(sizeof(*instr) == 1);
+  memcpy(&type, instr + kPrintfTypeOffset, sizeof(type));
+
+  const char * format = reinterpret_cast<const char *>(x0());
+  ASSERT(format != NULL);
+
+  // Pass all of the relevant PCS registers onto printf. It doesn't matter
+  // if we pass too many as the extra ones won't be read.
+  int result = 0;
+  if (type == CPURegister::kRegister) {
+    result = printf(format, x1(), x2(), x3(), x4(), x5(), x6(), x7());
+  } else if (type == CPURegister::kFPRegister) {
+    result = printf(format, d0(), d1(), d2(), d3(), d4(), d5(), d6(), d7());
+  } else {
+    ASSERT(type == CPURegister::kNoRegister);
+    result = printf("%s", format);
+  }
+  set_x0(result);
+
+  // TODO: Clobber all caller-saved registers here, to ensure no assumptions
+  // are made about preserved state.
+
+  // The printf parameters are inlined in the code, so skip them.
+  set_pc(instr->InstructionAtOffset(kPrintfLength));
+
+  // Set LR as if we'd just called a native printf function.
+  set_lr(reinterpret_cast<uint64_t>(pc()));
+}
+
+}  // namespace vixl
diff --git a/disas/libvixl/src/a64/simulator-a64.h b/disas/libvixl/src/a64/simulator-a64.h
new file mode 100644
index 0000000..0c22f9e
--- /dev/null
+++ b/disas/libvixl/src/a64/simulator-a64.h
@@ -0,0 +1,576 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_A64_SIMULATOR_A64_H_
+#define VIXL_A64_SIMULATOR_A64_H_
+
+#include "globals.h"
+#include "utils.h"
+#include "a64/instructions-a64.h"
+#include "a64/assembler-a64.h"
+#include "a64/disasm-a64.h"
+#include "a64/instrument-a64.h"
+
+namespace vixl {
+
+enum ReverseByteMode {
+  Reverse16 = 0,
+  Reverse32 = 1,
+  Reverse64 = 2
+};
+
+// Printf. See debugger-a64.h for more informations on pseudo instructions.
+//  - type: CPURegister::RegisterType stored as a uint32_t.
+//
+// Simulate a call to printf.
+//
+// Floating-point and integer arguments are passed in separate sets of
+// registers in AAPCS64 (even for varargs functions), so it is not possible to
+// determine the type of location of each argument without some information
+// about the values that were passed in. This information could be retrieved
+// from the printf format string, but the format string is not trivial to
+// parse so we encode the relevant information with the HLT instruction under
+// the type argument. Therefore the interface is:
+//    x0: The format string
+// x1-x7: Optional arguments, if type == CPURegister::kRegister
+// d0-d7: Optional arguments, if type == CPURegister::kFPRegister
+const Instr kPrintfOpcode = 0xdeb1;
+const unsigned kPrintfTypeOffset = 1 * kInstructionSize;
+const unsigned kPrintfLength = 2 * kInstructionSize;
+
+
+// The proper way to initialize a simulated system register (such as NZCV) is as
+// follows:
+//  SimSystemRegister nzcv = SimSystemRegister::DefaultValueFor(NZCV);
+class SimSystemRegister {
+ public:
+  // The default constructor represents a register which has no writable bits.
+  // It is not possible to set its value to anything other than 0.
+  SimSystemRegister() : value_(0), write_ignore_mask_(0xffffffff) { }
+
+  inline uint32_t RawValue() const {
+    return value_;
+  }
+
+  inline void SetRawValue(uint32_t new_value) {
+    value_ = (value_ & write_ignore_mask_) | (new_value & ~write_ignore_mask_);
+  }
+
+  inline uint32_t Bits(int msb, int lsb) const {
+    return unsigned_bitextract_32(msb, lsb, value_);
+  }
+
+  inline int32_t SignedBits(int msb, int lsb) const {
+    return signed_bitextract_32(msb, lsb, value_);
+  }
+
+  void SetBits(int msb, int lsb, uint32_t bits);
+
+  // Default system register values.
