用改进的界限法求解大型稀疏线性方程组

在求解大型稀疏线性方程组中,基于索引信息界限思想,本文提出一个新的数据结构,以非零元同等数量索引信息压缩存储稀疏系数矩阵。除每行第一个非零元存入一个界限信息外,其他非零元均只存入相应列标。矩阵的压缩还原过程仅需执行O(n)次加减运算.消元过程还运用存储复盖技术,从而进一步大幅度降低存储量,提高运算速度。该法已编制标准过程并以实例上机通过。

A MODIFICATON OF BOUNDARY METHOD FOR SOLVING LARGE SPARSE LINEAR EQUATIONS

Based on the Index Information Control[2], a new data structure for solving large sparse linear equations is proposed in this paper. In the modified boundary method, a sparse matrix is stored by compacting the nonzero elements along with the same number of the indexing information. The first nonzero element of each row is stored with a boundary information. But the other nonzero elements of each row are stored only with the oorresponding column number. The compacting and recovering procedures only require O(n) additions or subtractions. Furthermore, memory overlay technique is used in elimination so that much storage ean be saved and the efficiency of operation can be improved. This method bas been given in standard procedure and tested in computers.