Efficient Parallelization of the Preconditioned Conjugate Gradient Method

被引:0
作者
Accary, Gilbert [1 ]
Bessonov, Oleg [2 ]
Fougere, Dominique [3 ]
Gavrilov, Konstantin [4 ]
Meradji, Sofiane [3 ]
Morvan, Dominique [5 ]
机构
[1] Univ St Esprit Kaslik, BP 446 Jounieh, Beirut, Lebanon
[2] Russian Acad Sci, Inst Problems Mech, Moscow 119526, Russia
[3] Modelisat Mecan Lab, F-13451 Marseille, France
[4] Perm State Univ, Perm 614990, Russia
[5] Univ Aix Marseille 2, UNIMECA, F-13453 Marseille, France
来源
PARALLEL COMPUTING TECHNOLOGIES, PROCEEDINGS | 2009年 / 5698卷
关键词
LINEAR-SYSTEMS;
D O I
暂无
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
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 (MILD) 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).
引用
收藏
页码:60 / +
页数:2
相关论文
共 15 条