fcd is a LLVM-based native program optimizing decompiler. Most of the code is licensed under the GNU GPLv3 license, though some parts, like the executable parsing code, are licensed under a less restrictive scheme.
It implements pattern-independent structuring to provide a goto-free output (when decompilation succeeds).
Fcd currently only supports x86_64 programs. It implements a (partial) x86 emulator in C++, with one function per instruction, and compiles it to LLVM bytecode. To produce its output, fcd disassembles the target program and inlines each instruction's function's bytecode into a result function. This allows painless extension of the supported instruction set and powerful testing. Instructions that aren't implemented by the emulator are emitted as assembly statements; but since fcd uses Capstone, it can at least tell which registers the instruction reads and writes and still produce useful code when that happens.
Fcd's goal is not to produce fidel output of the disassembly. Rather, it aims to produce code that looks more natural and more readable. This is especially valuable in the case of obfuscated executables, where fidel output is next to useless.
However, in some conditions, it may make it harder to detect vulnerabilities. This means that fcd is usually more helpful for reverse engineering tasks than for exploitation tasks.
To assist in reverse engineering, fcd can load Python scripts as LLVM optimization passes to clean up custom obfuscation schemes.
Fcd is still a work in progress. You can contribute by finding ways to produce
a more readable output, by making it more reliable, or by tackling an issue
outlined in the FUTURE.md file.