计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (5): 189-192.DOI: 10.3778/j.issn.1002-8331.2009.05.056

• 工程与应用 • 上一篇    下一篇

POP海洋模式在四核至强集群上的并行计算

张理论,赵 军,吴建平,宋君强   

  1. 国防科学技术大学 计算机学院,长沙 410073
  • 收稿日期:2008-06-18 修回日期:2008-09-10 出版日期:2009-02-11 发布日期:2009-02-11
  • 通讯作者: 张理论

Parallel computing of POP ocean model on quad-core intel xeon cluster

ZHANG Li-lun,ZHAO Jun,WU Jian-ping,SONG Jun-qiang   

  1. School of Computer Science,National University of Defense Technology,Changsha 410073,China
  • Received:2008-06-18 Revised:2008-09-10 Online:2009-02-11 Published:2009-02-11
  • Contact: ZHANG Li-lun

摘要: 分析了POP海洋模式原理、离散方法。在四核至强集群上,研究分析POP模式中计算局部块技术和平衡并行数据剖分及其对模式性能的影响。针对模式的通信性能瓶颈,采用聚合通信优化技术。研究结果表明局部块技术和数据剖分方式对于POP模式并行性能影响显著;通过通信聚合优化,POP模式在四核集群上性能获得明显提升。

Abstract: Based on analysis of the equations and numerical discretization of POP ocean model,discuss the local blocking technique and load-balance data distribution,focusing on their influence upon parallel performance on quad-core cluster.Aggregation optimization is adopted for the communication bottleneck.Research and experiment show that local blocking technique and way of data distribution have remarkable effect on performance,and the aggregation optimization is effective on quad-core cluster,especially for the case of load-balance data distribution.