+  static SimSystemRegister DefaultValueFor(SystemRegister id);
+
+#define DEFINE_GETTER(Name, HighBit, LowBit, Func)                            \
+  inline uint32_t Name() const { return Func(HighBit, LowBit); }              \
+  inline void Set##Name(uint32_t bits) { SetBits(HighBit, LowBit, bits); }
+#define DEFINE_WRITE_IGNORE_MASK(Name, Mask)                                  \
+  static const uint32_t Name##WriteIgnoreMask = ~static_cast<uint32_t>(Mask);
+
+  SYSTEM_REGISTER_FIELDS_LIST(DEFINE_GETTER, DEFINE_WRITE_IGNORE_MASK)
+
+#undef DEFINE_ZERO_BITS
+#undef DEFINE_GETTER
+
+ protected:
+  // Most system registers only implement a few of the bits in the word. Other
+  // bits are "read-as-zero, write-ignored". The write_ignore_mask argument
+  // describes the bits which are not modifiable.
+  SimSystemRegister(uint32_t value, uint32_t write_ignore_mask)
+      : value_(value), write_ignore_mask_(write_ignore_mask) { }
+
+  uint32_t value_;
+  uint32_t write_ignore_mask_;
+};
+
+
+class Simulator : public DecoderVisitor {
+ public:
+  explicit Simulator(Decoder* decoder, FILE* stream = stdout);
+  ~Simulator();
+
+  void ResetState();
+
+  // TODO: We assume little endianness, and the way in which the members of this
+  // union overlay. Add tests to ensure this, or fix accessors to no longer
+  // require this assumption.
+  union SimRegister {
+    int64_t x;
+    int32_t w;
+  };
+
+  union SimFPRegister {
+    double d;
+    float s;
+  };
+
+  // Run the simulator.
+  virtual void Run();
+  void RunFrom(Instruction* first);
+
+  // Simulation helpers.
+  inline Instruction* pc() { return pc_; }
+  inline void set_pc(Instruction* new_pc) {
+    pc_ = new_pc;
+    pc_modified_ = true;
+  }
+
+  inline void increment_pc() {
+    if (!pc_modified_) {
+      pc_ = pc_->NextInstruction();
+    }
+
+    pc_modified_ = false;
+  }
+
+  inline void ExecuteInstruction() {
+    // The program counter should always be aligned.
+    ASSERT(IsWordAligned(pc_));
+    decoder_->Decode(pc_);
+    increment_pc();
+  }
+
+  // Declare all Visitor functions.
+  #define DECLARE(A)  void Visit##A(Instruction* instr);
+  VISITOR_LIST(DECLARE)
+  #undef DECLARE
+
+  // Register accessors.
+  inline int32_t wreg(unsigned code,
+                      Reg31Mode r31mode = Reg31IsZeroRegister) const {
+    ASSERT(code < kNumberOfRegisters);
+    if ((code == 31) && (r31mode == Reg31IsZeroRegister)) {
+      return 0;
+    }
+    return registers_[code].w;
+  }
+
+  inline int64_t xreg(unsigned code,
+                      Reg31Mode r31mode = Reg31IsZeroRegister) const {
+    ASSERT(code < kNumberOfRegisters);
+    if ((code == 31) && (r31mode == Reg31IsZeroRegister)) {
+      return 0;
+    }
+    return registers_[code].x;
+  }
+
+  inline int64_t reg(unsigned size,
+                     unsigned code,
+                     Reg31Mode r31mode = Reg31IsZeroRegister) const {
+    switch (size) {
+      case kWRegSize: return wreg(code, r31mode) & kWRegMask;
+      case kXRegSize: return xreg(code, r31mode);
+      default:
+        UNREACHABLE();
+        return 0;
+    }
+  }
+
+  inline void set_wreg(unsigned code, int32_t value,
+                       Reg31Mode r31mode = Reg31IsZeroRegister) {
+    ASSERT(code < kNumberOfRegisters);
+    if ((code == kZeroRegCode) && (r31mode == Reg31IsZeroRegister)) {
+      return;
+    }
+    registers_[code].x = 0;  // First clear the register top bits.
