Multidimensional integration in a heterogeneous network environment
Computational Physics
2009-10-30 v1 High Energy Physics - Phenomenology
Abstract
We consider several issues related to the multidimensional integration using a network of heterogeneous computers. Based on these considerations, we develop a new general purpose scheme which can significantly reduce the time needed for evaluation of integrals with CPU intensive integrands. This scheme is a parallel version of the well-known adaptive Monte Carlo method (the VEGAS algorithm), and is incorporated into a new integration package which uses the standard set of message-passing routines in the PVM software system.
Cite
@article{arxiv.physics/9710017,
title = {Multidimensional integration in a heterogeneous network environment},
author = {S. Veseli},
journal= {arXiv preprint arXiv:physics/9710017},
year = {2009}
}
Comments
19 pages, latex, 5 postscript figures included