GPU parallelization for accelerating 3D primitive equations of ocean modeling

Abstract

Graphics processing unit (GPU) has become a powerful computation platform not only for graphic rendering purposes, but also for multi-purpose computations. Using various software, such as NVIDIA’s Compute Unified Device Architecture (CUDA) programming model, the developers can use the GPU without a graphics programming background. In this paper, we describe the implementation of 3D primitive equations solver for incompressible and inviscid fluid flow in rotating frame with hydrostatic balance using desktop platform equipped with a GPU. The governing equations for this study consist of six dependent variables, three velocity components, temperature, salinity, and pressure. The finite difference method (FDM) is used to discretize the mathematical model based on forward-time backward-space (FTBS) scheme. It is realized that using a single Tesla K20c GPU card, the CUDA implementation of the ocean circulation model within two days simulation runs 216 times faster than a serial C++ code running on a single core of an Intel(R) Xeon(R) CPU E5-2620 2.10 GHz processor. The results reveal that the ocean circulation is feasible on this type of platform and that model can be run within minutes.