Main

• CPython

  • The main implementation of Python. It's also interesting to note that the language semantics are mostly based on mimicking CPython for other languages.

  • Curiosities:

    • CPython has a GIL (Global Interpreter Lock)

• Numba

  • Numba is an LLVM based JIT for CPython. It's numpy-aware and especially suited for algorithmic code.

• Cython:

  • A static compiler targeting CPython (with initial support for PyPy).
  • Can compile Python code and mixed Cython/C/C++ code.
  • Can also be used as a way to create bindings for existing C/C++ libraries.
  • It's predecessor was Pyrex (which is no longer actively maintaned).
  • Supports Python 2.6 and 3.2 onwards.

• MicroPyton:

  • Micro Python is an implementation of Python 3 optimized to run on a microcontroller.

• Nuitka:

  • Nuitka targets compiling Python code using the libpython library (thus giving a very compatible support for Python code).
  • As it uses libpython, much of its code is still interpreted (and development is going into providing more optimizations).
  • Given its compliant nature, it can be used as a way for distributing Python code (as it's effectively a way to compile Python code).

• PyPy:

  • PyPy is a Python interpreter with a builtin JIT which is able to grok Python code very well and make run it fast.
  • It can use cffi to interface with C code.
  • It also has a GIL (Global Interpreter Lock), but is working in a branch to create a GIL-less version using STM (Software Transactional Memory).

• Stackless Python:

  • Stackless Python is a CPython fork which has its own stack which can do context-switching to provide micro-threads and continuations.
  • Alternatives to Stackless Python exist as libraries: greenlet or asyncio can be used for that purpose.

Experimental (in development)

• Pyston:

  • Pyston is a Python implementation which has a builtin JIT (using LLVM). It's backed by Dropbox.

Integration with other languages

• Java: Jython

• .NET: IronPython

• JavaScript: pyjs

• JavaScript: RapydScript

Deprecated/Unsupported libraries

Over the years, other technologies were also created (and many resemble the ones above, although they're no longer actively mantained).

• Psyco

  Psyco was an optimizing JIT for CPython (developed from 2006-2010 by Armin Rigo) which was later abandoned because it was increasingly harder to maintain given its approach. Also, its main developer went on to work on PyPy ;)

• Shedskin

  Developed from 2008-2013 by Mark Dufour, it targeted translating pure (but implicitly typed) Python into C++.

• Unladen Swallow

  Back in 2009, Google sponsored a project which targeted integrating a JIT into CPython (using LLVM). The project was later abandoned (although it had nice side effects, such as improving LLVM and providing a structure for performance tests).