+    registers_[code].w = value;
+  }
+
+  inline void set_xreg(unsigned code, int64_t value,
+                       Reg31Mode r31mode = Reg31IsZeroRegister) {
+    ASSERT(code < kNumberOfRegisters);
+    if ((code == kZeroRegCode) && (r31mode == Reg31IsZeroRegister)) {
+      return;
+    }
+    registers_[code].x = value;
+  }
+
+  inline void set_reg(unsigned size, unsigned code, int64_t value,
+                      Reg31Mode r31mode = Reg31IsZeroRegister) {
+    switch (size) {
+      case kWRegSize:
+        return set_wreg(code, static_cast<int32_t>(value & 0xffffffff),
+                        r31mode);
+      case kXRegSize:
+        return set_xreg(code, value, r31mode);
+      default:
+        UNREACHABLE();
+        break;
+    }
+  }
+
+  #define REG_ACCESSORS(N)                                 \
+  inline int32_t w##N() { return wreg(N); }                \
+  inline int64_t x##N() { return xreg(N); }                \
+  inline void set_w##N(int32_t val) { set_wreg(N, val); }  \
+  inline void set_x##N(int64_t val) { set_xreg(N, val); }
+  REGISTER_CODE_LIST(REG_ACCESSORS)
+  #undef REG_ACCESSORS
+
+  // Aliases.
+  #define REG_ALIAS_ACCESSORS(N, wname, xname)                \
+  inline int32_t wname() { return wreg(N); }                  \
+  inline int64_t xname() { return xreg(N); }                  \
+  inline void set_##wname(int32_t val) { set_wreg(N, val); }  \
+  inline void set_##xname(int64_t val) { set_xreg(N, val); }
+  REG_ALIAS_ACCESSORS(30, wlr, lr);
+  #undef REG_ALIAS_ACCESSORS
+
+  // The stack is a special case in aarch64.
+  inline int32_t wsp() { return wreg(31, Reg31IsStackPointer); }
+  inline int64_t sp() { return xreg(31, Reg31IsStackPointer); }
+  inline void set_wsp(int32_t val) {
+    set_wreg(31, val, Reg31IsStackPointer);
+  }
+  inline void set_sp(int64_t val) {
+    set_xreg(31, val, Reg31IsStackPointer);
+  }
+
+  // FPRegister accessors.
+  inline float sreg(unsigned code) const {
+    ASSERT(code < kNumberOfFPRegisters);
+    return fpregisters_[code].s;
+  }
+
+  inline uint32_t sreg_bits(unsigned code) const {
+    return float_to_rawbits(sreg(code));
+  }
+
+  inline double dreg(unsigned code) const {
+    ASSERT(code < kNumberOfFPRegisters);
+    return fpregisters_[code].d;
+  }
+
+  inline uint64_t dreg_bits(unsigned code) const {
+    return double_to_rawbits(dreg(code));
+  }
+
+  inline double fpreg(unsigned size, unsigned code) const {
+    switch (size) {
+      case kSRegSize: return sreg(code);
+      case kDRegSize: return dreg(code);
+      default: {
+        UNREACHABLE();
+        return 0.0;
+      }
+    }
+  }
+
+  inline void set_sreg(unsigned code, float val) {
+    ASSERT(code < kNumberOfFPRegisters);
+    // Ensure that the upper word is set to 0.
+    set_dreg_bits(code, 0);
+
+    fpregisters_[code].s = val;
+  }
+
+  inline void set_sreg_bits(unsigned code, uint32_t rawbits) {
+    ASSERT(code < kNumberOfFPRegisters);
+    // Ensure that the upper word is set to 0.
