This code measures the x86-64 disassembling performance of Capstone
(http://www.capstone-engine.org/), Intel XED (https://intelxed.github.io/), and
Zydis (https://zydis.re).

parse_elf.py takes ELF binary file as input and generates x86-64 instructions.
disasm_perf.c measures on average how many CPU cycles are taken to
disassemble each instruction.

Environment
-----------
The code was tested in Ubuntu 16.04 with 4.4.0 kernel. It should be able to run
in other Linux systems as well.

Compile code
------------
* Install Capstone. In Ubuntu, run "sudo apt-get install libcapstone-dev".
* Follow the steps in https://intelxed.github.io/build-manual/ to build and
  install Intel XED. Make sure headers and static libraries are installed in
  path where GCC can find.
* Follow the steps in https://zyantific.com/doc/zydis/ to build and install
  Zydis.
* Install The GNU Multiple Precision Arithmetic Library. In Ubuntu, run "sudo
  apt-get install libgmp3-dev".
* Run "make".

Run test
--------
./parse_elf.py <elf_file> | ./disasm_perf <option>

<elf_file> is the ELF executable's file path, e.g., /bin/ls. <option> selects
the disassembling engines and output modes.

Below are several examples:

./parse_elf.py <elf_file> | ./disasm_perf --capstone
Choose Capstone to disassemble x86-64 instructions.

./parse_elf.py <elf_file> | ./disasm_perf --capstone --verbose
Choose Capstone to disassemble x86-64 instructions, and print per-instruction 
disassembling time.

./parse_elf.py <elf_file> | ./disasm_perf --capstone --detail
Choose Capstone to disassemble x86-64 instructions, and get more details for
the disassembled instruction, e.g., implicit register read/write.

./parse_elf.py <elf_file> | ./disasm_perf --xed
Choose XED to disassemble the x86-64 instructions. XED always gives detailed
disassembling result.

./parse_elf.py <elf_file> | ./disasm_perf --zydis
Choose Zydis to disassemble the x86-64 instructions. Zydis always gives
detailed disassembling result.