ITLApplied  Computational Mathematics Division
ACMD Seminar Series
Attractive Image NIST
 
Up


Sparse iterative techniques for the solution of the 3D Maxwell equations in boundary element formulation

Luc Giraud
Parallel Algorithms Project, CERFACS, Toulouse, France

Friday, October 31, 2003 15:00-16:00,
Room 145, NIST North (820)
Gaithersburg
Friday, October 31, 2003 13:00-14:00,
Room 4550
Boulder

Abstract: CERFACS, CERMICS/INRIA and EADS/AIRBUS, we have been investigating the solution of large dense complex linear solvers arising from the discretization of the 3D Maxwell equations via boundary element methods. Nowadays for typical applications the size of those linear systems exceed one million so that direct solvers are no longer affordable even on very large parallel computers and the iterative solvers are the only alternative. In this talk we will describe the solution techniques we have been developing that are essentially based on sparse approximate inverse and multi-level preconditioners combined with Krylov solvers. On this range of problems, the matrix-vector product is performed using the fast multipole method that imposes some additional constrains on the construction of the preconditioners. We will illustrate the numerical and parallel behaviour of the solvers on test examples coming from both academic and industrial applications.
Contact: A. J. Kearsley

Note: Visitors from outside NIST must contact Robin Bickel; (301) 975-3668; at least 24 hours in advance.



Privacy Policy | Disclaimer | FOIA
NIST is an agency of the U.S. Commerce Department.
Last updated: 2011-01-12.
Contact