[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] Good news
From: |
Tanguy URVOY |
Subject: |
[gnugo-devel] Good news |
Date: |
Fri, 31 Jan 2003 16:43:15 +0100 |
User-agent: |
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.2) Gecko/20021120 Netscape/7.01 |
Hello,
I made some experiments with gnugo.
Instead of building a fully incremental matcher
(I am too lasy and busy for that),
I tryed to see what would occur if we only try to 'cache'
the patterns.
The principle is to store the address of the last reached
state : if it does not change we do not need to rescan
the board.
A shared cache would gives a 40% hits score.
Individual caches seems more promissing with
a mean of 95% hits !!!
Such a pattern cache would reduce
dramatically the time wasted in
pattern matching.
semeai[16:30]%make owl
./eval.sh owl.tst
owl_defendpat cache hits: 90.3% of 37112
owl_attackpat cache hits: 90.0% of 30344
owl_defendpat cache hits: 95.9% of 96960
owl_attackpat cache hits: 94.9% of 66288
owl_defendpat cache hits: 97.9% of 15144
owl_attackpat cache hits: 97.9% of 9432
A change to 'dfa_pos' broke my patch :(
But you can replace matchpat.c by my version
and it will work fine.
http://www.irisa.fr/prive/Tanguy.Urvoy/gnugo/stat_matchpat.c
Bye
--
---------------------------------------
Tanguy Urvoy http://www.irisa.fr/galion
---------------------------------------
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [gnugo-devel] Good news,
Tanguy URVOY <=