[Back]


Talks and Poster Presentations (with Proceedings-Entry):

G. Barany:
"pylibjit: A JIT Compiler Library for Python";
Talk: 17. Kolloquium Programmiersprachen und Grundlagen der Programmierung (KPS 13), Lutherstadt Wittenberg, Deutschland; 2013-09-29 - 2013-10-02; in: "17. Kolloquium Programmiersprachen und Grundlagen der Programmierung (KPS 13)", (2013), 9 pages.



English abstract:
We present pylibjit, a Python library for generating machine code at load time or run time. The library has two distinct modes of use: First, it aims to provide a high-level interface for generating code at run time. This is achieved by using language features such as operator overloading and Python's context managers and decorators.

Second, Python's reflection features allow us to access functions' abstract syntax trees. pylibjit can make use of this to generate machine code for functions originally written in Python. Compiling a Python function can be as easy as attributing it with a decorator providing type information, without changing the function itself in any way. Such compiled functions are executed transparently within interpreted Python programs. This makes it convenient to develop applications in Python and then speed up only the hot spots using compilation.

The development of pylibjit is at a very early stage, but we can already present some working examples. For simple numeric programs, we achieve speedups of up to 50x over standard interpreted Python.

Keywords:
compilers, interpreters, jit, python, programming languages


Electronic version of the publication:
http://publik.tuwien.ac.at/files/PubDat_220249.pdf


Created from the Publication Database of the Vienna University of Technology.