A Jacobi Method by Blocks on a Mesh of Processors

Domingo Gimenez
Departamento de Informatica y Sistemas
Univ. de Murcia
Aptdo 4021
30001 Murcia
Spain
domingo@dif.um.es
Vicente Hernandez
Departmento de Sistemas Informaticos y Computacion
Univ. Politecnica de Valencia
Aptdo 22012
46071 Volencia
Spain
cpvhg@dsic.upv.es
Robert A. van de Geijn
Department of Computer Sciences
University of Texas
Austin, TX 78712
rvdg@cs.utexas.edu
Antonio M. Vidal
Departmento de Sistemas Informaticos y Computacion
Univ. Politecnica de Valencia
Aptdo 22012
46071 Volencia
Spain
cpavm@dsic.upv.es

Abstract

In this paper, we study the parallelization of the Jacobi method to solve the symmetric eigenvalue problem on a mesh of processors. To solve this problem obtaining a theoretical efficiency of 100%, it is necessary to exploit the symmetry of the matrix. The only previous algorithms we know exploiting the symmetry on multicomputers is that in [18], but that algorithm uses a storage scheme adequate for a logical ring of processors, so having low scalability. In this paper we show how matrix symmetry can be exploited on a logical mesh of processors obtaining a higher scalability that that obtained with the algorithm in [18]. In addition, we show how the storage scheme exploiting the symmetry can be combined with a scheme by blocks to obtain a highly efficient and scalable Jacobi method for solving the symmetrix eigenvalue problem for distributed memory parallel computers. We report performance results from the Intel Touchstone DELTA.

D. Gimenez, V. Hernandez, R. van de Geijn, and A. Vidal, A Jacobi Method by Blocks on a Mesh of Processors, submitted to Concurrency: Practice and Experience.