The parallelization of the Princeton Ocean Model

Abstract

In this paper we present the parallel implementation of the Princeton Ocean Model (POM) using message passing. Domain decomposition techniques are used for the horizontal discretization whereas in the vertical direction each column per grid point is computed. The required interprocessor communication was implemented on the PVM message passing library. Three different data partitioning schemes are studied. It is found that the checkerboard partitioning produces the best results when the number of processors becomes larger than 15, otherwise row block striped partitioning is to be preferred. © Springer-Verlag Berlin Heidelberg 1999.