[Couenne] Runnig time

[Giacomo Nannicini]

Dear Nathalie,
as far as I know, this problem is not due to Couenne, and it appears
on some Linux machines, depending on your kernel (do you have Ubuntu?
I've seen it happen on Ubuntu machines, mostly). I believe that there
is a bug in the way CPU time is computed.
This is the fix I found: in CoinUtils/src/CoinTime.hpp, look for the
CoinCpuTime() function and, around line 130, write the following:

cpu_temp = (static_cast<double>(usage.ru_utime.tv_sec) +
             static_cast<double>(usage.ru_stime.tv_sec));
cpu_temp += 1.0e-6*(static_cast<double>(usage.ru_utime.tv_usec) +
                     static_cast<double>(usage.ru_stime.tv_usec));

instead of what is there - which I do not remember right now, but was
something along the lines of:

cpu_temp = static_cast<double>(usage.ru_utime.tv_sec);
cpu_temp += 1.0e-6*(static_cast<double>(usage.ru_utime.tv_usec));

That should fix it. Basically you have to add User and System CPU time
together. Recompile and it should work. Let me know if it doesn't.

Giacomo

On Tue, Dec 21, 2010 at 10:10 AM, njamett at uantof.cl <njamett at uantof.cl> wrote:
> Hi
>
> I'm working on a MINLP problem and solving it with Couenne.
>
> Sometimes it get stuck trying to get an optimum always at the same time (256
> seconds).
>
> Cbc0010I After 81300 nodes, 30318 on tree, -1.12361e+11 best solution, best
> possible -2.45689e+11 (256.52 seconds)
> Cbc0010I After 81400 nodes, 30353 on tree, -1.12361e+11 best solution, best
> possible -2.45689e+11 (256.52 seconds)
> Cbc0010I After 81500 nodes, 30393 on tree, -1.12361e+11 best solution, best
> possible -2.45689e+11 (256.52 seconds)
> Cbc0010I After 81600 nodes, 30429 on tree, -1.12361e+11 best solution, best
> possible -2.45689e+11 (256.52 seconds)
>
> What's these mean? Is it a formulation problem? Or a PC capacity problem?
>
> Thank you
>
> Best regards
>
> Nathalie Jamett
> PhD Student
> University of Antofagasta
> Chile
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/couenne
>