+    set_dreg_bits(code, 0);
+
+    set_sreg(code, rawbits_to_float(rawbits));
+  }
+
+  inline void set_dreg(unsigned code, double val) {
+    ASSERT(code < kNumberOfFPRegisters);
+    fpregisters_[code].d = val;
+  }
+
+  inline void set_dreg_bits(unsigned code, uint64_t rawbits) {
+    ASSERT(code < kNumberOfFPRegisters);
+    set_dreg(code, rawbits_to_double(rawbits));
+  }
+
+  inline void set_fpreg(unsigned size, unsigned code, double value) {
+    switch (size) {
+      case kSRegSize:
+        return set_sreg(code, value);
+      case kDRegSize:
+        return set_dreg(code, value);
+      default:
+        UNREACHABLE();
+        break;
+    }
+  }
+
+  #define FPREG_ACCESSORS(N)                             \
+  inline float s##N() { return sreg(N); }                \
+  inline double d##N() { return dreg(N); }               \
+  inline void set_s##N(float val) { set_sreg(N, val); }  \
+  inline void set_d##N(double val) { set_dreg(N, val); }
+  REGISTER_CODE_LIST(FPREG_ACCESSORS)
+  #undef FPREG_ACCESSORS
+
+  bool N() { return nzcv_.N() != 0; }
+  bool Z() { return nzcv_.Z() != 0; }
+  bool C() { return nzcv_.C() != 0; }
+  bool V() { return nzcv_.V() != 0; }
+  SimSystemRegister& nzcv() { return nzcv_; }
+
+  // TODO(jbramley): Find a way to make the fpcr_ members return the proper
+  // types, so this accessor is not necessary.
+  FPRounding RMode() { return static_cast<FPRounding>(fpcr_.RMode()); }
+  SimSystemRegister& fpcr() { return fpcr_; }
+
+  // Debug helpers
+  void PrintSystemRegisters(bool print_all = false);
+  void PrintRegisters(bool print_all_regs = false);
+  void PrintFPRegisters(bool print_all_regs = false);
+  void PrintProcessorState();
+
+  static const char* WRegNameForCode(unsigned code,
+                                     Reg31Mode mode = Reg31IsZeroRegister);
+  static const char* XRegNameForCode(unsigned code,
+                                     Reg31Mode mode = Reg31IsZeroRegister);
+  static const char* SRegNameForCode(unsigned code);
+  static const char* DRegNameForCode(unsigned code);
+  static const char* VRegNameForCode(unsigned code);
+
+  inline bool coloured_trace() { return coloured_trace_; }
+  inline void set_coloured_trace(bool value) { coloured_trace_ = value; }
+
+  inline bool disasm_trace() { return disasm_trace_; }
+  inline void set_disasm_trace(bool value) {
+    if (value != disasm_trace_) {
+      if (value) {
+        decoder_->InsertVisitorBefore(print_disasm_, this);
+      } else {
+        decoder_->RemoveVisitor(print_disasm_);
+      }
+      disasm_trace_ = value;
+    }
+  }
+  inline void set_instruction_stats(bool value) {
+    if (value != instruction_stats_) {
+      if (value) {
+        decoder_->AppendVisitor(instrumentation_);
+      } else {
+        decoder_->RemoveVisitor(instrumentation_);
+      }
+      instruction_stats_ = value;
+    }
+  }
+
+ protected:
+  // Simulation helpers ------------------------------------
+  bool ConditionPassed(Condition cond) {
+    switch (cond) {
+      case eq:
+        return Z();
+      case ne:
+        return !Z();
+      case hs:
+        return C();
+      case lo:
+        return !C();
+      case mi:
+        return N();
+      case pl:
+        return !N();
+      case vs:
+        return V();
+      case vc:
+        return !V();
+      case hi:
+        return C() && !Z();
+      case ls:
+        return !(C() && !Z());
+      case ge:
+        return N() == V();
+      case lt:
+        return N() != V();
+      case gt:
+        return !Z() && (N() == V());
+      case le:
+        return !(!Z() && (N() == V()));
+      case nv:  // Fall through.
