Efficient Parallelization of the Preconditioned Conjugate Gradient Method

Authors:
Gilbert Accary;Oleg Bessonov;Dominique Fougère;Konstantin Gavrilov;Sofiane Meradji;Dominique Morvan
Affiliations:
Université Saint-Esprit de Kaslik, Jounieh, Lebanon;Institute for Problems in Mechanics of Russian Academy of Sciences, Moscow, Russia 119526;Laboratoire de Modélisation, Mécanique et Procédés Propres, L3M---IMT, La Jetée, Technopôle de Château-Gombert, Marseille Cedex 20, France 13451;Perm State University, Russia 614990;Laboratoire de Modélisation, Mécanique et Procédés Propres, L3M---IMT, La Jetée, Technopôle de Château-Gombert, Marseille Cedex 20, France 13451;Université de la Méditerranée, UNIMECA, Marseille Cedex 13, France 13453
Venue:
PaCT '09 Proceedings of the 10th International Conference on Parallel Computing Technologies
Year:
2009

Citing 7
Cited 1

Introduction to Parallel & Vector Solution of Linear Systems

Introduction to Parallel & Vector Solution of Linear Systems
Parallelization of robust multigrid methods: ILU factorization and frequency decomposition method

SIAM Journal on Scientific and Statistical Computing
Parallelism in ILU-preconditioned GMRES

Parallel Computing
OpenMP: An Industry-Standard API for Shared-Memory Programming

IEEE Computational Science & Engineering
Preconditioning techniques for large linear systems: a survey

Journal of Computational Physics
Iterative Methods for Sparse Linear Systems

Iterative Methods for Sparse Linear Systems
Optimized parallel approach for 3D modelling of forest fire behaviour

PaCT'07 Proceedings of the 9th international conference on Parallel Computing Technologies

OpenMP parallelization of a CFD code for multicore computers: analysis and comparison

PaCT'11 Proceedings of the 11th international conference on Parallel computing technologies

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper we present methods for efficient parallelization of the solution of pressure Poisson equation arising in 3D CFD forest fire modeling. The solution procedure employs the Conjugate Gradient method with implicit Modified ILU (MILU) preconditioner. The basic idea for parallelizing recursive incomplete-decomposition algorithms is to use a direct nested twisted approach in combination with a staircase method. Parallelization of MILU-CG solver is implemented in OpenMP environment for Non-uniform memory (NuMA) computer systems. Performance results of the parallelized algorithm are presented and analyzed for different number of processors (up to 16).