Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old October 15th 05, 08:56 PM
Mr. Wizard
 
Posts: n/a
Default Disdributed Computing for Chess

Does anyone know if there are any websites with information on using
disdributed computing for chess. That is, is it possible to have many
computers analyzing different lines of a chess game to come up with the
best move.

  #2   Report Post  
Old October 16th 05, 10:20 AM
Guy Macon
 
Posts: n/a
Default Distributed Computing for Chess



Mr. Wizard wrote:

Does anyone know if there are any websites with information on using
distributed computing for chess.


http://www.chessbrain.net/
http://www.frayn.net/beowulf/

--
Guy Macon http://www.guymacon.com/

  #3   Report Post  
Old October 16th 05, 05:35 PM
Gian-Carlo Pascutto
 
Posts: n/a
Default Distributed Computing for Chess

Guy Macon wrote:
Mr. Wizard wrote:


Does anyone know if there are any websites with information on using
distributed computing for chess.



http://www.chessbrain.net/
http://www.frayn.net/beowulf/


No actual "information" on those sites, only marketing-like hype.

--
GCP
  #4   Report Post  
Old October 16th 05, 05:35 PM
Gian-Carlo Pascutto
 
Posts: n/a
Default Distributed Computing for Chess

Guy Macon wrote:
Mr. Wizard wrote:


Does anyone know if there are any websites with information on using
distributed computing for chess.



http://www.chessbrain.net/
http://www.frayn.net/beowulf/


No actual "information" on those sites, only marketing-like hype.

--
GCP
  #5   Report Post  
Old October 16th 05, 09:01 PM
Guy Macon
 
Posts: n/a
Default Distributed Computing for Chess




Gian-Carlo Pascutto wrote:

Guy Macon http://www.guymacon.com/ wrote:

Mr. Wizard wrote:

Does anyone know if there are any websites with information
on using distributed computing for chess.


http://www.chessbrain.net/
http://www.frayn.net/beowulf/


No actual "information" on those sites, only marketing-like hype.


Only if you are too stupid and/or lazy to actually read them.

The sites referenced contain full source code for a distributed
chess engine, as well as white papers such as _The ChessBrain
Project - Massively Distributed Inhomogeneous Speed-Critical
Computation_, which was presented Dr. Colin Frayn at the June
30th 2004 International Conference On Scientific and Engineering
Computation in Singapore.

Note: Mr. Pascutto is the author of Deep Sjeng, a commercial
chess engine that he sells for 35 euros (roughly 42 US dollars).
Here is where his program ranks on the 2005-07-29 SSDF list:
(Tested on 256MB Athlon 1200 MHz PCs)

1 Shredder 9.0 UCI 2821
2 Shredder 8.0 CB 2805
3 Shredder 7.04 UCI 2804
4 Junior 9.0 2789
5 Deep Fritz 8.0 2783
6 Junior 8.0 2766
7 Shredder 7.0 2765
8 Deep Fritz 7.0 2764
9 Fritz 8.0 2753
10 Deep Junior 8.0 2750
11 Gandalf 6.0 2742
12 Fritz 7.0 2739
13 Hiarcs 9.0 2736
14 Pro Deo 1.1 2723
15 Shredder 6.0P UCI 2721
15 Chess Tiger 2004 2721
17 Chess Tiger 14.0CB 2718
18 Chess Tiger 15.0 2717
19 Deep Fritz 2715
19 Chessmaster 9000 2715
21 Shredder 6.0 2714
22 Gambit Tiger 2.0 2713
23 Junior 7.0 2700
24 Rebel 12.0 2682
25 Hiarcs 8.0 2680
26 Ruffian 1.0.1 2674
27 Rebel Century 4.0 2673
28 Deep Sjeng 1.5a 2671



  #6   Report Post  
Old October 16th 05, 09:15 PM
Gian-Carlo Pascutto
 
Posts: n/a
Default Distributed Computing for Chess

Guy Macon wrote:

Only if you are too stupid and/or lazy to actually read them.

The sites referenced contain full source code for a distributed
chess engine, as well as white papers such as _The ChessBrain
Project - Massively Distributed Inhomogeneous Speed-Critical
Computation_, which was presented Dr. Colin Frayn at the June
30th 2004 International Conference On Scientific and Engineering
Computation in Singapore.


The same "paper" was in the ICGA journal, and believe me, I read them
attentively as I have great interest in this research. Your remark and
attempted snipe shows poor class and a great lack of understanding.

I have looked in vain for a detailed evaluation of the actual parallel
performance. And that means actual speedup numbers, not just the total
node speeds or even effective node speeds. How much faster is the
program actually reaching the same depth. This is the key matter,
unfortunately hard to measure accurately and there is very few published
about this in literature. I guess that has a lot to do with those
numbers usually looking a bit less impressive than total node speeds.
But it is *the* key *relevant* performance metric for parallel
chessprogram performance.

As for your poor snipe, I'll remark that Deep Sjeng finished third in
this years World Championship, and that enities such as ChessBrain are
always notably absent from the competition. (Or from the SSDF list, for
that matter :-P)

--
GCP
  #7   Report Post  
Old October 17th 05, 07:29 AM
HD
 
Posts: n/a
Default Distributed Computing for Chess

Gian-Carlo Pascutto wrote:
Guy Macon wrote:


Only if you are too stupid and/or lazy to actually read them.

