Re: Using LAPACK in octave

[David Bateman]

John W. Eaton wrote:
> On 11-Sep-2007, David Bateman wrote:
> 
> | David Bateman wrote:
> | > 
> | > I've often felt that Octave should change to use the same solver as
> | > Matlab, especially now as the sparse matrices in Octave use a QR solver
> | > with pivoting and don't give exactly the same results as the dense 
> solvers..
> | 
> | If you want another reason to change consider Figure 3.3 on the page
> | 
> | http://www.netlib.org/lapack/lug/node71.html
> | 
> | for N = 500 dgelss is about 35 times slower than dgelsy
> 
> Switching is fine with me if the results are approximately as good.  I
> chose dgelss because it was available and I knew about it.

If we want to stay with an SVD solver we should at least change to
dgelsd that is significantly faster than dgelss and gives essentially
the same results.

D.