Partial Differential Equations (PDEs) are vastly used in scientific computations and engineering applications. Since these equations cannot directly be solved, numerical methods like FDTD are used for solving them rather than analytical ones. Great efforts have been put into fast solving of these equations which have huge computations. Recently accelerating these computations using clusters of computers have been adopted. More recently, GPUs which have parallel architectures has been used for solving general purpose computations. In this paper we first review FTDT method and then suggest a method for solving it on a cluster of computers and evaluate its performance on our hardware. Also, we propose a new method which reduces the amount of memory needed for implementing FDTD. Improving the speed of FTDT algorithm, we implement it using our new method on a cluster of GPUs taking advantage of CUDA technology. By so doing, we achieved the speedup of up to forty factors.
展开▼
