high performance computing on graphics processing units: hgpu.org

This paper proposes and describes a methodology developed to port complex scientific applications originally written in FORTRAN to nVidia CUDA. The significance of this lies in the fact that, despite the performance improvement and programmer-friendliness provided by CUDA, it presently lacks support for FORTRAN. The methodology described in this paper addresses this problem using a multiple step process that includes identification of software modules that benefit from being ported, familiarization with the code, porting, optimizing, and verifying the ported code. It was developed and carried out by porting an existing module of a weather forecasting application written in FORTRAN. Using this approach, we obtained a functional prototype of the ported module in approximately 3 months, despite our lack of knowledge of the theory of the weather code. Considering the relevance of this application to other scientific applications also written in FORTRAN, we believe that the proposed porting methodology described can be successfully utilized in several other existing scientific applications.