[SCIP] About variances in solving time

[Stefan Vigerske]

Hi,

On 27/11/2024 00:03, s schnug wrote:
> I'm somewhat afraid *and this is a question for SCIP-devs* that SCIP is not
> (opposed to some other solvers based on cooperative-scheduling /
> task-interleaving) necessarily following the same path before being cut-off
> by some time-limit (or terminating with a solution). I interpret the
> source-code like "#define DEFAULT_HEURTIMELIMIT    60.0 /**< time limit for
> a single heuristic run */" as indication, that effects/influences from 1
> can make two solver-runs "diverge" in their decisions (run A can work with
> heuristic success solution; run B cannot) => the bad thing here is that
> decisions can have exponentially bad influence on solving-time (compare
> with "performance variability in mixed-integer programming").

SCIP by default does not use timelimits for components like primal 
heuristics. So if the user does not enable time limits (e.g., 
limits/time), then it should behave deterministic.

The DEFAULT_HEURTIMELIMIT mentioned here is in a primal heuristic 
("scheduler") that is disabled by default. I cannot say why this 
heuristic has such an option, maybe it is some left over from some 
experiments. It would be nice to change its default to -1 (no limit), so 
that the heuristic behaves deterministically if enabled.

Stefan