El projecte PyPy pretén crear una implementació de Python més ràpida i
flexible. Amb aquesta idea en ment es reimplementa un intèrpret de Python en
el mateix llenguatge de programació. Potser algun dia serà més
ràpida que la implementació en C, qui sap...
Un sprint per la gent de PyPy no és sols
programar com a bojos: és un bon moment per donar a coneixer el seu
projecte. Ricardo Galli va conseguir
que aquesta gent pogués usar el laboratori conegut com
AulaLinux per dur a terme les seves tasques de desenvolupament i
organització. Abans de marxar ens volen oferir una conferència on
s'explicaran els principis del projecte, en què està basat i perquè fan
conses tan curioses com aquesta. La ponència començarà a les 15:30h, durarà
casi una hora i després hi haurà torn de preguntes. La major part de la
ponència serà en anglès.
Cartell anunciant l'acte.
Aquest és el seu missatge sobre l'acte del divendres:
27th th of January 2006/UIB
Title: PyPy and the Muenchhausen approach
Talkers: Carl Friedrich Bolz, Armin Rigo, Samuele Pedronis
Time: 45 minutes + questions
Abstract:
The PyPy team is "sprinting" (one week programming workshop) at the
UIB this week, working in the Aula GNU/Linux. The team has been
invited by Ricardo Galli at the Computer Science departement.
We present this talk as a good way to get to understand the novel
language design aspects of the project as well as our contribution
back to the UIB for graciously hosting PyPy.
We present our first self-contained Python virtual machine that uses
parts of itself to translate itself to low level languages ("the
Muenchhausen approach"). The PyPy approach could solve problems at
language/interpreter-level that formerly required complex
frameworkish solutions at user-level.
Description:
PyPy is a reimplementation of Python written in Python
itself, flexible and easy to experiment with. Our
long-term goals are to target a large variety of
platforms, small and large, by adapting the compiler
toolsuite we developed to produce custom Python versions.
Platform, Memory and Threading models will become
aspects of the translation process - as opposed to
encoding low level details into a language implementation
itself.
Basically, we think it's a good way to avoid writing
n x m x o interpreters for n dynamic languages and m
platforms with o crucial design decisions. In PyPy
any one of these can be changed independently.
We are going to briefly describe the concepts of
object spaces, abstract interpretation and translation
aspects and how they led us to a first self-contained
very compliant Python implementation in August 2005,
completely independent from the current mainstream
CPython implementation. We go through a translation
example of a Python program with control-flow-graphs
and the according translated lowlevel C and
LLVM (Low level Virtual Machine) code.
We'll also try to relate PyPy's architectural concepts
(known roughly for 2-3 years now) to similar upcoming
concepts in e.g. pugs/Perl 6 development and we'll give an
outlook on our starting Just-In-Time Compiler efforts and
approaches.
Lastly, we intend to discuss experimental new
language/interpreter-level solutions to long-standing problems such as
distributed computing,persistence and security/sandboxing.
| 
