Re: [gnugo-devel] semeai_3_15.1a

[bump]

Gunnar wrote:

> > -  struct dragon_data *d = &(dragon[pos]);
> > -  struct dragon_data2 *d2 = &(dragon2[d->id]);
> > +  struct dragon_data *d;
> > +  struct dragon_data2 *d2;
> >    
> > +  d = &(dragon[pos]);
> > +  d2 = &(dragon2[d->id]);
> 
> I don't see why this would be better.
> 
> > RCS file: /cvsroot/gnugo/gnugo/engine/optics.c,v
> > retrieving revision 1.63
> > diff -u -r1.63 optics.c
> > --- engine/optics.c 2 Jan 2003 00:23:28 -0000       1.63
> > +++ engine/optics.c 4 Jan 2003 03:09:46 -0000
> > @@ -425,7 +425,7 @@
> >      }
> >    }
> >    
> > -  if (0 && (debug & DEBUG_EYES)) {
> > +  if (1 && (debug & DEBUG_EYES)) {
> 
> I'm rather negative to having this on by default. When I do owl tuning
> I frequently use DEBUG_EYES and then the output for the primary
> domains are badly cluttering up the traces. Should we split off this
> as a separate debug flag?

This was accidentally left in after I did some debugging, sorry.
I've reverted both these changes.

> Why remove this? I think it's a big win not having to test whether
> move is NULL, like this,
> 
> > -      *move = PASS_MOVE;
> > +      if (move) *move = PASS_MOVE;
> 
> some 15 times in the function.

I didn't change this. I don't see why this is a big win. One way you
test (move) being a null pointer, the other way you set a dummy
variable only to discard the result. In any case, it's misleading to
point out that this happens 15 times in the function since this will
probably happen only once per call to do_owl_analyze_semeai.

Nando wrote:

> A short comment : loosing or winning semeais may (and often do) have very
> large repercussions. Limiting ourselves to 250 nodes looks a bit too few to
> me.

I agree and will test to see how expensive it is to double the semeai
nodes to 500.

Dan