The Great Lakes of North America have ice covers during winter that change the pattern of the wind-induced current in the lakes. Offshore and coastal engineers are interested in these changes, because the current may impose significant loads on the offshore manmade structures in the lakes. The current also influence the elevation of the water level. Physical oceanography researchers have developed numerical models to simulate this phenomenon. These models typically require heavy computations and faster computation methods. High Performance Fortran (HPF) is one of the powerful tools to execute such large programs faster, by allowing users to simultaneously use multiple CPUs. Converting legacy FORTRAN applications to HPF is an attractive path to parallelism.
The first part of this thesis describes the physical oceanography problem and the numerical model. Computational solution of this model requires massive computations that lead to the need for HPF. The second part of the thesis summarizes features of HPF and the parallel computing hardware environment, the IBM SP2. Then we detail the conversion of the FORTRAN 77 program to an HPF program and make a comparison of the run times and efficiency of the parallelism.