[Scip] SCIP doesn't start the solution process after presolving

[Stefan Vigerske]

Hi,

> yeah you were right. I stopped it after the root node and the statistics 
> showed it took more than 3 hours to solve the LP. I guess I was 
> expecting some kind of output as soon as the main solution process 
> starts. That's why I assumed it was spending this much time to load the 
> problem.
> 
> Cplex' barrier method solves the problem in 5 s. However I used it as a 
> stand alone solver. When I call the barrier method in SCIP using
> SCIP_CALL( SCIPsetCharParam(scip, "lp/initalgorithm", 'b') );
> I get an assertion error:
> lp->nlpicols == 0 && lp->nlpirows == 0, file [...]\src\scip\lp.c line 9244
> what is going wrong?

Above that assert, there is a @todo in the code:
"the assert might be to hard since in case of solving the LP with the
barrier the lpistate is also NULL"
So maybe you try commenting out this assert and see whether that works
better?

> Another question I have: I want to disable presolving. I use
> SCIP_CALL( SCIPsetBoolParam(scip, "lp/presolving", FALSE) );
> but presolving is executed nonetheless.  

This only disables the presolving in the LP solver.
To disable CIP presolve, you can set the parameters
 presolving/maxrounds = 0
 presolving/maxrestarts = 0

See also the settings file settings/presolving/off.set.

Stefan

> I load the default plug-ins
>      SCIP_CALL( SCIPincludeDefaultPlugins(scip) );
> and then call
>      SCIP_CALL_ABORT(SCIPsolve(scip));
> 
> Thank you
> Sebastian
> 
> On 16/08/2010 9:12 PM, Tobias Achterberg wrote:
>> Hi Sebastian,
>>
>> my guess is that the root LP solve is just incredibly hard (or the LP 
>> solver has bad luck and struggles a lot with degeneration).
>>
>> You can activate the LP solver output with the "display/lpinfo" 
>> parameter. This should give you a clue what happens. If it is the LP 
>> solve that takes so long, you may want to try the barrier algorithm 
>> (which is available by using CPLEX as LP solver, but not for Soplex): 
>> "lp/initalgorithm" = "b".
>>
>>
>> Tobias
>>
>>
>> Sebastian Ruther wrote:
>>>    Hello,
>>>
>>> I am trying to solve a MIP but SCIP or Cplex gets stuck somewhere.
>>> I load the default plug-ins and after presolving it doesn't start the
>>> main solution process
>>> .
>>> Presolving deletes 1300 variables and 800 constraints. After that the
>>> problem has 16000 constraints and 19600 variables with 19200 being
>>> binary and the rest continuous.
>>> After displaying the presolving time (which is 157 sec) nothing happens
>>> for at least one hour. The CPU is busy but like I said the main solution
>>> process is not started. What's going on? Is SCIP transferring the
>>> problem to Cplex? That shouldn't take so long for a medium seized 
>>> problem.
>>> After 1-4 hours it starts the solution process.
>>>
>>> Regards,
>>> Sebastian
>>> _______________________________________________
>>> Scip mailing list
>>> Scip at zib.de
>>> http://listserv.zib.de/mailman/listinfo/scip
>> .
>>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>