一种基于Doolittle LU分解的线性方程组并行求解方法

Parallel Solving Method of Linear Equations Based on Doolittle LU Decomposition

矩阵方程的快速求解是矩量法计算电大问题的关键,LU分解是求解线性方程组的有效方法。该文详细地分析了Doolittle LU分解过程,基于分解过程的特点,在MPI(Message-Passing interface)并行环境下,提出了按直角式循环对进程进行任务分配的并行求解方法。实验证明该方法可以有效地减少进程间数据通信量,从而加快计算速度。

The fast matrix solving is the key of the moment method when computing the electrically large issues.LU decomposition is a efficient algorithm for solving linear equations.In this paper,Doolittle LU Decomposition is described detailedly.Based on the decomposition characteristics,a parallel solving method looping over squares is proposed in MPI （Message-Passing interface） parallel environment.The experiments indicate that the method can decrease communication quantity between processes and accelerate computing speed efficiently.