GSL, FFI, GCL, Windows; was: find_root with bfloats

I am solving a convex optimization problem with linear 
inequality constraints, and lbgfs (augmented_lagrangian) 
handles only equalities.

I remember your donlp2 effort, and recall that it fell 
though for some reason.

But, I also realize that Robert is right, and the CL 
interface would pass double floats to the GSL's C code.  But 
then again, given the functionality that this would open up, 
it wouldn't be such a bad thing, would it?

			Kostas

On 09/29/10 02:27 PM, Raymond Toy wrote:
>   On 9/29/10 1:56 PM, Kostas Oikonomou wrote:
>> I agree that the ideal is a library that fits the computer algebra way
>> of doing things, sometimes symbolic, sometimes numeric.  The GSL is
>> purely numeric, as far as I know.  But maxima is missing a lot of
>> numerical capabilities: optimization is the one that is prominent for
>> me right now, I have to do it in Mathematica and move the results to
>> maxima.  I was encouraged by the ease by which Ray converted find_root
>> to use bigfloats, and thought something similar might be doable with
>> the GSL and its Common Lisp interface.
> What kind of optimization are you trying to do?  There is lbfgs and
> minpack_lsquares.  lbfgs does unconstrained minimization and
> minpack_lsquares is unconstrained least squares.
>
> I also have an interface to donlp2 (constrained and unconstrained
> nonlinear optimization with equality and inequality constraints) which
> worked nicely.   But I found out later that it can't be distributed and
> you must get a license for it if you want to use it.  Or something.  I'm
> not really sure what the license constraint is.
>
> Ray
>
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima