[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] Irregular edges
From: |
Daniel Bump |
Subject: |
Re: [gnugo-devel] Irregular edges |
Date: |
Mon, 15 Oct 2001 21:00:19 -0700 |
> I'm aware of this problem, but am adding the following functionality
> before fixing it:
> - CPU time used per problem.
> - other useful statistics per problem.
> - capturing of -t output for all problems.
> - storing sgf trace files for owl_* problems.
> - saving problem solution info to XML, rather than HTML directly,
> allowing much more flexible reporting/sorting, etc.
Some of this doesn't seem high priority to me.
But for the genmove problems the html view really does need
the results of top_moves. Gunnar seems to have done some
debugging around October 3 and wrote:
> > This is much worse than that. In fact the same move isn't even
> > generated by "gg_genmove black" run twice:
> >
> > loadsgf games/trevor/trevor_25.sgf 31
> > = black
> >
> > gg_genmove black
> > = G2
> >
> > gg_genmove black
> > = F4
> >
> > Needless to say, this requires an investigation.
>
> As I suspected this aspect of the engine is broken by design (my
> design, as it happens). The problem is that the caching scheme in
> examine_position() causes make_worms() and make_dragons() not to be
> rerun the second time do_genmove() is called, but the list of move
> reasons is reset and the ones generated in make_worms() (and possibly
> make_dragons() too) are lost.
I assume this to the NEEDS_UPDATE scheme.
Inge wrote:
> I simple fix would be to cache the generated move reasons in the
> worm[] and dragon[] arrays. Then they could be quickly reissued when
> examine_position is called again.
and Gunnar seemed to endorse this idea. I think we do
need to do something about this.
Dan