Parallel Processing Technology In The Fdtd Algorithm

A parallel FDTD method is the important method to enhance the speed and reduce the requirement of single computer memory costs in multiple data FDTD operation. Parallel FDTD method proceeded in a cluster of PCs is a good solution to get the low-cost and high efficiency. The work done in this thesis is based on the application of parallel FDTD in computational electromagnetics.This thesis firstly introduces the basic theory of parallel algorithms, and then introduces the system of PC clusterand the mechanism of MPI(Message Pass Interface). According to this theory, a simple cluster of PCs based on Solaris/Windows and MPIwas established which satisfies the laboratory environment. There are great differences between the newly devised MPI programming mode and the traditional one. Secondly, this thesis discusses the method of how to proceed the MPI parallel programming in cluster environment and how to implement the two dimensional MPI parallel FDTD programs. The time differences and memory cost between parallel FDTD and traditional one are also compared. Finally, from the theoretical and practical study, the feasibility of proceeding parallel FDTD in cluster of PCs are discussed and some meaningful conclusions are also presented.