Multiprocessing ocean circulation: Modeling, implementation, and performance on the Intel Paragon

Abstract

In this paper we present the modeling and implementation of a grand challenge problem in the field of scientific computation: the primitive-equation numerical ocean circulation model. We present the mathematical formulation of the model and propose a scheme for its parallel implementation. Optimizations are made through collective communications and various partitioning schemes. In our experiments, which use up to 100 processors on the Intel Paragon parallel computer, the proposed strategy yields an encouraging speedup and exhibits a sustained scalability with increasing problem and machine sizes. We consider barotropic continental shelf waves in a periodic channel as a test problem. The model has numerous applications in environmental studies and ocean sciences.