Based On The Computational Speed Of The Gpu-accelerated Fdtd Simulation

The Finite Different Time Domain Method has been widely used in the field of the calculation electromagnetic. As the FDTD method has the nature characteristic of parallelism, so its application has become more and more universal. But the calculation that used the CPUs in the FDTD has met some problems. The CPUs have no longer meet the requirements of the memory space and the speed of the FDTD calculation. The emerge of the Graphic Processing Unit (GPU) solves the problems. GPU is a hardware that specially designed for the parallel computing.Because of the good of the FDTD in parallelism,just right for the GPUs.Aimed at the lack of the traditional FDTD algorithm, in this paper, we apply the GPU computing platform of the Tesla c1060.Through the theoretical analysis and numerical simulation I studied and implemented FDTD algorithm for the one-dimension and two-dimensions using the Tesla c1060.We can conclude that using the GPU to calculate the FDTD is right.In addition,the GPU really play the role of accelerating the FDTD calculation speed.At the same time,I compare the computing time of the GPU and the CPU under the different numbers of the grids.The results show that the more dates we have to simulate,the more advantage GPU over the CPU.