[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] dfa optimizer
From: |
Paul Pogonyshev |
Subject: |
Re: [gnugo-devel] dfa optimizer |
Date: |
Sat, 8 Feb 2003 19:23:36 +0200 |
User-agent: |
KMail/1.4.3 |
Dan wrote:
> The patch paul_3_17.2a is now in the CVS. This means you need to run
> configure before recompiling.
i wrote:
> i haven't measured speed, but i'm quite certain that if it has changed,
> then to the better side ;)
here are several first strings of two profiles (nngs2.tst):
non-optimized dfas:
Each sample counts as 0.00195312 seconds.
% cumulative self self total
time seconds seconds calls s/call s/call name
12.00 51.01 51.01 134909984 0.00 0.00 scan_for_patterns
4.90 71.85 20.83 163184617 0.00 0.00 fastlib
3.62 87.22 15.37 61184027 0.00 0.00 do_play_move
optimized dfas:
Each sample counts as 0.00195312 seconds.
% cumulative self self total
time seconds seconds calls s/call s/call name
9.59 39.06 39.06 134838784 0.00 0.00 scan_for_patterns
5.11 59.88 20.82 163184636 0.00 0.00 fastlib
3.86 75.60 15.71 61182242 0.00 0.00 do_play_move
almost 2.5% of speed gain. the total speedup on nngs2.tst seems to
be more than 2.5%, but that must be caused by random change in owl
and reading nodes. on some test files the change may be positive,
on others it is negative. it is caused by the fact that after the
optimization patterns are sorted in different order in some cases.
Paul