To provide a parallel processing method of an operation for performing the operation of an eigenvalue problem at a high speed in a shared memory scalar computer.
A method for solving the eigenvalue problem is divided into three stages, and is composed of a step for triply diagonalizing a matrix, a step for determining an eigenvalue and a proper vector from a triple diagonal matrix, and a step for determining the proper vector of an original matrix by converting the proper vector determined from the triple diagonal matrix. Particularly, since an operation load of the step for triply diagonalizing the matrix and the step for determining the proper vector of the original matrix is large, the eigenvalue problem is solved at a high speed by executing processing by parallelizing these steps.
COPYRIGHT: (C)2004,JPO
JP8166941A | ||||
JP9212489A | ||||
JP3156568A |
分散メモリ型並列計算機によるブロック化Householder法の性能評価,情報処理学会論文誌,社団法人情報処理学会,1998年 7月15日,第39巻、第7号,p2391~2394
並列固有値ソルバーの実現とその性能,情報処理学会研究報告(97-HPC-69),社団法人情報処理学会,1997年12月12日,第97巻、第121号,p49~54
Motoaki Hisagi