首页 | 本学科首页   官方微博 | 高级检索  
     

基于矩阵分解的周期块三对角线性方程组的并行直接解法
引用本文:樊艳红,吕全义,李纪华,宋东红.基于矩阵分解的周期块三对角线性方程组的并行直接解法[J].纺织高校基础科学学报,2008,21(4):483-486.
作者姓名:樊艳红  吕全义  李纪华  宋东红
作者单位:1. 西北工业大学,应用数学系,陕西,西安,710072
2. 西北工业大学,经济研究中心,陕西,西安,710072
基金项目:陕西省自然科学基金资助项目  
摘    要:提出了分布式环境下求解周期块三对角线性方程组的一种并行算法.该算法充分利用系数矩阵结构的特殊性,通过对系数矩阵进行适当分解及近似处理,使算法只在相邻处理机间通信2次,并从理论上给出了算法有效的一个充分条件.最后,在HP rx2600集群上进行了数值试验,结果表明,实算与理论是一致的,并行性也很好.

关 键 词:周期块三对角线性方程组  矩阵分解  并行算法  并行效率  HP  rx2600集群

A parallel direct algorithm based on matrix decomposition for solving periodical block-tridiagonal linear equations
FAN Yan-hong,LU Quan-yi,LI Ji-hua,SONG Dong-hong.A parallel direct algorithm based on matrix decomposition for solving periodical block-tridiagonal linear equations[J].Basic Sciences Journal of Textile Universities,2008,21(4):483-486.
Authors:FAN Yan-hong  LU Quan-yi  LI Ji-hua  SONG Dong-hong
Affiliation:FAN Yan-hong1,LU Quan-yi1,LI Ji-hua2,SONG Dong-hong2
Abstract:A parallel algorithm for periodical block-tridiagonal linear equations on distributed-memory multi-computers is presented. Making full use of the special structure of the coefficient matrix, the algorithm is based on decomposing the coefficient matrix properly and approximately disposing the matrix. The communication only needs twice between the adjacent processors. Theoretically, a sufficient condition about effectivity of this algorithm is given. Finally, some numerical results on HP rx2600 cluster demonstrate that practice computing is consistent with theory. The algorithm's parallelism is preferable.
Keywords:periodical block-tridiagonal linear equations  factorization of matrix  parallel algorithm  parallel efficiency  HP rx2600 cluster
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号