Toward GPU on CUDA for Simulating Non-hydrostatic Wave Model

The goal of this paper is to calculate the performance of the parallel scheme of non-hydrostatic Gravitational Surface Wave using GPU based on CUDA. Here, Navier-Stokes equations used as a model which is simplified by removing non-linear and friction terms. In this paper, simple finite different method is used as a numerical method. The discretization of pressure, vertical and horizontal velocity variables uses Arakawa C-Grid which leads to solve Poisson equation. The numerical simulation of propagating waves generated by surface pressure is given. The parallel computing is shown satisfied. Using large number of grids (N
_x
=1024,N
_y
=512), serial and parallel CPU time are observed 20197.66 and 6521.85 seconds respectively. Therefore the usage of parallel computing in this simulation produces speedup approximately 3 times of serial computing.