期刊文献+

集群下Cholesky分解的核外预取算法

Out-of-core prefetching algorithm of Cholesky decomposition on clusters
下载PDF
导出
摘要 核外计算中,由于I/O操作速度比较慢,所以对文件的访问时间占的比例较大。如果使文件操作和计算重叠则可以大幅度地提高运行效率。软件数据预取是一种有效的隐藏存储延迟的技术,通过预取使数据在实际使用之前从硬盘读到缓存中,提高了缓存(cache)的命中率,降低了读取数据的时间。通过设置两个缓冲区来轮流存放本次和下一次读入的数据块,实现访存完全命中cache的效果,使Cholesky分解并行程序执行核外计算的效率得到了大幅度的提高。同时,I/O操作的时间与CPU的执行时间的比例也是影响效率的主要因素。 The time of accessing files is larger proportion because the I/O operation is slow in out-of-core computation. If the file operation and the calculation is overlap, the operational efficiency can be improved greatly. Data prefetching of software is an effective technique to hide memory latency, because it can improve the cache hit rate and reduce the time of reading data that make the data read from the hard drive to the cache before using the date actually.We set the two buffers to store the time and the next block of reading data in turn, to achieve the effect of accessing memory hitting cache completely and the execution effi- ciency of the parallel program of Cholesky decomposition has been improved greatly in out-of-core computation. At the same time, the proportion between I/O operation time and CPU execution time is also an important factors effecting efficiency.
作者 刘凤 刘青昆
出处 《微型机与应用》 2011年第4期14-17,共4页 Microcomputer & Its Applications
关键词 预取 核外 并行 集群 CHOLESKY分解 prefetching out-of-core parallel clusters Cholesky decomposition
  • 相关文献

参考文献7

二级参考文献20

  • 1迟学斌.Transputer上Cholesky分解的并行实现[J].计算数学,1993,15(3):289-294. 被引量:4
  • 2周树荃 邓绍忠 等.变带宽大型稀疏线性方程组的并行直接解法及其在YH-1的实现.航空科学基金论文集[M].北京:航空工业出版社,1993.216-220.
  • 3[1]Dongarra J et al. A set of level 3 basic linear algebra s ubprograms. ACM Trans on Mathematical Software, 1990, 16(1): 1~17
  • 4[2]Anderson E et al. LAPACK Users' Guide. 2nd ed. Philadelphia: Socie ty for Industrial and Applied Mathematics, 1995
  • 5[3]Gustavson F. Recursion leads to automatic variab1e blocking for dense linear a1gebra. IBM Journal of Research and Development, 1997, 41(6): 737~755
  • 6[4]Demmel J W. App1ied Numerical Linear Algebra. Philadelphia: Society fo r Industrial and Applied Mathematics, 1997
  • 7[5]Metcalf S, Reid J. FORTRAN90/95 Explained. Oxford: Oxford University P ress, 1996
  • 8[6]Engineering and Scientific Subroutine Library, Guide and Ne w York: IBM, 1994
  • 9Manish M,Shivakant M.Enhanced Server FaultTolerance for Improved User Experience.International Conference on Dependable Systems & Networks (DSN'08),2008,167-176.
  • 10http://server.ctocio.com.cn/news/17517684175.shtrrd,20 08.

共引文献18

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部