high performance computing on graphics processing units: hgpu.org

In this paper, it was presented numerical results related to three dimensional simulation of motion of a vortex ring. For the simulation it was chosen the Vortex In Cell method. The method was shortly described in the paper. The numerical results were obtained on the single processor (x86) architecture. The disadvantage of the single processor computation is a very long time of computation. To menage this problem, we switched to the parallel architecture. In our first approach to the multicore architecture we tested the possibility and algorithms for the solution of the algebraic system of equations that resulted form discretization of the Poisson equation. We presented the results obtained with CUDA architecture. In order to better understand how does the parallel algorithms work on CUDA architecture, it was shortly presented a scheme of the device and how programs are executed on it. We showed also our results which are related to the parallelization of some simple iterative methods like the Jacobi method and Red-Black Gauss-Seidel method for solution of the algebraic system. The results were encouraging. For the Red Black Gauss-Seidel using GTX480 card, the calculations were 90-times shorter than on a single processor. As we know the solution to the Poisson equation is equivalent to the solution to the algebraic systems.