Parallel Processing White Papers
A Parallel Algebraic Multigrid Preconditioner Using Algebraic Multicolor Ordering for Magnetic Finite Element Analyses
Overview Recently, Finite Element (FE) analyses of magnetic fields have played a major role in the design of various electromagnetic machines. In the analyses, most computation time is consumed by the solution of large-scale sparse linear systems of equations derived from FE formulations. Algebraic Multigrid (AMG) techniques are known to be good preconditioners that efficiently accelerate the convergence of basic iterative methods for linear systems of equations with sparse matrices. This paper studies parallel processing of the AMG preconditioner for FE analyses with coloring strategies. In parallelization of the AMG preconditioner, it is important to devise parallelization of the smoother. Since AMG techniques have been developed as black-box multigrid solvers, it is desirable that parallel processing does not destroy the black-box property of the AMG techniques.
| Publisher | John von Neumann Institute for Computing | File Format | |
|---|---|---|---|
| Date Published | December 2006 | Downloads | 1 |
| Format | White Papers | ||
| Topics | |||
-
Did you find this white paper useful? 3 out of 8 users found this white paper useful
White Paper RSS
Download
Register
