This paper uses a coastal ocean model simulation as motivation for a parallel processing study. The goal was to study the parallelization of this model, not to improve or enhance the model. Because of this, boundary conditions (like surface wind) were not imposed.

There are several statements made with no support. For example, when comparing two methods, it is stated that one method “captures the flow features...much better” than the other method. There is no description of how this was determined. Also, when comparing the parallel code to the serial code, the serial code partitioned the grid into blocks (just as the parallel code does). Why was this partitioning done? Were there memory restrictions? If not, then the comparisons are not as valid.

The paper devotes a section to discussing a technique for speeding up processing by reducing what the authors call “idle overhead.” Although this is a reasonable metric to consider, this is not the goal. The goal is to reduce the time to get the answer. This study did present a case where there was a reduction of “idle overhead,” but the time to get an answer was longer. There were no discussions of scalability or processor efficiency.

The paper’s most significant contribution is that is shows the successful blending of the message passing interface (MPI) and OpenMP models.