[Osi] osi/clp goes to infinite loop

Renaud Lepère renaud.lepere at alma.fr
Wed Apr 28 05:21:57 EDT 2010


Hello,

I am using clp through osi. But i have a problem : when i solve
one of my lp problem clp goes into an infinite loop. And i can't
even stop it with maximum iterations (since the number of iterations
do not increase).

Attached is a generated code that shows the problem (RawValue is
a hack to avoid precision problems on double).

Apparently my problem is unfeasible ; but is there a way to fix this
problem ?  Otherwise, is it possible to stop after the pre-solve if the 
problem is infeasible (even at a given precision) ?

Thanks,

Renaud

Following is the trace i have got with osi-1.02.1 and after
the trace with osi-0.97.1

Clp0024I Matrix will be packed to eliminate 97 small elements
Coin0507I Presolve determined that the problem was infeasible with 
tolerance of 1e-008
Clp3003W Analysis indicates model infeasible or unbounded
Clp0006I 0  Obj 5585.84 Primal inf 2.3448e+009 (50)
Clp0006I 48  Obj 6698.48 Primal inf 978.444 (48)
Clp0006I 89  Obj 7006.6 Primal inf 505.002 (24)
...
Clp0006I 243  Obj 5585.84 Primal inf 3.84652e+021 (80) Dual inf 
3.59101e+013 (35)
Clp0029I End of values pass after 243 iterations
Clp0006I 243  Obj 5585.84 Primal inf 3.84652e+021 (80) Dual inf 
3.59101e+013 (35)
Clp0029I End of values pass after 243 iterations
...




Previously:
Clp0024I Matrix will be packed to eliminate 97 small elements
Coin0507I Presolve determined that the problem was infeasible with 
tolerance of 1e-008
Clp3003W Analysis indicates model infeasible or unbounded
Clp0006I 0  Obj 5585.84 Primal inf 2.3448e+009 (50)
Clp0006I 48  Obj 6698.48 Primal inf 978.444 (48)
Clp0006I 89  Obj 7006.6 Primal inf 505.002 (24)
Clp0001I Primal infeasible - objective value 7006.6
Clp0032I PrimalInfeasible objective 7006.597715 - 89 iterations time 0.032
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: osi_100427.inl
URL: <http://list.coin-or.org/pipermail/osi/attachments/20100428/c7abff65/attachment.ksh>


More information about the Osi mailing list