The Research Of Three-dimensional FDTD Parallel Algorithm Based On MPI And OpenMP

As a result of a large number of scientific and engineering computing needs, more and more attention has been attracted by the calculation of electromagnetic fields. The analytical solution of the classical Maxwell equations only can be got in a simple and special model, so a large number of approximate numerical methods emerged. Restricting by the development level of computer, many numerical methods theoretically works, but can not be achieved in practice. Until the emergence of computer technology came, the numerical methods of electromagnetic field had been widely used. 1966, the prototype of FDTD method has been carry out by K. S. Yee, which has been continuously developing and improving in the next few decades. It has become widely used in the calculation of a variety of electromagnetic field as an effective method nowadays.With the continuous development of science and technology, problems in the research and application of electromagnetic fields become more and more complicated, and the volume of calculation proliferates. Restricted by the storage capacity and speed, a single computer has been far from to meet the requirements. Through the computing power of a single computer continues to be heightened with the hardware and software technology development, which is able to make some effect, the problems can not be solved fundamentally. As the parallel technology proposed, computers can be used to calculate a part of the problem at the same time, which means in principle the computing power of the entire system can be increased indefinitely by increasing the number of computers in the system. In this way, the bottlenecks of numerical calculation caused by storage size and time can be solved.This paper mainly introduces the parallel models and characteristics of MPI and OpenMP. Based on them, making parallel computing of the FDTD algorithm to come true, and verify the correctness and efficiency. Then discuss from the different ways of division (dimension and scale ), the different means of communication (standard blocking and standard non-blocking) and the different model of parallel (MPI and MPI + OpenMP), to find these factors on the efficiency of parallel programming. And then a number of ways has put forward to improve the speed-up rate. Finally, this article summarizes the research aspects in the future.