Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Current Crafty strength on SMP?

13 views
Skip to first unread message

Charlton Harrison

unread,
Apr 29, 1998, 3:00:00 AM4/29/98
to

Does anyone have any estimate/guestimate on how strong the current
Crafty v15 is? How about on an SMP machine?
...or does anyone have an idea of how it fairs agains the most recent
programs
like Rebel, Fritz, etc?


Charlton


bruce moreland

unread,
Apr 29, 1998, 3:00:00 AM4/29/98
to

The real "Crafty" account isn't doing that great, I think.

But "Moron", which is a dual P2/300, is a monster for some reason.

My program (running on 533 mhz Alpha) usually wins blitz and bullet
matches with Crafty accounts, but against Moron it has been lucky to
draw the matches.

I can't compare Crafty to commercial programs, but I think I'd rather
play most anything at 5 0 rather than that dual P2/300.

bruce


Robert Hyatt

unread,
Apr 29, 1998, 3:00:00 AM4/29/98
to

Charlton Harrison <char...@dynet.com> wrote:
: Does anyone have any estimate/guestimate on how strong the current
: Crafty v15 is? How about on an SMP machine?
: ...or does anyone have an idea of how it fairs agains the most recent
: programs
: like Rebel, Fritz, etc?


: Charlton

This isn't a good question. IE Crafty on a quad-pentium-pro as significantly
faster than a commercial program on the same machine. A 3x speed handicap is
not just a small advantage...

Main question then, is whether you like how it plays, and whether you want
a "research program" or a "commercial program". Both have advantages. You
likely won't see a commercial program using multiprocessing, since NT is the
only platform they could use and it won't work on Dos/win95. But with a
commercial program you get a much more polished interface with many more
features. Either can beat 99+ percent of all the players in the world...

--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170

Robert Hyatt

unread,
Apr 29, 1998, 3:00:00 AM4/29/98
to

bruce moreland <bru...@seanet.com> wrote:
: On Wed, 29 Apr 1998 01:18:14 -0500, Charlton Harrison
: <char...@dynet.com> wrote:

:>Does anyone have any estimate/guestimate on how strong the current
:>Crafty v15 is? How about on an SMP machine?
:>...or does anyone have an idea of how it fairs agains the most recent
:>programs
:>like Rebel, Fritz, etc?

: The real "Crafty" account isn't doing that great, I think.

: But "Moron", which is a dual P2/300, is a monster for some reason.

: My program (running on 533 mhz Alpha) usually wins blitz and bullet
: matches with Crafty accounts, but against Moron it has been lucky to
: draw the matches.

: I can't compare Crafty to commercial programs, but I think I'd rather
: play most anything at 5 0 rather than that dual P2/300.

: bruce

Most of "crafty"'s problems are "mine". IE I have a few glitches in the
SMP code, and I'm debugging. I flag about 1 game every 30 due to my debug
code crashing or hanging. Which means (with crafty currently rated at
around 2850 in blitz) that every 30 games, I will drop 30 rating points or
so, and its unlikely that I'll pick them back up over the next 29 games
before I crash again...

Roman commented on Moron when we talked last week. I noticed that "scrappy"
was also a bear at fast games on the dual PII/300 I had for a while. One
important note is that 2 processors come closer to being 2x faster than do
4 processors come close to being 4x faster.

But my rating has always been "volatile" since I do 99% of my debugging on
ICC. Moron is running "semi-stable" code (version 15.3 I believe) which, as
we both know, goes a *long* way toward producing a high rating. Any glitches
and down it comes.

I'm getting closer to "sanity" on ICC... but for the past two weeks I haven't
had much time to work on this stuff. And the problems are complicated enough
that 30 minutes here and there don't help...

You should look at the latest source however... it is *damned clever* in the
way it avoids requiring massive modifications to the engine... although I am
not sure my comments in thread.c are enough to explain the dynamic way that
threads are *never* idle and work together... Complicated in principle, but
quite simple in terms of code... which is why it was running so quickly when
compared with Cray Blitz...

0 new messages