Community Research and Development Information Service - CORDIS

Abstract

To speed up calculations on a hydrodynamic model of sea circulation patterns (ISPRAMIX), a parallelization by domain decomposition was implemented on a Cray J916 shared memory parallel computer. Increasing the number of domains for a constant number of processors leads to an optimum number of domains with respect to the speed up factor of 4.3 was attained using 6 processors instead of 1 processor. By reducing the waiting time induced by synchronization using suitable measures, the wall clock time could be decreased further, increasing the speed up factor from 4.3 to 5.3.

Additional information

Authors: HIBMA N W, University of Twente, Department of Computer Science (NL);EIFLER W, JRC Ispra (IT)
Bibliographic Reference: Paper presented: Parallel CFD '96, Capri (IT), May 20-23, 1996
Availability: Available from (1) as Paper EN 39710 ORA
Record Number: 199610629 / Last updated on: 1996-05-27
Category: PUBLICATION
Original language: en
Available languages: en