Download:
|
by W. Joubert, R. Janardhan, W. Dearholt
http://www.c3.lanl.gov/~wdj/papers/3dic.ps.Z
Add To MetaCart
Abstract:
Abstract. This paper presents a fully parallel implementation of the global lexicographically ordered M/ILU preconditioning for linear systems arising from structured 3-D problems. M/ILU preconditionings, which are popular on conventional serial architectures, have not been easily amenable to parallelization due to their sequentiality, leading to the need to reorder the gridpoints or precondition on subdomains, commonly resulting in degraded convergence. This paper presents a technique for circumventing this problem for a class of problems of practical importance, namely, 3-D structured problems. For numbers of processors (100-1000) and problem sizes (millions of grid cells) which are characteristic of large scale computations, the parallelization strategy results in parallel efficiencies of up to 70-80 percent or more. The details of the implementation are given in this paper, with numerical results.
Citations
|
410
|
E.: Methods of conjugate gradients for solving linear systems
– Hestenes, Stiefel
- 1952
|
|
60
|
A class of first order factorization methods
– Gustafsson
- 1978
|
|
36
|
Solution of the FirstOrder Form of the 3-D Discrete Ordinates Equation on a Massively Parallel
– Koch, Baker, et al.
- 1992
|
|
13
|
Ewing Lusk and Anthony Skjellum, “Using MPI- Portable Parallel Programming with the Message Passing Interface
– Gropp
- 1994
|
|
12
|
The Cray T3D Address Space and How to Use It
– Numrich
- 1994
|
|
11
|
Modified incomplete Cholesky (MIC) methods
– Gustafsson
- 1983
|
|
8
|
Stability and rate of convergence of modified incomplete Cholesky factorization methods
– Gustafsson
- 1979
|
|
7
|
der Vorst, "An iterative solution method for linear systems of which the coefficient matrix is a symmetric M-matrix
– Meijerink, van
- 1977
|
|
5
|
Efficient Implementation of a Class of Conjugate Gradient Methods
– Eisenstat
- 1981
|
|
4
|
Improved SSOR and incomplete Cholesky solution of linear equations on shared and distributed memory parallel computers
– Joubert, Oppe
- 1994
|
|
2
|
On Vectorizing Incomplete Factorization Preconditioners
– Ashcraft, Grimes
- 1988
|
|
2
|
Meurant, "The Effect of Ordering on Preconditioned Conjugate Gradient
– Duff, A
- 1989
|
|
2
|
Olaf Lubeck and Bart van Bloemen Waanders, "Falcon: A Production Quality Distributed Memory Reservoir Simulator
– Shiralkar, Stephenson, et al.
- 1997
|
|
1
|
Performance Fortran Language Specification. Unpublished report
– High
- 1993
|
|
1
|
Fortran at 10 Gigaflops : The Connection Machine Convolution Compiler
– Bromley, Heller, et al.
- 1991
|
|
1
|
van Bloemen Waanders, Robert Stephenson, Gautam Shiralkar, "Next Generation Oil Reservoir Simulations
– Joubert, Koch, et al.
|
|
1
|
The High Performance Fortran Handbook
– Zozel
- 1994
|
