gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] regress 13x13b.17


From: kevin yong
Subject: [gnugo-devel] regress 13x13b.17
Date: Sun, 6 Apr 2003 12:07:29 -0400 (EDT)

Hi, Evan:

  sorry i sent you a in-complete version of this
message before.

  i did some work on 13x13b.17:

  problem: H5 is wrong, should F1.

  i look at the log file: H1 got value 1.54, i dont
think anything wrong with that; F1 got value 0, that
is the problem.

  how F1 got value 0:

  because at examine_move_safety() stage, identify F1
as 'unsafe' move, but the result is from old chched
result;

  then i trace back to owl_reasons() stage,

    owl_does_defend A4 with F1
    Current stack: B:F1 W:B1 B:C1 
    Variation -1: DEAD 1 (cached)
  
  and the cached result dump is:
    
    Komaster 0 (0, 7) Routine 6, (-1, 15), depth:
      16 Result: 0 12, (39, 18)

  automataly, i can not figure out when the original
cached result was saved.

  somewhere at early stage, it was found that A4
failed to defend, and the result cached. later on,
when value F1, the old result was retrieved and failed
F1.

  i also used start_sgftrace/finish_sgftrace trace.sgf

as you suggested. in this case, i got a 400+k (221
pages in MS word) .sgf file. (for case 13x13.6, i got
a 29mg .sgf.)

  inside this trace.sgf, there are so many variations
around A4,D1,E1,F1. All info we need is:

  1. at which step F1 was ORIGINALLY identified as
fail and cached.
  2. at which step this cached result was retrieved (i
think i already know this)
  
  but i have difficult to find answer for Qustion-1.

  i did read reading.texi doc and source code of
cache.h & cache.c, and source code in owl.c, it seems
the result was stored by MACROs READ_RETURNs, i insert
some 'gprint' into these macros, bu i did not get any
print out.

  another question is: has this cached result been
correctly used at the later stages?

  could you help me to give me some indication how to
trace where the reading result was cashed? and how to
effectively understand the trace.sgf file?

  thank you very much for you help !

kevin.



  

______________________________________________________________________ 
Post your free ad now! http://personals.yahoo.ca




reply via email to

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