Parallel multistage preconditioners based on a hierarchical graph decomposition for SMP cluster architectures with a hybrid parallel programming model

Authors:
Kengo Nakajima
Affiliations:
Department of Earth and Planetary Science, The University of Tokyo, Tokyo, Japan
Venue:
HPCC'07 Proceedings of the Third international conference on High Performance Computing and Communications
Year:
2007

Citing 7
Cited 0

GPBi-CG: Generalized Product-type Methods Based on Bi-CG for Solving Nonsymmetric Linear Systems

SIAM Journal on Scientific Computing
Iterative Methods for Sparse Linear Systems

Iterative Methods for Sparse Linear Systems
Parallel Iterative Solvers of GeoFEM with Selective Blocking Preconditioning for Nonlinear Contact Problems on the Earth Simulator

Proceedings of the 2003 ACM/IEEE conference on Supercomputing
A Parallel Multistage ILU Factorization Based on a Hierarchical Graph Decomposition

SIAM Journal on Scientific Computing
Parallel Preconditioning Methods with Selective Fill-Ins and Selective Overlapping for Ill-Conditioned Problems in Finite-Element Methods

ICCS '07 Proceedings of the 7th international conference on Computational Science, Part III: ICCS 2007
Communication Bandwidth of Parallel Programming Models on Hybrid Architectures

ISHPC '02 Proceedings of the 4th International Symposium on High Performance Computing
The impact of parallel programming models on the performance of iterative linear solvers for finite element applications

VECPAR'06 Proceedings of the 7th international conference on High performance computing for computational science

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this work, the Parallel Hierarchical Interface Decomposition Algorithm (PHIDAL) and a hybrid parallel programming model were applied to finite-element based simulations of linear elasticity problems in media with heterogeneous material properties using parallel preconditioned iterative solvers. Reverse Cuthill-McKee reordering with cyclic multicoloring (CM-RCM) was applied for parallelism on each SMP node through OpenMP. The developed code has been tested on the IBM p5-575 and the TSUBAME Grid Cluster using up to 512 cores. Preconditioners based on PHIDAL provide a superior scalable performance and robustness on both architectures in comparison to conventional block Jacobi-type localized preconditioners.