Number of times this page has been accessed since July 21, 1995:

Parallel Matrix Decompositions: Have we been doing it all wrong?

Carter Edwards
Po Geng
Abani Patra
Texas Institute for Computational and Applied Mathematics
University of Texas at Austin
Austin, TX 78712
Robert van de Geijn
Texas Institute for Computational and Applied Mathematics
and
Department of Computer Sciences
University of Texas at Austin
Austin, TX 78712

Abstract

The basic premise of this essay is that traditional matrix decompositions for distributing matrices on distributed memory parallel architectures are in practice too restrictive. The primary problem lies with the fact that such decompositions start with the matrix, not with the underlying physical problem. Through a series of examples, we show how this hampers convenient interfaces between applications and libraries. In some instances, we show how it hampers performance in general. We propose a new data decomposition, {\em Physically Based Matrix Decompositions}, which appear to show promise for solving the encountered problems. Some traditionally used distributions are shown to be a special, but often unnatural, case of this more general class of decompositions.

C. Edwards, P. Geng, A. Patra, and R. van de Geijn, "Parallel Matrix Decompositions: have we been doing it all wrong?" TR-95-39, Department of Computer Sciences, University of Texas, Oct. 1995.