[Scip] scip: automated tests
Tobias Achterberg
achterberg at zib.de
Wed Feb 3 16:04:56 MET 2010
You need to look in the log file (*.out I think). My guess is that SCIP was overshooting
the time limit too much and was therefore killed by the system process time limit (I think
it is set to 110% of the time limit). If the process is killed, then there is no
statistics output which the scripts can parse.
Tobias
Robert Schwarz wrote:
> Hi all,
>
> I'm running some automated tests with a small timelimit. Normally, that
> results in some instances not being solved to optimality, producing a
> line like this:
>
> ibm32 BP 30 2790 30 2790 475.438791 485 2.0 183137 37 1200.0 timeout
>
> Here we can see the dual bound 475.4 from the lp relaxation and the
> primal bound 485 from a heuristic that is run once in the beginning.
> The last column states correctly, that the time limit stopped the
> solving process.
>
> Sometimes, however, I get lines like this:
>
> gd95c -- 0 0 0 0 -1e+20 1e+20 -- 0 0 1200.0 abort
>
> Where it says `abort' instead of `timeout' and no bounds are given at
> all. But when I run this instance by hand, I can see that the primal
> bound is computed after a couple and some value is displayed as dual
> bound from the LP (which is really small before separation).
>
> Why is this, and what can I do about it?
>
> I guess, the bounds are already computed in the second case, but yet
> reported back globally, because the current subproblem (root node) is
> still active or something. But isn't really accurate, since we have some
> weak bounds already after a short amount of time.
>
> Cheers,
>
More information about the Scip
mailing list