Load-balanced parallel architectures for 2-D water quality model PARATUNA-WQ on OpenMP

Abstract

Because of the potential speedup, parallel algorithms have recently been developed for improving serial applications in ocean and coastal hydrodynamics and water quality simulations. Developing a parallel program, however, is a difficult task that requires special and expensive processing resources. Motivated by the potential benefits of parallelization, this paper develops a load-balanced parallel architecture on OpenMP to improve on an in-house serial two-dimensional water quality simulation model to a parallel application named PARATUNA-WQ. Analysis of the performance of speedup is discussed to justify the use of parallel architecture in water quality simulation model. Speedup achieved by PARATUNA-WQ is close to the maximum theoretical speedup predicted by the Amdahl Law. Further enhancement for application to very large computational domain consisting of 25 million computational nodes is possible by integrating MPI architecture into the framework of OpenMP, the result of which will be reported in a subsequent paper.