Posted by Paul Wallner on September 08, 1998 at 10:27:24:

I work with rather big double precision matrices.

For these matrices I have to perform a singular value decomposition,

but I need only the smallest singular value.

Common routines like LAPACK's dgesvd compute the whole set of SV

and this leads to a big loss of performance.

Who knows an appropriate routine which computes the smallest SV?