High performance solution of the complex symmetric eigenproblem (Q1275444)

The emphasis of the paper is on the computation of the eigenvalues and eigenvectors of full symmetric, but non-Hermitian matrices if only a part of the eiqenspectrum, e.g. \(20\%\) is needed. A sketch of the theory is presented with reference to preceding publications. Interesting numerical experiments are carried out for a comparison of the new algorithm and the LAPACK routines on a single processor of a Cray J90, a DEC Alpha 8400 server and SGI Power Challenge 8000 and 10.000.