Domain decomposition for nonsymmetric systems of equations  Examples from computational fluid dynamics
Abstract
A blockpreconditioned Krylov method which combines features of several previously developed techniques in domain decomposition and iterative methods for large sparse linear systems is described and applied to a few illustrative problems. The main motivation of the work is to examine the gracefulness of parallelization under the domain decomposition paradigm of the solution of systems of equations typical of finitedifferenced fluid dynamical applications. Such systems lie outside of the realm of selfadjoint scalar elliptic equations for which most of the theory has been developed, and the present contribution is merely a first step in an attempt to approach them, raising several issues and settling none. However, results of tests run on an Encore Multimax with up to 16 processors show that even this first step has utility in the coarsegranularity parallelization of hydrocodes of practical importance.
 Publication:

Domain Decomposition Methods
 Pub Date:
 1989
 Bibcode:
 1989ddcm.proc..321K
 Keywords:

 Computational Fluid Dynamics;
 Iterative Solution;
 Partial Differential Equations;
 Asymmetry;
 Elliptic Differential Equations;
 Finite Difference Theory;
 Linear Equations;
 Matrices (Mathematics);
 Preconditioning;
 Fluid Mechanics and Heat Transfer