基于分布式集群的线性方程组求解

在很多应用领域中都需要求解大规模的线性方程组,然而由于单机和多处理器系统的计算能力有限,这个问题没有得到很好地解决。该文提出了一种求解这类方程组的并行算法及其在Linux集群上的实现。在介绍了该算法的矩阵分块、通信机制和改进的高斯消元法之后,描述了该算法的实现流程,并进一步对其通信开销、计算开销和存储需求进行了分析。

Distributed Cluster-based Solution Techniques for Linear Equations

In many applications,very large-scale matrix need s to be solved,however,because single or multiprocessor systems have some limitations of computing resource,the problem was not solved perfectly.This paper will discuss a distributed cluster-based solution techniques for linear equations and the implementation on Linux cluster.Our works include partition of matrix,communication mechanism,improved Guassian Elimination and a master-slaver algorithm etc.What's more,Pseudcode of the algorithm in C language is presented.Finally,we give the cost of computing,memory and communication of this algorithm.