Computing the singular values of a complex matrix using one-sided Jacobi method on the intel-paragon machine

Abstract

An algorithm for computing the singular values of a complex matrix based on Rijk's improvement of the one-sided Jacobi method [12] is developed.  The aim was to find the most efficient Jacobi algorithm for the complex case retaining the same characteristics of numerical stability. Its parallelisation is given for MIMD machines and in particular the tests have been done on the Intel-Paragon machine with an estimation of a maximum of 55% gain in computation time between the classical algorithm and the improved one.