稀疏线性方程组并行解法的优化

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,重点研究了并行环境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 RX2600集群系统上,基于MPI(Message Passing 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.

Parallel Algorithm's Optimization of Sparse Linear Equations

In this paper,the optimal storage method of coefficient matrix of sparse linear equations under the circumstance of single computer is introduced,with an emphasis on the achievement of distributed optimal storage method under the parallel circumstance and the achievement of Multi-splitting parallel algorithm of sparse linear equations.Optimal algorithm and notoptimal algorithm are adopted in HP RX2600 cluster to carry out numerical experiments based on MPI(Message Passing Interface) C program language parallel program.It is indicated through the results that optimal algorithm not only saves storage,but also reduces computational time.At the same time,the algorithm has certain applied value because it extends the scale of processing problem because of the decrease of storage.