Parallel 3D Finite Difference Time Domain Simulations on Graphics Processors with Cuda

Parallel Finite Difference Time Domain (FDTD) method has been explored over past few years because of the expensive computation needed for its application. And General Purpose Graphics Processing Units (GPGPU), especially Computer Unit Device Architecture (CUDA) model, has been offered an efficient and simple solution. This paper analyzes parallel FDTD method and CUDA architecture, presents a GPU based implementation of three-dimensional FDTD which is solved by two-dimensional grid of threads and extra shared memory is used in our application for optimal memory accessing. With a GT200 GPU as coprocessor, tens of times speedup is obtained compared with traditional PC computation.