Parallel algorithms for reduction of a symmetric matrix to tridiagonal form on a shared memory multiprocessor

Abstract

This study is concerned with parallel algorithms for the reduction of a real symmetric matrix to tridiagonal form. A variety of algorithms are investigated, which involve varying amounts of overlap between different parts of the calculation. Empirical comparison was carried out using C++ and the THREADS package oil a shared memory Encore Multimax multiprocessor [Encore Computer Corporation, Encore Parallel Threads Manual No. 724-06210 Rev. A, 1988]. In this testing the filial version which involves most overlap was found to be the most efficient algorithm, and its efficiency is very high when the algorithm has some minor changes did give some improvements. (c) 2004 Elsevier Inc. All rights reserved.