Resources for Parallel Computing in Python

Authors:	Lisandro Dalcin Rodrigo Paz Mario Storti
Organization:	International Center for Computer Methods in Engineering (CIMEC)
Address:	PTLC, (3000) Santa Fe, Argentina.
Contact:	dalcinl@users.sourceforge.net
Web Site:	http://www.cimec.org.ar/python/
Date:	2007-26-04

Abstract

At CIMEC, we are experimenting with the Python programming language as a viable alternative for rapid parallel application development and unpainful integration of parallelized source code from different sources.

This page lists all available resources and provides links to download already developed components.

Contents

Objectives
Get the Software

Objectives

Our efforts have been directed toward the development of the following components:

A port to MPI libraries (with emphasis in the widely used implementations MPICH and LAM, and the recent Open MPI). We want to achieve full support for managemet of groups, intra/inter-communicators and virtual topologies as well as the communication of general Python object between processes.
A port to PETSc libraries to provide capabilities for the flexible, efficient and scalable solution of large systems of linear and non-linear equations and integration with current Python packages specialized in numerical computations (as NumPy).
A port to ParMETIS libraries for parallel partition of graphs and meshes, which repeatedly arise in scientific applications modeled by partial differential equations.

Get the Software

MPI for Python
- Web Site: http://mpi4py.scipy.org
- Download: http://cheeseshop.python.org/pypi/mpi4py
PETSc for Python
- Web Site: http://code.google.com/p/petsc4py
- Download: http://code.google.com/p/petsc4py/downloads/list
ParMETIS for Python

Not available yet. Please contact developers if you really want to try it.