Templates for truncated QR, RQ, and QZ methods for large scale eigenproblems

Dan Sorensen

Rice University Department of Computational & Applied Mathematics - MS 134 6100 Main St. Houston, TX 77005-1892


Abstract

The implicit restarting technique that provides the algorithmic foundation for the ARPACK eigenvalue package and also for the eigs command in Matlab has recently been generalized. These generalizations provide a family of algorithms suitable for the standard problem Ax = lambda x and for the generalized problem Ax = lambda Bx. We shall discuss how to put these in a unified template form to interface with the required matrix-vector products and/or linear system solves.