[Coin-discuss] problem with CLP

[John J Forrest]

Mathieu,

Sounds like a bug in Clp to me.

To get round it try your suggestion - very little overhead in copying but
you will be starting from an all slack basis which may be slower if more
reliable.  Or just use getModelPtr()->allSlackBasis which should give same
effect.

To help debug - try using getModelPtr()->saveModel as well as writing mps
file.  Then see if it has problems using restoreModel.  If so send me the
file.

John Forrest


                                                                           
             mathieu lacroix                                               
             <mathieu.lacroix@                                             
             isima.fr>                                                  To 
             Sent by:                  coin-discuss at list.coin-or.org       
             coin-discuss-boun                                          cc 
             ces at list.coin-or.                                             
             org                                                   Subject 
                                       [Coin-discuss] problem with CLP     
                                                                           
             09/08/06 03:22 AM                                             
                                                                           
                                                                           
             Please respond to                                             
             Discussions about                                             
                open source                                                
               software for                                                
                Operations                                                 
                 Research                                                  
             <coin-discuss at lis                                             
              t.coin-or.org>                                               
                                                                           
                                                                           




Hi,
I develop a branch and cut algorithm with BCP for a metric constraint
based formulation. To separate metric constraints, I solve a linear
program (wich is totally different from the linear relaxation of the
integer linear program I want to solve) with an Osi/Clp object. In each
separation, only the objective function is changed in this linear
program. Sometimes, a bug appears in the resolution of the metric
separation linear program : whether the resolution takes a long time
(more than 10 minutes) or it does not terminate (I had to kill it after
1000 minutes of computation whereas the limit time is 300 minutes).
However, for each of these linear programs which make troubles, if I
solve it using Osi/CLP (by reading the mps file and using the osi
example "basic" for instance), it takes less than 10 seconds to solve
it. I guess the bug comes from the fact that the Osi object is not in
the same state after one resolution. Is there any solution for the
problem different than creating the linear program at each iteration (I
have not yet tried this solution)?

Best regards,
Mathieu

--
Mathieu LACROIX
mathieu.lacroix at isima.fr
Université Blaise Pascal - Clermont Fd II
Laboratoire LIMOS
Bâtiment ISIMA - Bureau D112
Complexe scientifique des Cézeaux
63 177 Aubière, Cedex - France
Tel : 04.73.40.79.48
Fax : 04.73.40.76.39

_______________________________________________
Coin-discuss mailing list
Coin-discuss at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/coin-discuss