Search
loader loading subjects...
cover
Evaluation of Parallel Sparse Matrix Partitioning Software for Parallel Multilevel ILU Preconditioning on Shared-Memory Multiprocessors
Authors
José I. Aliaga, Matthias Bollhöfer, Alberto F. Martín, Enrique S. Quintana-Ortí
Pages
125 - 132
DOI
10.3233/978-1-60750-530-3-125
Series
Advances in Parallel Computing
Ebook
Volume 19: Parallel Computing: From Multicores and GPU's to Petascale
Abstract

In this paper we analyze the performance of two parallel sparse matrix partitioning software packages, ParMETIS and PT-SCOTCH. We focus our analysis on the parallel performance of the nested dissection partitioning stage as well as its impact on the performance of the numerical stages of the solution of sparse linear systems via multilevel ILU preconditioned iterative methods. Experimental results on a shared-memory platform with 16 processors show that ParMETIS seems to be the best choice for our approach.