[ADOL-C] What does it means "Compressed sparse structures will not be available / Only sparsity patterns can be computed" ?

Antonello Lobianco antonello at lobianco.org
Thu Aug 16 10:24:57 EDT 2012


Oh, sorry, I didn't express myself correctly.
I know how to install ColPack and having it recognised by ADOL-C (e.g.
configuring ColPack to use the "/usr" prefix).
However there are some troubles when using ColPack for optimisation
purposes (IPOPT), so I was wondering if enabling the sparse drivers without
ColPack I could still benefit of some "extra" features, in my case e.g. if
enabling sparsity without ColPack I would have been able to calculate and
call the Jacobian only at a user-provided row/column points.
In the povided code snipped the jacobian is called only over these points,
but I don't know if it is stll calculated for all the points.

Thank you,
Antonello


2012/8/16 Kshitij Kulshreshtha <kshitij at math.upb.de>

> Hello, this means that you enabled sparse drivers in the configure
> script but it was not able to find the binary of libColPack or the
> ColPackHeaders.h file. Please make sure you used --with-colpack=DIR in
> the configure command and make sure ColPack is installed correctly in
> DIR. On 64-bit systems you may have to tell ColPack to install the
> binary in ${prefix}/lib64 by passing --libdir=${prefix}/lib64 to the
> configure command of ColPack.
>
> As on 2012-08-16 10:34, Antonello Lobianco did write:
> > Hello, what does this exactly means?
> >
> > Build sparse drivers:          yes
> > Build with ColPack:            no
> > Compressed sparse structures will not be available
> > Only sparsity patterns can be computed
> >
> > I'm using ADOL-C as AD tool for the IPOPT solver and having problems
> > with ColPack when I need to run multiple optimisation problems (in
> > series) I implemented an algorithm for finding sparsity structure for my
> > specific model, at least for the Jacobian.
> >
> > Then I replaced the eval_jac_g() function in the LuksanVlcek1 or
> > MittelmannDistCntrlNeumA examples
> > (ADOL-C/examples/additional_examples/ipopt) with:
> >
> > eval_jac_g(){
> >   if (values == NULL) {
> >     // return the structure of the jacobian,
> >     for(Index idx=0;idx<nzjelements.size();idx++){
> >       iRow[idx] = nzjelements[idx][0];
> >       jCol[idx] = nzjelements[idx][1];
> >     }
> >  }
> >   else {
> >     // return the values of the jacobian of the constraints
> >     jacobian(tag_g,m,n,x,Jac);
> >     for(Index idx=0;idx<nzjelements.size();idx++){
> >       values[idx] = Jac[nzjelements[idx][0]][nzjelements[idx][1]];
> >     }
> >   }
> >   return true;
> > }
> >
> > where vector < vector <Index> > nzelements is the vector of non zero
> > elements in my model, with nzelements[i][0] the row position and
> > nzelements[i][1] the column one.
> >
> > Is this the most efficient implementation or there could be a better way
> > to compute sparsity patterns?
> >
> > Thank you,
> >       Antonello
> >
> >
> >
> >
> > --
> > Antonello Lobianco
> > INRA, Laboratoire d'Economie Forestière
> > 14 Rue Girardet - 54000 Nancy, France
> > Tel: +33.652392310
> > Email: antonello.lobianco at nancy-engref.inra.fr
> > <mailto:antonello.lobianco at nancy-engref.inra.fr>
> > http://antonello.lobianco.org <http://antonello.lobianco.org/>
> >
> >
> > _______________________________________________
> > ADOL-C mailing list
> > ADOL-C at list.coin-or.org
> > http://list.coin-or.org/mailman/listinfo/adol-c
> >
>
> --
> Dr. Kshitij Kulshreshtha
>
> Institut für Mathematik,
> Universität Paderborn,
> Warburger Straße 100,
> 33098 Paderborn.
>
> Büro: A3.235
>
> Privatanschrift:
> Arnikaweg 62
> 33100 Paderborn.
> _______________________________________________
> ADOL-C mailing list
> ADOL-C at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/adol-c




-- 
Antonello Lobianco
INRA, Laboratoire d'Economie Forestière
14 Rue Girardet - 54000 Nancy, France
Tel: +33.652392310
Email: antonello.lobianco at nancy-engref.inra.fr
http://antonello.lobianco.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/adol-c/attachments/20120816/902c6bb6/attachment.html>


More information about the ADOL-C mailing list