| With the development of large scale parallel computing technology in recent years, especially in the multi-core technique, the multi-thread computing, which is based on share memory architecture, has played an increasingly important role in parallel computing area, and it is of great practical significance to use this technique for improve computing efficiency of software. If this technique is applied to CHIPIC, which is developed by our own country and of intellectual property, the computational efficiency would be enhanced greatly under existing computer environment to finally achieve practicality.There are two standard in parallel computing area: one is OpenMP, which is an application program interface of share memory architecture; the other is MPI, which is a standard with the message passing interface. In this paper, the execution performance of CHIPIC is studied when using OpenMP, and combining MPI and OpenMP, the effect of the mixed program on the CHIPIC is also discussed.The CHIPIC is based on Finite-Difference Time-Domain (FDTD) algorithm and Particle-In-Cell (PIC) algorithm. In the paper, some basic theories of particular simulation and OpenMP standard are shown in the first. Then, for a simple two-dimensional waveguide model, some important factors which may affect the efficiency of electromagnetic FDTD parallel program are investigated. Put this result into the CHIPIC, it can be able to reach a higher computing performance of taking full advantage of OpenMP statement. At last, combining MPI and OpenMP standard, the mixed parallel computation is designed in the three-dimensional of CHIPIC, which can be used in the cluster networks with multi-thread and multi-computer.The test which compared serial and parallel computing of CHIPIC shows that, it can get a great performance in parallel computing with using shared memory standard, and the mixed parallel mode is the best effective way of getting a best performance. |
PDF Full Text Request