+      case al:
+        return true;
+      default:
+        UNREACHABLE();
+        return false;
+    }
+  }
+
+  bool ConditionFailed(Condition cond) {
+    return !ConditionPassed(cond);
+  }
+
+  void AddSubHelper(Instruction* instr, int64_t op2);
+  int64_t AddWithCarry(unsigned reg_size,
+                       bool set_flags,
+                       int64_t src1,
+                       int64_t src2,
+                       int64_t carry_in = 0);
+  void LogicalHelper(Instruction* instr, int64_t op2);
+  void ConditionalCompareHelper(Instruction* instr, int64_t op2);
+  void LoadStoreHelper(Instruction* instr,
+                       int64_t offset,
+                       AddrMode addrmode);
+  void LoadStorePairHelper(Instruction* instr, AddrMode addrmode);
+  uint8_t* AddressModeHelper(unsigned addr_reg,
+                             int64_t offset,
+                             AddrMode addrmode);
+
+  uint64_t MemoryRead(const uint8_t* address, unsigned num_bytes);
+  uint8_t MemoryRead8(uint8_t* address);
+  uint16_t MemoryRead16(uint8_t* address);
+  uint32_t MemoryRead32(uint8_t* address);
+  float MemoryReadFP32(uint8_t* address);
+  uint64_t MemoryRead64(uint8_t* address);
+  double MemoryReadFP64(uint8_t* address);
+
+  void MemoryWrite(uint8_t* address, uint64_t value, unsigned num_bytes);
+  void MemoryWrite32(uint8_t* address, uint32_t value);
+  void MemoryWriteFP32(uint8_t* address, float value);
+  void MemoryWrite64(uint8_t* address, uint64_t value);
+  void MemoryWriteFP64(uint8_t* address, double value);
+
+  int64_t ShiftOperand(unsigned reg_size,
+                       int64_t value,
+                       Shift shift_type,
+                       unsigned amount);
+  int64_t Rotate(unsigned reg_width,
+                 int64_t value,
+                 Shift shift_type,
+                 unsigned amount);
+  int64_t ExtendValue(unsigned reg_width,
+                      int64_t value,
+                      Extend extend_type,
+                      unsigned left_shift = 0);
+
+  uint64_t ReverseBits(uint64_t value, unsigned num_bits);
+  uint64_t ReverseBytes(uint64_t value, ReverseByteMode mode);
+
+  void FPCompare(double val0, double val1);
+  double FPRoundInt(double value, FPRounding round_mode);
+  double FPToDouble(float value);
+  float FPToFloat(double value, FPRounding round_mode);
+  double FixedToDouble(int64_t src, int fbits, FPRounding round_mode);
+  double UFixedToDouble(uint64_t src, int fbits, FPRounding round_mode);
+  float FixedToFloat(int64_t src, int fbits, FPRounding round_mode);
+  float UFixedToFloat(uint64_t src, int fbits, FPRounding round_mode);
+  int32_t FPToInt32(double value, FPRounding rmode);
+  int64_t FPToInt64(double value, FPRounding rmode);
+  uint32_t FPToUInt32(double value, FPRounding rmode);
+  uint64_t FPToUInt64(double value, FPRounding rmode);
+  double FPMax(double a, double b);
+  double FPMin(double a, double b);
+
+  // Pseudo Printf instruction
+  void DoPrintf(Instruction* instr);
+
+  // Processor state ---------------------------------------
+
+  // Output stream.
+  FILE* stream_;
+  PrintDisassembler* print_disasm_;
+
+  // Instruction statistics instrumentation.
+  Instrument* instrumentation_;
+
+  // General purpose registers. Register 31 is the stack pointer.
+  SimRegister registers_[kNumberOfRegisters];
+
+  // Floating point registers
+  SimFPRegister fpregisters_[kNumberOfFPRegisters];
+
+  // Program Status Register.
+  // bits[31, 27]: Condition flags N, Z, C, and V.
+  //               (Negative, Zero, Carry, Overflow)
+  SimSystemRegister nzcv_;
+
+  // Floating-Point Control Register
+  SimSystemRegister fpcr_;
+
+  // Only a subset of FPCR features are supported by the simulator. This helper
+  // checks that the FPCR settings are supported.
+  //
+  // This is checked when floating-point instructions are executed, not when
+  // FPCR is set. This allows generated code to modify FPCR for external
+  // functions, or to save and restore it when entering and leaving generated
+  // code.
+  void AssertSupportedFPCR() {
+    ASSERT(fpcr().DN() == 0);             // No default-NaN support.
+    ASSERT(fpcr().FZ() == 0);             // No flush-to-zero support.
+    ASSERT(fpcr().RMode() == FPTieEven);  // Ties-to-even rounding only.
