Integrating high performance computing strategies with existing finite element hydrodynamic codes

Using serial finite-element codes for large scale modeling of estuarine flows is limited by the computational environment. As the modeling domain becomes more complex and larger, it is imperative to move the application platform from serial to parallel architectures. It is our belief that all future scientific computing in general will be on parallel computers. Gaining synergy from the recent advances made in the field of high performance computing, we demonstrate the advantages of running three different serial finite-element codes in a parallel environment. The serial codes, including RMA2, have been ported over to parallel environment using Domain Decomposition principles based on MPI. The tests, up to 32 processors, indicate that these parallelized serial codes are efficient, scalable and portable.