> Not that this is relevant > for a symmetric (Hermitian) eigensolver, which will always have real > eigenvalues. > > Anne Yep, sorry I failed to notice this wasn't an extension to eig, but about a new function eigs only for symmetric matrices. My bad.