Study On Parallel Performance OF FDTD Under Environment Based On Linux In PC Cluster System

被引：1

作者：

Ye, Xiaoli ^{[1
]}

Lai, Qinwu ^{[1
]}

Chen, Rongling ^{[1
]}

Xu, Zao ^{[2
]}

Ye, Xiangqun ^{[3
]}

Luo, Dongxiang ^{[3
]}

机构：

[1] Nanchang Univ, Dept Elect, Nanchang, Peoples R China

[2] Nanjing Univ Aeronaut & Astronaut, Nanjing, Jiangsu, Peoples R China

[3] Nanchang Univ, Sch Foreign Language Studies, Nanchang, Jiangxi, Peoples R China

来源：

2009 INTERNATIONAL SYMPOSIUM ON INTELLIGENT UBIQUITOUS COMPUTING AND EDUCATION | 2009年

关键词：

MPI; finite difference time domain(FDTD); COW; region division; parallel computing;

D O I：

10.1109/IUCE.2009.80

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This paper describes the parallelizing algorithm of finite difference time domain(FDTD), which is realized by a method based on Message Passing on the parallel computing system of COW constituted by high-performance PC through network. Every subregion deliverys the data of the field values to its contiguous subregion on the boundary by dividing the computation region so as to realize the parallelizing of FDTD algorithm. In this paper, we utilize the parallel FDTD to study the scattering of dielectric layer of electromagnetic wave. It shows that the results of the parallel computing are consistent with those of the serial computing. Meanwhile, we also analyze & discuss the computing efficiency of FDTD and other factors that can have impacts on the parallel efficiency. Finally, we present an improved algorithm to increase the parallel efficiency.

引用

页码：514 / +

页数：2

共 4 条

[1] MRTD analysis of a transient electromagnetic pulse propagating through a dielectric layer [J].