Re: [gnugo-devel] GNU Go 3.1.22

[Trevor Morris]

http://www.public32.com/regress/2002-01-20/?unexpected=1

This shows 9 FAILS & 6 PASSES, so we still have platform
specific issues.  Perhaps one of the new defaults isn't
correctly configured under VC?  It did run the regressions
much faster (about 3 hours; used to take 4).

-Trevor


At 07:54 AM 1/20/2002 -0800, Daniel Bump wrote:
>
>GNU Go 3.1.22 is now available at:
>
>http://www.gnu.org/software/gnugo/devel.html
>
>This version will be faster because computation of owl threats
>is turned off by default. There are three compile time options
>worth mentioning
>
>--enable-owl-threats             compute owl threats
>--enable-experimental-semeai     use new semeai module
>--enable-experimental-influence  use Arend's experimental influence
>
>The first two options should result in more PASS'd tests,
>and presumably make the program slower at the expense of speed.
>I hope the experimental semeai can become the default after
>Zobrist hashing gets implemented. I'm working on this.
>
>There are a *lot* of changes to test results, 118 unexpected PASS
>and 63 unexpected FAIL. Much of this is due to the patch gunnar_1_22.4.
>
>Patches in this release:
>
>arend_1_22.4. Experimental influence. 
>trevor_1_22.5. Tuning. 
>owl_1_threats_22.1. Owl threat computation made optional. 
>arend_1_22.2. Tuning etc. 
>gunnar_1_22.6. Filllib improvement. 
>dfa_1_22.1. Makes dfa the default. 
>trevor_1_22.3. Test case revisions. 
>trevor_1_22.4. Various changes. 
>komaster_1_22.1. Bugfix. 
>teun_1_22.1. Rot revision. 
>arend_1_22.3. New NNGS tests. 
>gunnar_1_22.5. Tuning and new connection tests. 
>gunnar_1_22.4. Atari_atari revisions. 
>gunnar_1_22.3. Optics improved. 
>arend_1_22.1. Redundant move reasons. 
>trevor_1_22.1. engine.dsp. 
>gunnar_1_22.2. Various small changes. 
>gunnar_1_22.1. Revision of lazarus.tst. 
>dfa_1_22.1. Increased DFA_MAX_MATCHED. 
>sgf_1_22.1. Bugfix in sgftreeAddStone. 
>
>- find_backfilling_move() in filllib.c revised
>- dfa made standard
>- owl threat computation made optional (and not default)
>- comments added, cleanup around discard_redundant_move_reasons
>- new (unused) experimental influence configure and run time options
>- unsafe tactical moves now taken care by discard_redundant_move_reasons
>- FIXME added for reevaluate_ko_threats
>- atari_atari_attack_patterns() uses a goal array to limit search area
>- new functions compute_aa_values(), get_aa_value(), and update_aa_goal()
>- ko contingent defenses excluded in get_saved_worms(), get_saved_dragons()
>- valuation of atari_atari move reasons revised
>- recognize_eye() revised
>- spelling corrections
>- fflush() added a few places
>- owl tries tactical defense for small groups
>- owl logs number of variations per sub-tree in SGF file
>- [ox]play_(attack_either|defend_both) handles ko correctly.
>- added [ox]play_connect and dis_connect helpers
>- allow followup value if defense threat is contingent on ko.
>- only allow territory move values if move is deemed safe.
>- reduce cutstone values in case of ko.
>- debug printing in influence.c slightly restructured
>- new function influence_moyo_color_opposite()
>- one trace output reenabled in confirm_safety()
>- bugfix in sgftreeAddStone
>- bugfix in komaster_trymove
>- mascot in dragon.c
>- tuning
>- eye tuning
>- atari_atari tuning
>- connection tuning
>- fuseki tuning
>- owl tuning
>- test revisions
>
>Dan
>
>_______________________________________________
>gnugo-devel mailing list
>address@hidden
>http://mail.gnu.org/mailman/listinfo/gnugo-devel
>