[Top][All Lists]

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

[gnugo-devel] eye valuation patch

From: Gunnar Farneback
Subject: [gnugo-devel] eye valuation patch
Date: Sun, 15 Sep 2002 20:43:41 +0200
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 huge patch at

This patch changes the representation of eye values, according to the
scheme described in

- life code completely disabled
- global variables life and life_eyesize removed
- options --life and --life-eyesize removed
- new eye value scheme; all code handling eye values revised
- new functions set_eyevalue(), min_eyes(), max_eyes(),
  add_eyevalues(), eye_move_urgency(), eyevalue_to_string()
- eyes.db revised for new eye value scheme and substantially extended

All support for the eye valuation code in life.c is removed with this
patch. It has long been unused and it would be way too much work, if
at all possible, to upgrade it to support the new eye value scheme.
The file life.c still remains but can safely be dropped at any time.
The options --life and --life-eyesize have naturally been removed too.

Obviously all code handling eye values has been revised. In many cases
this has been straightforward, e.g. changing value->mineye to
min_eyes(value). In other cases the changes are bigger and do change
the behaviour of the code. In particular this is true for
owl_determine_life(). There's some risk that the experimental semeai
code needs some further adjustments to this, but that shouldn't be too

The eye shape database eyes.db has been completely revised and
substantially extended. It should now be much more complete for eye
shapes up to size six.

The total regression delta compared to current CVS is 19 passes and 13
failures. The passes are:

owl: 229
owl_rot: 170
trevora: 520
nngs1: 46
neurogo: 13
lazarus: 6, 16
trevorb: 660
nicklas1: 1801
nicklas3: 401
trevorc: 940
vie: 50
trevord: 120
nngs2: 40, 240
auto01: 1, 2, 3
auto04: 2

and the failures:

trevora: 530, 550
nngs1: 26
rosebud: 1
strategy2: 60
nicklas5: 1212
nngs: 1700
global: 14, 32
arend: 29
trevord: 1070
strategy5: 232
safety: 5

While the delta is positive it's not exactly great. Generally speaking
the eye space analysis is on the whole not good enough to take full
advantage of the new eye value scheme. But the potential for further
improvements is good.

This patch will most likely clash badly with some other people's
changes. I'm sorry about that but I think we had better get this in
now. It has been long in preparing.


reply via email to

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