On Sun, 02 Nov 2003 17:25:34 +0100 Rob Hooft wrote: Rob> - lapack_routine = lapack_lite.dgelss Rob> + lapack_routine = lapack_lite.dgelsd Well, here the underlying LAPACK routine was changed to the newer and significantly faster divide-and-conquer routine. (Same holds for complex version.) This could be the problem, which you should test. See LAPACK documentation for details. Nevertheless I would advise against reversing that change, as performance diffferences can really be large (although I haven't used either one of these specific functions here). Maybe you can keep a copy of the old version in your own project if really necessary? (After all there seems to be some agreement that you were just "lucky" to find a working algorithm in the first place.) Greetings, Jochen -- Einigkeit und Recht und Freiheit http://www.Jochen-Kuepper.de Liberté, Égalité, Fraternité GnuPG key: CC1B0B4D (Part 3 you find in my messages before fall 2003.)