计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (1): 35-42.

• 理论研究、研发设计 • 上一篇    下一篇

基于CFIO的参数自动选择研究

孙新华1,2,张  程1,2,杨广文1,2   

  1. 1.清华大学 计算机科学与技术系,北京 100084
    2.清华大学 地球系统科学研究中心 地球系统数值模拟教育部重点实验室,北京 100084
  • 出版日期:2015-01-01 发布日期:2015-01-06

Research on automatic parameter selection for CFIO

SUN Xinhua1,2, ZHANG Cheng1,2, YANG Guangwen1,2   

  1. 1.Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China
    2.Ministry of Education Key Laboratory for Earth System Modeling, Center for Earth System Science, Tsinghua University, Beijing 100084, China
  • Online:2015-01-01 Published:2015-01-06

摘要: 为了使不同背景的使用者都可以在地球系统模式中发挥并行I/O库CFIO(Climate Fast I/O)库的最大效用,基于CFIO 设计并实现了一套参数自动选择的解决方案。该方案基于对目标模式程序的预运行收集到的I/O Tace信息,分为带宽测试子程序、Trace分析子程序和参数选择子程序,用户仅需要在命令行中运行这三个子程序即可自动地获得最优的CFIO参数。实验结果表明,依照该解决方案提供的参数进行设置可以有效地将模式的I/O过程隐藏在计算过程中,同时又避免了计算资源的浪费。

关键词: 地球系统模式, I/O Trace信息, 参数自动选择

Abstract: In order to meet the demands of the users with different backgrounds to get the best performance of CFIO(Climate Fast I/O) Library, the paper designs and implements a solution of automatic parameter selection, which is based on CFIO. The solution based on the I/O Trace collected by the pre-excution contains the bandwidth testing routine, the trace analysis routine and the parameter selection routine. The user will get the best CFIO parameters after running these routines in command line terminal. The results of the experiments show that under the parameters these routines provided, the I/O phase can be perfectly overlapped by the computing phase with the least computing resource.

Key words: Climate Models, I/O Trace, automatic parameter selection