用于数值分析的分块求解法

A Block Solving Technique for Numerical Computation

本文提出了用于数值分析的分块求解方法,即利用计算机的硬盘存贮,对线性代数方程组的系数矩阵进行分块高斯消去求解,在计算机内存中最少只需开辟系数矩阵中两行元素的存贮空间,本方法的突出优点是,能大大节省计算机内存,可使计算机的解题能力提高数十倍,为求解大规模的题目提供了方便.文中还给出了相应的 FORTRAN 子程序框图,经实例计算表明,该分块求解法经济有效,简单实用.

In this paper,a block solving technique is put forward to solve linear al- gebraic equations in numerical computation.By making use of hard pan storage of com- puter,linear algebraic equations can be solved through Gaussian—Jordon elimination of the coefficient matrix block by block.In this method,only two rows of the whole coeffi- cient matrix need be stored in the CPU memory of a computer.As a result,a great amount of CPU memory can be saved and the computer's solving ability will be raised by tens of times.