+
+    // The simulator does not support half-precision operations so fpcr().AHP()
+    // is irrelevant, and is not checked here.
+  }
+
+  static inline int CalcNFlag(uint64_t result, unsigned reg_size) {
+    return (result >> (reg_size - 1)) & 1;
+  }
+
+  static inline int CalcZFlag(uint64_t result) {
+    return result == 0;
+  }
+
+  static const uint32_t kConditionFlagsMask = 0xf0000000;
+
+  // Stack
+  byte* stack_;
+  static const int stack_protection_size_ = 256;
+  // 2 KB stack.
+  static const int stack_size_ = 2 * 1024 + 2 * stack_protection_size_;
+  byte* stack_limit_;
+
+  Decoder* decoder_;
+  // Indicates if the pc has been modified by the instruction and should not be
+  // automatically incremented.
+  bool pc_modified_;
+  Instruction* pc_;
+
+  static const char* xreg_names[];
+  static const char* wreg_names[];
+  static const char* sreg_names[];
+  static const char* dreg_names[];
+  static const char* vreg_names[];
+
+  static const Instruction* kEndOfSimAddress;
+
+ private:
+  bool coloured_trace_;
+
+  // Indicates whether the disassembly trace is active.
+  bool disasm_trace_;
+
+  // Indicates whether the instruction instrumentation is active.
+  bool instruction_stats_;
+};
+}  // namespace vixl
+
+#endif  // VIXL_A64_SIMULATOR_A64_H_
diff --git a/disas/libvixl/src/globals.h b/disas/libvixl/src/globals.h
new file mode 100644
index 0000000..859ea69
--- /dev/null
+++ b/disas/libvixl/src/globals.h
@@ -0,0 +1,66 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_GLOBALS_H
+#define VIXL_GLOBALS_H
+
+// Get the standard printf format macros for C99 stdint types.
+#define __STDC_FORMAT_MACROS
+#include <inttypes.h>
+
+#include <assert.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include "platform.h"
+
+
+typedef uint8_t byte;
+
+const int KBytes = 1024;
+const int MBytes = 1024 * KBytes;
+const int GBytes = 1024 * MBytes;
+
+  #define ABORT() printf("in %s, line %i", __FILE__, __LINE__); abort()
+#ifdef DEBUG
+  #define ASSERT(condition) assert(condition)
+  #define CHECK(condition) ASSERT(condition)
+  #define UNIMPLEMENTED() printf("UNIMPLEMENTED\t"); ABORT()
+  #define UNREACHABLE() printf("UNREACHABLE\t"); ABORT()
+#else
+  #define ASSERT(condition) ((void) 0)
+  #define CHECK(condition) assert(condition)
+  #define UNIMPLEMENTED() ((void) 0)
+  #define UNREACHABLE() ((void) 0)
+#endif
+
+template <typename T> inline void USE(T) {}
+
+#define ALIGNMENT_EXCEPTION() printf("ALIGNMENT EXCEPTION\t"); ABORT()
+
+#endif  // VIXL_GLOBALS_H
diff --git a/disas/libvixl/src/platform.h b/disas/libvixl/src/platform.h
new file mode 100644
index 0000000..a2600f3
--- /dev/null
+++ b/disas/libvixl/src/platform.h
@@ -0,0 +1,43 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef PLATFORM_H
+#define PLATFORM_H
+
+// Define platform specific functionalities.
+
+namespace vixl {
+#ifdef USE_SIMULATOR
+// Currently we assume running the simulator implies running on x86 hardware.
+inline void HostBreakpoint() { asm("int3"); }
+#else
+inline void HostBreakpoint() {
+  // TODO: Implement HostBreakpoint on a64.
