稀疏线性方程组并行解法的优化 - 昆明理工大学学报.pdf

33 2 ( ) V ol 33 N o2 20084 Journal ofK unm ing U n iversity of Sc ience and T echnology ( Sc ience and T echnology) Ap r 2008 1, 2 1 1, 2 段西发 , 吕全义, 齐培艳 ( 1 , 7 10072; 2 , 030024) : 简单介绍了单机下稀疏线性方程组系数矩阵的优化存储方式, 重点研究了并行环 下分 布式优化存储的实现及求解稀疏线性方程组的多分裂并行算法的实现 应用优化算法与优化前 算法在 HP RX 2600集群系统上, 基于M P I( M essage P assing Interface) 的 C 语言并行程序进行数 值试验 结果表明, 相对于优化前算法, 优化算法既节省了内存空间, 又减少了计算时间 同时, 由 于存储的减少, 扩大了处理问题的规模, 具有一定的应用价值 : 稀疏矩阵; 优化算法; 并行算法; H P RX 2600集群 : TP302 1 : A : 1007 - 855X ( 2008) 02 - 0100 - 08 ParallelA lgorithms Optim ization of Sparse L inear Equations 1, 2 1 1, 2 DUA N X ifa , LV Q uany i, Q I Pe iyan ( 1 Schoo l o f Sc ience, N orthw estern P o ly techn ica l U n ive rsity, X ian 71007 2, Ch ina; 2 School o f A pplied Sc ience, T a iyuan U n ive rsity o f Science and T echno logy, T aiyu an 030024, Ch ina) Abstract: In th is p aper, the optmi a l storage m ethod o f coe ff icient m atrix o f sparse linear equat ions under the cir cum stance of sing le com puter is introdu ced, w ith an em ph asis on the achievem ent o f distributed optmi al sto rage m ethod under the p arallel circum stance and the ach ievem ent ofM u lt i- splitting para llel a lgorithm o f sparse linear equ ation s Opt mi al algor ithm and no t- optmi a l a lgorithm are adopted in H P RX 2600 c lu ster to carry out num eri ca l expermi ents b ased on M P I ( M essage Passing Interface) C program langu age para lle l program It is ind icated through the resu lts th at optmi al a lgorithm not on ly saves storage, but also reduces com putat iona