high performance computing on graphics processing units: hgpu.org

In this manuscript, the parallel computation is investigated including reviewing different programming APIs and architectures. Two specific parallel API-MPI and CUDA C are deeply analyzed. Two sorting algorithms and a visual mathematic problem are implemented with MPI alone with performance analysis. A stable fluid dynamics simulation has been experimented with CUDA. We also present a parallel way to implement a visualization application in the cluster with integration of multi-thread, CUDA and OpenGL. The performance analysis is given and future improvements are also pointed out.