+}
+#endif
+}  // namespace vixl
+
+#endif
diff --git a/disas/libvixl/src/utils.cc b/disas/libvixl/src/utils.cc
new file mode 100644
index 0000000..6f85e61
--- /dev/null
+++ b/disas/libvixl/src/utils.cc
@@ -0,0 +1,120 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "utils.h"
+#include <stdio.h>
+
+namespace vixl {
+
+uint32_t float_to_rawbits(float value) {
+  uint32_t bits = 0;
+  memcpy(&bits, &value, 4);
+  return bits;
+}
+
+
+uint64_t double_to_rawbits(double value) {
+  uint64_t bits = 0;
+  memcpy(&bits, &value, 8);
+  return bits;
+}
+
+
+float rawbits_to_float(uint32_t bits) {
+  float value = 0.0;
+  memcpy(&value, &bits, 4);
+  return value;
+}
+
+
+double rawbits_to_double(uint64_t bits) {
+  double value = 0.0;
+  memcpy(&value, &bits, 8);
+  return value;
+}
+
+
+int CountLeadingZeros(uint64_t value, int width) {
+  ASSERT((width == 32) || (width == 64));
+  int count = 0;
+  uint64_t bit_test = 1UL << (width - 1);
+  while ((count < width) && ((bit_test & value) == 0)) {
+    count++;
+    bit_test >>= 1;
+  }
+  return count;
+}
+
+
+int CountLeadingSignBits(int64_t value, int width) {
+  ASSERT((width == 32) || (width == 64));
+  if (value >= 0) {
+    return CountLeadingZeros(value, width) - 1;
+  } else {
+    return CountLeadingZeros(~value, width) - 1;
+  }
+}
+
+
+int CountTrailingZeros(uint64_t value, int width) {
+  ASSERT((width == 32) || (width == 64));
+  int count = 0;
+  while ((count < width) && (((value >> count) & 1) == 0)) {
+    count++;
+  }
+  return count;
+}
+
+
+int CountSetBits(uint64_t value, int width) {
+  // TODO: Other widths could be added here, as the implementation already
+  // supports them.
+  ASSERT((width == 32) || (width == 64));
+
+  // Mask out unused bits to ensure that they are not counted.
+  value &= (0xffffffffffffffffUL >> (64-width));
+
+  // Add up the set bits.
+  // The algorithm works by adding pairs of bit fields together iteratively,
+  // where the size of each bit field doubles each time.
+  // An example for an 8-bit value:
+  // Bits:  h  g  f  e  d  c  b  a
+  //         \ |   \ |   \ |   \ |
+  // value = h+g   f+e   d+c   b+a
+  //            \    |      \    |
+  // value =   h+g+f+e     d+c+b+a
+  //                  \          |
+  // value =       h+g+f+e+d+c+b+a
+  value = ((value >> 1) & 0x5555555555555555) + (value & 0x5555555555555555);
+  value = ((value >> 2) & 0x3333333333333333) + (value & 0x3333333333333333);
+  value = ((value >> 4) & 0x0f0f0f0f0f0f0f0f) + (value & 0x0f0f0f0f0f0f0f0f);
+  value = ((value >> 8) & 0x00ff00ff00ff00ff) + (value & 0x00ff00ff00ff00ff);
+  value = ((value >> 16) & 0x0000ffff0000ffff) + (value & 0x0000ffff0000ffff);
+  value = ((value >> 32) & 0x00000000ffffffff) + (value & 0x00000000ffffffff);
+
+  return value;
+}
+}  // namespace vixl
diff --git a/disas/libvixl/src/utils.h b/disas/libvixl/src/utils.h
new file mode 100644
index 0000000..4e0b367
--- /dev/null
+++ b/disas/libvixl/src/utils.h
@@ -0,0 +1,126 @@ 
+// Copyright 2013, ARM Limited
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are met:
+//
+//   * Redistributions of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//   * Redistributions in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//   * Neither the name of ARM Limited nor the names of its contributors may be
+//     used to endorse or promote products derived from this software without
+//     specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS CONTRIBUTORS "AS IS" AND
+// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef VIXL_UTILS_H
+#define VIXL_UTILS_H
+
+
+#include <string.h>
+#include "globals.h"
+
+namespace vixl {
+
+// Check number width.
