============================
PyPy
============================

:Authors: Maciej Fijalkowski
:Location: Adam Mickiewicz University
:Date: 15.IV.2007

Next 45 minutes
=======================================

* Quick intro and motivation, including
  CPython problems

* Introduction to features unique to PyPy,
  including the JIT

* A little talk about our plans

What is PyPy?
==============

* PyPy is:
* A Python interpreter
* A very
  flexible compiler framework (with some
  features that are especially useful for
  implementing interpreters)
* An open source project (MIT license)
* A STREP (''Specific Targeted REsearch
  Project''), partially funded by the EU
* A lot of fun!

About the project
====================

* Distributed -- the 12 paid developers live in
  6 countries, contributers from many more
* Sprint driven development -- focussed week
  long coding sessions, every ~6 weeks during
  funding period
* Extreme Programming practices: pair
  programming, test-driven development

Motivation
===========

* PyPy grew out of a desire to modify/extend
  the implementation of Python, for example to:
* increase performance (psyco-style JIT
  compilation, better garbage collectors)
* add expressiveness (stackless-style
  coroutines, logic programming, distribution)
* ease porting (to new platforms like the
  JVM or CLI or to low memory situations)

Lofty goals, but first...
===========================
* CPython is a fine implementation of Python
  but:
* it's written in C, which makes porting to,
  for example, the CLI hard
* while psyco and stackless exist, they are
  very hard to maintain as Python evolves
* some implementation decisions are very
  hard to change (e.g. refcounting)

Compiler toolchain
===================

* Translates so called RPython
  (restricted Python) to lower level language
* Targets C, LLVM, CLI, JVM, JavaScript...
* Can translate any program, but
  is aimed at implementing interpreters

RPython
=========

* Starts from living python objects, so
  all dynamic tricks are allowed for initialization
* Full implicit type inference
* Type system like Java's, but with function pointers
  and more
* 20-60x faster than CPython

Things that make PyPy unique
===============================

* The Just-In-Time compiler (and the way it
  has been made)
* Transparent Proxies
* Runtime modifiable Grammar
* Thunk object space
* Multiple backends
* Logic programming
* whatever you imagine ...

Just-In-Time compiler generator
=================================

* generates just in time compiler given
  the interpreter source + few hints
* applicable to any possible interpreter
  (written in RPython)

Multiple object implementations
================================

* Every python-level object can have multiple
  implementations
* Examples are: String-key specialized dicts,
  ropes, tree-based dicts

Transparent proxies
====================

* Built on top of multiple object implementations

* Provides new implementation for each object

* can access everything in transparent manner,
  including such things like interpreter frames

Distribution
================

* Built on top of transparent proxies
* Can hide network layer
* Copies only small chunks of data,
  avoiding too much traffic
* Still very experimental!

Thunk object space
==================

* Built as different object space
  (object operation basis)

* Can do lazy computation of objects

Taint object space
===================

* Also built as different object space

* Few lines hack

* Can track values as taint, which cannot
  cross I/O borders implicitely

JavaScript backend
===================

* translates RPython to JavaScript

* has built-in AJAX functionality,
  which maps method calls to
  JSON calls

* can be tested on top of CPython
  (common case in PyPy)

Interpreter status
==========================

* We still use CPython :)

* for some specific use cases,
  we use PyPy when CPython is
  incapable

* It's a very friendly platform
  for experiments

* Threads and extension modules
  are not working

Interpreters in RPython
=============================

* Useful
* There is a working python interpreter
* prolog interpreter
* JavaScript interpreter coming soon
* Ruby? Perl?? (with JIT almost for free)

Test tools
===========

* We use py.test, developed together with PyPy
* distributed testing
* easy to extend (testing docs, JavaScript snippets etc.)

What future holds
====================

* More JIT! (faster than C)
* Extension modules
* Holiday! (funding period ended in March)

Further points of interest
==============================

* http://codespeak.net/pypy
* http://play1.codespeak.net/
* #pypy on freenode
* pypy-dev@codespeak.net