The parallel computation of the smallest eigenpair of an acoustic problem with damping

Abstract

Acoustic problems with damping may give rise to large quadratic eigenproblems, which require efficient and parallelizable solution algorithms. This paper describes such an algorithm: the Jacobi–Davidson solution method for quadratic eigenproblems. In particular, it describes its parallelization according to the Bulk Synchronous Programming model, and its implementation on the massively parallel CRAY T3D. Experimental results for a large-scale acoustic problem show that the method is efficient and parallelizes well, i.e. scales almost linearly up to the maximum number of available processors. Copyright © 1999 John Wiley & Sons, Ltd.