gslclapack

[James Bergstra <james dot bergstra at umontreal dot ca>]

Please excuse a sort of continued email (further to my last comment on cholesky
decomposition).  I notice that lapack implements cholesky decomposition of real
symmetric matrices in a file called dpotrf.f.  In that file, there is apparently
a 'blocked' version of the decomposition algorithm that uses level3 blas
routines instead of level2.

On one hand, I'm jealous that lapack has a fancy routine, and react "let's
implement that for gsl too" and on the other hand, I think "why is gsl
re-implementing (what I believe is) a very standard library?"

What would be the pros and cons of slowly migrating the number-crunching from
gsl_linalg routines to gsl_clapack ones, while leaving argument-checking to the
gsl_linalg ones.  I think the division works well in the case of gsl_blas vs.
gsl_cblas.

My assumption through all this, is that clapack offers more routines, and faster
implementations than those of the GSL.  The main advantage I hope for is that
the lapack implementation is seperate from the gsl, and can be chosen at
application link-time.

-- 
james bergstra
http://www-etud.iro.umontreal.ca/~bergstrj