Re: [eclipse-users] Memory management for Eplex with OSI clpnbncblpcbplp

Dear Daniel,

Thanks for reporting the problem. I have looked at it, and indeed memory 
is being leaked, about 100K per iteration. As this memory is allocated 
and lost in the clpcbc code, you would not see it in the ECLiPSe statistics.

I have tracked the problem down to the way a MIP problem model is 
created (this is done even if you are solving a linear problem). I will 
need to check with the COIN people to see how this can be fixed.

Cheers,

Kish


Daniel Lucraft wrote:
> Hi,
>
> when I run this:
>
> %----------------
> :-lib(eplex).
> :- eplex_instance(prob).
>
> main:-
>    (
>      for(_, 1, 100000)
>    do
>      solve(_, _)
>    ).
>
> solve(Cost, Vars) :-
>    Vars = [A1,A2,A3,B1,B2,B3,C1,C2,C3,D1,D2,D3],
>    prob: eplex_cleanup,
>    prob: (Vars $:: 0.0..1.0Inf),
>    prob: (A1 + A2 + A3 $= 21),
>    prob: (A3 + B3 + C3 + D3 $=< 40),
>    prob: eplex_solver_setup(min(7*D3)),
>    prob: eplex_solve(Cost),
>    prob: eplex_cleanup.
> %----------------
>
> with the OSI clpcbc solver, the memory of the process shoots up and away 
>   until it reaches enormous levels. It was using 850MB of RAM and 1.5GB 
> of  swap space just now when I killed it. Running in TkEclipse the 
> 'statistics' panel doesn't show any of this.
>
> Is this expected, normal or an error in my code?
>
> thanks,
> Dan
>
> _______________________________________________
> ECLiPSe-Users mailing list
> ECLiPSe-Users_at_...2...
> http://www.crosscoreop.com/mailman/options/eclipse-users
>