S-HARP: A parallel dynamic spectral partitioner (A short summary)

Document Type

Conference Proceeding

Publication Date

1-1-1998

Abstract

Computational science problems with adaptive meshes involve dynamic load balancing when implemented on parallel machines. This dynamic load balancing requires fast partitioning of computational meshes at rtm time. We present in this report a scalable parallel partitioner, called S-HARP. The underlying principles of S-HARP are the fast feature of inertial partitioning and the quality feature of spectral partitioning. S-HARP partitions a graph from scratch, requiring no partition information from previous iterations. Two types of parallelism have been exploited in S-HARP, fine-grain loop-level parallelism and coarse-grain recursive parallelism. The parallel partitioner has been implemented in Message Passing Interface on Cray T3E and IBM SP2 for portability. Experimental results indicate that S-HARP can partition a mesh of over 100,000 vertices into 256 partitions in 0.2 seconds on a 64-processor Cray T3E. S-HARP is much more scalable than other dynamic partitioners, giving over 15-fold speedup on 64 processors while ParaMeTiS 1.0 gives a few-fold speedup. Experimental results demonstrate that S-HARP is three to 10 times faster than the dynamic partitioners ParaMeTiS and Jostle on six computational meshes of size over 100,000 vertices.

Identifier

42549153054 (Scopus)

ISBN

[3540648097, 9783540648093]

Publication Title

Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics

External Full Text Location

https://doi.org/10.1007/bfb0018554

e-ISSN

16113349

ISSN

03029743

First Page

376

Last Page

385

Volume

1457 LNCS

This document is currently not available here.

Share

COinS