求解广义特征值问题的并行保域行列式查找法

结构分析领域有着重要应用的广义特征值问题的并行算法，因为难度很大，且当问题的规模较大时还必须有先进的计算环境支持，所以迄今研究得很少。文中提出了一种适用于流水线型向量机的求解大型稀疏实对称矩阵广义特征值问题的并行保域行列式查找法。该方法不但保持了传统的行列式查找法的优点，而且克服了其迭代不收敛、漏根等缺点，并具有较高的速度加速比。该算法在ＹＨ－１计算机上进行了数值实验，结果表明该法是一种求解大型对

Parallel Region-Preserving Determinant Search Method for Solving Generalized Eigenproblem

The parallel algorithm of generalized eigenproblems applied significantly to the structural analysis domain has been studied little so far, because this study is extremely difficult, and needs the support of advanced computing circumstances for large scale problems. The parallel region preserving determinant search method for solving the generalized eigenproblem of a large sparse real symmetric matrix is presented in this paperl which is suited to pipelined vector computers. This method not only retains the advantages of the conventional determinant search method, but also overcomes its disadvantages such as leaking roots and nonconvergence, and has a higher speed-up. The author have put the method into practice on the YH-1 computer. The numerical results show that it is an efficient algorithm for solving large symmetric generalized eigenproblems.