引用本文: | 赵军,宋君强,张理论,等.全球气象资料客观分析系统的分布式并行计算[J].国防科技大学学报,2000,22(3):39-43.[点击复制] |
ZHAO Jun,SONG Junqiang,ZHANG Lilun,et al.Distributed Parallel Computation of the Global Atmospheric Data Objective Analysis System[J].Journal of National University of Defense Technology,2000,22(3):39-43[点击复制] |
|
|
|
本文已被:浏览 6677次 下载 5349次 |
全球气象资料客观分析系统的分布式并行计算 |
赵军, 宋君强, 张理论, 朱小谦 |
(国防科技大学 计算机学院, 湖南 长沙 410073)
|
摘要: |
对已有的串行算法进行并行化, 是一项很困难的工作。通过对全球气象资料客观分析系统串行算法的研究, 提出了一种静态分配数据的分布式并行算法。该算法通过间隔选取分析盒子和模式格点纬圈行, 将数据分配给不同的处理机实现分布式并行。该并行算法负载平衡好, 并行效率高, 而且并行化代价较低, 具有良好的可扩展性。 |
关键词: 分布式并行 分析盒子 数据分配 客观分析 |
DOI: |
投稿日期:1999-12-07 |
基金项目:国家863计划资助项目(863-306-ZD11-03-8) |
|
Distributed Parallel Computation of the Global Atmospheric Data Objective Analysis System |
ZHAO Jun, SONG Junqiang, ZHANG Lilun, ZHU Xiaoqian |
(College of Computer, National Univ. of Defense Technology , Changsha 410073, China)
|
Abstract: |
It is a difficult task to parallelize the existing sequential algorithm. A distributed parallelized algorithm is described by means of static data distribution after researching into the sequential algorithm of the global atmospheric data objective analysis system. We have achieved the parallelization by extracting the analysis boxes and model grid point latitude rows with leaped step and distributing data to different processors. The parallelization algorithm has a good load balance, a highly parallelization efficiency and a lower parallelization cost. Good scalability can be got from the algorithm. |
Keywords: distributed parallelization analysis box data distribution objective analysis |
|
|
|
|
|