high performance computing on graphics processing units: hgpu.org

Because Cellular Automata (CA) is a dynamic system with inherent parallelism, many studies are focused on mapping CA to the parallel system in order to obtain high performance computing capability, such as using clusters, supercomputers and networks of computers. But the application of these systems are too expensive and difficult to use on the occasions which need convenient computing. Recent developments in the General Purpose GPU can meet the desktop computing challenge, which have high performance at low cost. This paper presents a general-purpose approach to accelerate the CA in geographical domain in case of spatial simulation. A series of experiments are launched to test the performance of proposed method. Finally, the experimental results indicate the approach in this paper can obtain high performance and computational performance data using GPU based accelerating method are fifty to sixty times faster than the identical algorithms using CPU in test environment. It is worthy of (1) reducing the communication cost between GPU and CPU is crucial when visualization and processing are equally important in real time simulation and (2) improving the parallel capability in the CA functions is essential using GPU Programming.