[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[gnugo-devel] mixed patch

From: Gunnar Farneback
Subject: [gnugo-devel] mixed patch
Date: Sun, 25 Nov 2001 12:27:37 +0100
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode)

I've put up a patch with mixed contents at

- persistent caching added to owl_does_attack()
- owl_does_defend() and owl_does_attack() now understand ko
- owl_determine_life() revised
- new function silent_examine_position()
- find_more_owl_attack_and_defense_moves() revised
- ko valuation revised
- 1D revision of print_eye() in optics.c
- compute_eyes_pessimistic() revised
- bugfixes in evaluate_diagonal_intersection()
- further 1D revisions in sgfdecide.c
- new GTP commands owl_does_attack and owl_does_defend
- influence tuning
- owl tuning
- new test cases

The new function silent_examine_position() is identical to
examine_position() except that it turns off all traces, debug output,
and sgf output while examine_position() is run. This is something we
want to do fairly often, so silent_examine_position() simplifies the
code anywhere it applies.

I've improved owl_does_defend() and owl_does_attack() to keep track of
ko return codes. Also added persistent caching to owl_does_attack()
which didn't yet have it.

There was a fairly serious bug in evaluate_diagonal_intersection(),
which in some cases caused attack and defense points to be
interchanged. This may have caused some stupid moves to be tried
during the owl reading.

Compared to the latest reported results in BREAKAGE, this patch has
the following effects:



Of the failures, I only had arb:200 and nicklas2:102 before updating
with recent changes in CVS. Both these are okay. I haven't checked yet
what's up with the four other failures but I won't delay this patch
any more now.

I've added the patch to CVS.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]