高斯消去的并行化研究  

Study on Parallel of Gaussian Elimination

作　　者：刘琳[1] 刘青昆[1] 宋小雨[2] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116029 [2]大连职工大学经济管理系,辽宁大连116033

出　　处：《计算机工程》2011年第8期40-42,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(20633050)

摘　　要：为满足大规模线性方程组对内存容量的要求,针对对称方程组提出一种高斯消去的并行化方案。对称方程组在高斯消去过程中其子方阵的对称性仍然存在,因此在并行计算时只读入和计算三角部分的数据,从而减少储存空间的大小,提高并行效率。测试表明,该方案的并行效率优于传统算法,可应用于对称方程组的大规模数值计算中。In order to solve the memory requirement of large-scale symmetric equations, a new kind of parallel Gaussian elimination method which is based on the symmetric equations is put forward. Analysis shows that the sub-square is symmetry in the process of Gaussian elimination. In order to reduce the space application and ialprovc the efficiency of parallel program, the parallel program can only read and calculate triangle section of linear equations. After test, its efficiency is better than traditional algorithm. This method can be widely applied to large-scale symmetric equations in the numerical calculations.

关 键 词：并行计算 高斯消去 对称矩阵 矩阵划分 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]