A parallel algorithm for the partial eigensolution of sparse symmetric matrices on the CRAY Y-MP (Q1179256)

The author presents a parallel algorithm to calculate only the leftmost \(p\) eigenpairs of finite element matrices with very large dimensions. It is shown that the vectorized algorithm achieves an average speed up 3.34 with respect to the best scalar code. The parallel version improves this average speed up by factors 1.72, 2.24, 2.63 with two, three and four processors, respectively. The result suggests that the proposed method ranks among the most efficient parallel algorithms.