作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

›› 2005, Vol. 31 ›› Issue (22): 91-93.

• 论文 • 上一篇    下一篇

万亿次机群系统高性能应用软件运行现状分析

侯晓吻,张林波,张云泉   

  1. 北京邮电大学计算机科学与工程学院,中科院数学与系统科学研究院,中科院软件所 北京 100876,北京 100080,中科院计算机科学国家重点实验室,北京 100080
  • 发布日期:2005-11-25

State-of-art Analysis of High Performance Application Software on Domestic Tera-scale Linux Cluster System

HOU Xiaowen1,ZHANG Linbo2,ZHANG Yunquan3,4   

  1. (1.School of Computer Science and Engineering,Beijing University of Posts and Telecommunications,Beijing 100876;2.Institute of Computational Mathematics and Scientific/Engineering Computing,Chinese Academy of Sciences,Beijing 100080;3.Institute of Software,Chinese Academy of Sciences,Beijing 100080;4.State Key Lab of Computer Science,Chinese Academy of Sciences,Beijing 100080)
  • Published:2005-11-25

摘要: 通过调用PAPI(PerformanceApplicationProgrammingInterface)接口函数[6]对2004年3月~4月之间运行在国家应用“973”计划项目LSSC-II万亿次机群系统上部分应用程序进行了跟踪,收集到了大量宝贵的性能数据。依据这些性能数据信息,对我国当前高性能软件的运行情况给出了初步分析。分析结果表明,目前大部分应用程序性能都处于较低水平,并行程序使用处理器的数目范围一般为1~64个,处理器平均效率低于10%,平均性能低于300Mflops。

关键词: 高性能计算, 高性能应用软件, 性能分析

Abstract: All the programs running on the first national “973” LSSC-II tera-scale Linux cluster system from March to April in 2004 through calling PAPI(performance application programming interface)functions have been tailed after,and lots of important performance data are gathered.This paper presents a primary analysis to high performance application in nowadays in China.The analysis shows that the real performance level of many high performance software is still very low.The amount of CPU used by parallel programs is between 1 and 64.The average efficiency of application is lower than 10%.The average performance of CPU is lower than 300Mflops.

中图分类号: