块对角线性方程组的一种分布式并行迭代算法

An improved iteration algorithm for solving linear equations

提出了一种在分布式环境下求解块对角线性方程组的并行ILU(0)-JACOBI迭代算法,与传统算法不同之处是首先选取一个合适的预处理矩阵对原系数矩阵进行处理,以改善矩阵的条件数。数值结果表明,算法具有较快的收敛速度,在分布式并行环境下具有较好的并行度和较低的存储要求。

This paper presents an iteration algorithm named ILU（0） -JACOBI for solving diagonal linear systems on distributed computers. Different with traditional algorithms, it first chooses a proper preconditioner to transform the linear system, for improving the spectral properties of the coefficient matrix. Experimental results show that the algorithms and has quicker convergence speed, and have better parallel level as well as lower request for memory.