• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2010, Vol. 32 ›› Issue (12): 98-101.doi: 10.3969/j.issn.1007130X.2010.

• 论文 • 上一篇    下一篇

扩展双精度浮点并行计算:MPI方法

杨灿群,杨学军,易会战   

  1. (国防科学技术大学计算机学院,湖南 长沙 410073)
  • 收稿日期:2008-12-15 修回日期:2009-03-06 出版日期:2010-12-25 发布日期:2010-12-25
  • 通讯作者: 杨灿群
  • 作者简介:杨灿群(1968),男,湖南桃江人,副研究员,研究方向为编译技术;杨学军,教授,博士生导师,研究方向为并行体系结构、操作系统与编译;易会战,博士,研究方向为编译技术。
  • 基金资助:

    国家自然科学基金重点资助项目(60633050)

DoubleExtended Precision FloatingPoint Parallel Computing:MPI Implementation

YANG Canqun,YANG Xuejun,YI Huizhan   

  1. (School of Computer Science,National University of Defense Technology,Changsha 410073,China)
  • Received:2008-12-15 Revised:2009-03-06 Online:2010-12-25 Published:2010-12-25

摘要:

双精度浮点并行计算将不能满足高性能计算领域对计算精度的要求,但是目前还没有高性能的超双精度并行计算的解决方法。基于并行编程语言MPI,本文提出了扩展双精度浮点的并行计算实现方法,并且使用精度敏感的圆周率计算BBP算法验证了该方法的正确性和性能。

关键词: 扩展双精度, MPI, BBP算法

Abstract:

Parallel computing in doubleprecision floatingpoint arithmetic will be out of accord with the requirements of the computing precision in high performance computing in the future. However,there is no solution to parallel computing in super doubleprecision arithmetic with high performance. Based on the wildly used parallel programming language MPI,this paper presents a solution of doubleextended precision parallel computing. The correctness and performance of this solution is validated by a precisionsensitive BBP algorithm which is used for the computation of Pi.

Key words: doubleextended precision;MPI;BBP algorithm