Re: [gnugo-devel] Patch: Strategic effect for strategic attacks and defenses

[Arend Bayer]

Inge wrote:

> > If I am mistaken, i.e. if move_influence is recomputed here, then this
> > would be a big performance problem.
>
> Are you sure? I did a profile, and the run time for delta_territory
> was very insignificant.

The profile (if you mean the flat profile) only shows the time spent
in influence_delta_territory itself. What is time expensive is the
influence recomputation caused by calling compute_move_influence. This
expensive partly because of accumulate_influence, but mostly because of
the numerous influence pattern helpers with reading constraints that have
to be evaluated.

The call graph tries to include this information, too, but has
systematic inaccuracies (that undervalue compute_move_influence).

Arend