Block-Arnoldi and Davidson methods for unsymmetric large eigenvalue problems
Abstract
We present two methods for computing the leading eigenpairs of large sparse unsymmetric matrices. Namely the block-Arnoldi method and an adaptationof the Davidson method to unsymmetric matrices. We give some theoretical results concerning the convergence of these two methods when restarting is used and discuss implementation aspects of the two methods on an Alliant FX/80. Finally some results of numerical tests on a variety of matrices including matrices from the Harwell-Boeing test collection in which we compare these these two methods are reported.