A Parallel Subspace Iteration Method for Generalized Eigenvalue Problems Based on Multi-core Platform

Abstract

A parallel subspace iteration method for solving eigenvalue problem of based on multi-core platform is presented, which can solve several extreme eigenpair in parallel. Compared with Jacobi-Davidson method, the dimension number of the subspace in the method keeps unchanged, which makes it easier for the programming implementation. Numerical experiments are performed with a quad-core computer under the joint programming environment of Intel Fortran and OpenMp. The computation of the plane wing frequency and aircraft pylon for a real model airplane is taken as an example. As a result, the first 10 frequencies of a plane wing and an aircraft pylon are provided which shown the efficiency and applicability of our parallel computation algorithm.