PUMA
Istituto di Matematica Applicata e Tecnologie Informatiche     
Fornasari N., Gazzaniga G., Rovida S., Sacchi G. A parallel implementation of the three-fields algorithm on a HP-Convex exemplar SPP1600. Preprint ercim.cnr.ian//1997-1047, 1997.
 
 
Abstract
(English)
This work deals with the parallel implementation of a non-overlapping Domain Decomposition method. We applied the so called Three-Fields algorithm to an advection/diffusion, strongly advection dominated, problem choosing the programming model suggested by the functional decomposition of the procedure. The considered algorithm was implemented on a HP-Convex Exemplar SPP1600, following the explicit message passing style and using PVM to perform interprocessor communication. We provide a careful analysis of execution times and parallel scalability, in order to highlight the performance of the algorithm on the considered parallel machine. In particular, we discuss communication and computation costs and investigate the influence on the achieved results of workload balancing and of architectural features such as interconnect network and cache memories. Finally, we also discuss the gains in performance provided by multilevel parallelism.
Subject 68-XX



Icona documento Open access Icona documento Restricted Icona documento Private

 


Per ulteriori informazioni, contattare: Librarian http://puma.isti.cnr.it

Valid HTML 4.0 Transitional