+inline bool is_intn(unsigned n, int64_t x) {
+  ASSERT((0 < n) && (n < 64));
+  int64_t limit = 1L << (n - 1);
+  return (-limit <= x) && (x < limit);
+}
+
+inline bool is_uintn(unsigned n, int64_t x) {
+  ASSERT((0 < n) && (n < 64));
+  return !(x >> n);
+}
+
+inline unsigned truncate_to_intn(unsigned n, int64_t x) {
+  ASSERT((0 < n) && (n < 64));
+  return (x & ((1L << n) - 1));
+}
+
+#define INT_1_TO_63_LIST(V)                                                    \
+V(1)  V(2)  V(3)  V(4)  V(5)  V(6)  V(7)  V(8)                                 \
+V(9)  V(10) V(11) V(12) V(13) V(14) V(15) V(16)                                \
+V(17) V(18) V(19) V(20) V(21) V(22) V(23) V(24)                                \
+V(25) V(26) V(27) V(28) V(29) V(30) V(31) V(32)                                \
+V(33) V(34) V(35) V(36) V(37) V(38) V(39) V(40)                                \
+V(41) V(42) V(43) V(44) V(45) V(46) V(47) V(48)                                \
+V(49) V(50) V(51) V(52) V(53) V(54) V(55) V(56)                                \
+V(57) V(58) V(59) V(60) V(61) V(62) V(63)
+
+#define DECLARE_IS_INT_N(N)                                                    \
+inline bool is_int##N(int64_t x) { return is_intn(N, x); }
+#define DECLARE_IS_UINT_N(N)                                                   \
+inline bool is_uint##N(int64_t x) { return is_uintn(N, x); }
+#define DECLARE_TRUNCATE_TO_INT_N(N)                                           \
+inline int truncate_to_int##N(int x) { return truncate_to_intn(N, x); }
+INT_1_TO_63_LIST(DECLARE_IS_INT_N)
+INT_1_TO_63_LIST(DECLARE_IS_UINT_N)
+INT_1_TO_63_LIST(DECLARE_TRUNCATE_TO_INT_N)
+#undef DECLARE_IS_INT_N
+#undef DECLARE_IS_UINT_N
+#undef DECLARE_TRUNCATE_TO_INT_N
+
+// Bit field extraction.
+inline uint32_t unsigned_bitextract_32(int msb, int lsb, uint32_t x) {
+  return (x >> lsb) & ((1 << (1 + msb - lsb)) - 1);
+}
+
+inline uint64_t unsigned_bitextract_64(int msb, int lsb, uint64_t x) {
+  return (x >> lsb) & ((static_cast<uint64_t>(1) << (1 + msb - lsb)) - 1);
+}
+
+inline int32_t signed_bitextract_32(int msb, int lsb, int32_t x) {
+  return (x << (31 - msb)) >> (lsb + 31 - msb);
+}
+
+inline int64_t signed_bitextract_64(int msb, int lsb, int64_t x) {
+  return (x << (63 - msb)) >> (lsb + 63 - msb);
+}
+
+// floating point representation
+uint32_t float_to_rawbits(float value);
+uint64_t double_to_rawbits(double value);
+float rawbits_to_float(uint32_t bits);
+double rawbits_to_double(uint64_t bits);
+
+// Bits counting.
+int CountLeadingZeros(uint64_t value, int width);
+int CountLeadingSignBits(int64_t value, int width);
+int CountTrailingZeros(uint64_t value, int width);
+int CountSetBits(uint64_t value, int width);
+
+// Pointer alignment
+// TODO: rename/refactor to make it specific to instructions.
+template<typename T>
+bool IsWordAligned(T pointer) {
+  ASSERT(sizeof(pointer) == sizeof(intptr_t));   // NOLINT(runtime/sizeof)
+  return (reinterpret_cast<intptr_t>(pointer) & 3) == 0;
+}
+
+// Increment a pointer until it has the specified alignment.
+template<class T>
+T AlignUp(T pointer, size_t alignment) {
+  ASSERT(sizeof(pointer) == sizeof(uintptr_t));
+  uintptr_t pointer_raw = reinterpret_cast<uintptr_t>(pointer);
+  size_t align_step = (alignment - pointer_raw) % alignment;
+  ASSERT((pointer_raw + align_step) % alignment == 0);
+  return reinterpret_cast<T>(pointer_raw + align_step);
+}
+
+
+}  // namespace vixl
+
+#endif  // VIXL_UTILS_H