On Mon, Feb 9, 2009 at 16:25, M Trumpis firstname.lastname@example.org wrote:
I played around with a C translation of that test program, and found that dgesvd (but not dgesdd) happens to converge and return all non-negative singular values for both operators I was having trouble with. I'm also looking at the Octave code just now, and I think they're using dgesvd also. Any one know why numpy uses dgesdd? speed?