纺织高校基础科学学报2011，Vol.24Issue(1)：21-25,5.
Jacobi矩阵特征值的并行算法
The parallel algorithm of eigenvalues of Jacobi matrix
摘要
Abstract
A parallel algorithm for solving eigenvalues of real tridiagonal matrices is presented, which mainly apply to Jacobi matrices. This algorithm firstly isolates a small single region which only includes a root from the large region of roots of the polynomials by using the Sturm method; secondly looks for the accurate root by employing the Dichotomy and Newton iteration further. When considering the equally tasks among processors, the large region is divided into small regions, then these small regions are distributed among processors by tums. Every processor independently finishes the tasks of solving eigenvalues. During parallel computing, there are no communications between processors. This strategy can make the equally tasks among processors, the algorithm's parallelism is preferable. Finally, some numerical results demonstrate that the parallel efficiency can reach O. 85 at least, so this parallel algorithm is high efficiency.关键词
Jacobi矩阵/Sturm法/牛顿法/并行算法/并行效率Key words
Jacobi matrix/ Sturm method/ Newton method/ parallel algorithm /parallel efficiency分类
数理科学引用本文复制引用
刘艳红,吕全义..Jacobi矩阵特征值的并行算法[J].纺织高校基础科学学报,2011,24(1):21-25,5.基金项目
陕西省自然科学基金资助项目(2009JM1008) （2009JM1008）