The sites referenced contain full source code for a distributed
chess engine, as well as white papers such as _The ChessBrain
Project - Massively Distributed Inhomogeneous Speed-Critical
Computation_, which was presented Dr. Colin Frayn at the June
30th 2004 International Conference On Scientific and Engineering
Computation in Singapore.



The same "paper" was in the ICGA journal, and believe me, I read them
attentively as I have great interest in this research. Your remark and
attempted snipe shows poor class and a great lack of understanding.

I have looked in vain for a detailed evaluation of the actual parallel
performance. And that means actual speedup numbers, not just the total
node speeds or even effective node speeds. How much faster is the
program actually reaching the same depth. This is the key matter,
unfortunately hard to measure accurately and there is very few published
about this in literature. I guess that has a lot to do with those
numbers usually looking a bit less impressive than total node speeds.
But it is *the* key *relevant* performance metric for parallel
chessprogram performance.

As for your poor snipe, I'll remark that Deep Sjeng finished third in
this years World Championship, and that enities such as ChessBrain are
always notably absent from the competition. (Or from the SSDF list, for
that matter :-P)


Apart from the "thing" being very much under developement, how would you
suggest the SSDF testers to test it??
It's aimed at the use of distributed - volunteer - computers appantly on
the internet with (very) various numbers of contributers, not a AMD 1200
box in a SSDF testers house. It can hardly be compared to an ordinary
chess engine.

HD
  #8   Report Post  
Old October 17th 05, 04:29 PM
Guy Macon
 
Posts: n/a
Default Distributed Computing for Chess



Gian-Carlo Pascutto wrote:

I have looked in vain for a detailed evaluation of the actual parallel
performance. And that means actual speedup numbers, not just the total
node speeds or even effective node speeds. How much faster is the
program actually reaching the same depth. This is the key matter,


There is a large difference between not having the specific information
you list above and your false claim of "No actual 'information' on those
sites, only marketing-like hype." Would you like to retract that false
statement now? As I pointed out before, a web site that has the source
code for a distributed chess engine as well as informative articles such
as the one at http://www.frayn.net/beowulf/theory.html can hardly be
characterized as having "No actual 'information' ... only marketing-like
hype."

I'll remark that Deep Sjeng finished third


Tied for third, if I remember correctly.

in this years World Championship, and that enities such as
ChessBrain are always notably absent from the competition.


It is unclear whether Chessbrain as it exists today would be
allowed under the current rules. How do you insure that a
program using thousands of computers on the Internet is not
receiving help from a human?

Or from the SSDF list, for that matter


I *know* that Chessbrain as it exists today is not allowed by
the SSDF rules. The whole point of the SSDF is to pit program
against program on equal hardware. On equal (single CPU)
hardware, Deep Sjeng came in at 28th place, behind Shredder,
Junior, Deep Fritz, Fritz, Deep Junior, Gandalf, Hiarcs, Pro Deo,
Chess Tiger, Chessmaster 9000, Gambit Tiger, Junior, Rebel, Hiarcs,
Ruffian, and Rebel Century. Clearly you are at a disadvatage on a
single CPU system. This is nothing to be ashaded about; many of the
programs that beat you can't even use a second processor, and it takes
more skill to write an efficient multithreaded program.

The WCCC, on the other hand, compares that subset of chess engines
that have enough financial backing to be able to go to Reykjavik,
running on the best and fastest computers they can afford, including
custom hardware (not that doing that has helped in recent years).

Under those conditions Deep Sjeng tied Shredder for 3rd place, behind
Zappa and Fruit. I believe (correct me if I am wrong) that Zappa and
Deep Sjeng ran on an AMD64 Dual Core at 2.2MHz, Fruit ran on a single
AMD64 at 2.4MHz, and Shredder ran on quad AMD64s at 2.6MHz.

SSDF and WCCC measure different things.

Your comment that "entities such as ChessBrain are always notably
absent from the competition" seems to imply that you think Deep
Sjeng can beat ChessBrain (if so, I tend to agree; I don't think
that any distributed chess engine has solved the intercommunication
problem well enough to beat a top-tier PC program). Why not challenge
it to a match and prove it?

--
Guy Macon http://www.guymacon.com/

  #9   Report Post  
Old October 17th 05, 09:39 PM
HD
 
Posts: n/a
Default Distributed Computing for Chess

Guy Macon wrote:
Gian-Carlo Pascutto wrote:


I have looked in vain for a detailed evaluation of the actual parallel
performance. And that means actual speedup numbers, not just the total
node speeds or even effective node speeds. How much faster is the
program actually reaching the same depth. This is the key matter,



There is a large difference between not having the specific information
you list above and your false claim of "No actual 'information' on those
sites, only marketing-like hype." Would you like to retract that false
statement now? As I pointed out before, a web site that has the source
code for a distributed chess engine as well as informative articles such
as the one at http://www.frayn.net/beowulf/theory.html can hardly be
characterized as having "No actual 'information' ... only marketing-like
hype."


I'll remark that Deep Sjeng finished third



Tied for third, if I remember correctly.


in this years World Championship, and that enities such as
ChessBrain are always notably absent from the competition.



It is unclear whether Chessbrain as it exists today would be
allowed under the current rules. How do you insure that a
program using thousands of computers on the Internet is not
receiving help from a human?


Or from the SSDF list, for that matter



I *know* that Chessbrain as it exists today is not allowed by
the SSDF rules. The whole point of the SSDF is to pit program
against program on equal hardware. On equal (single CPU)
hardware, Deep Sjeng came in at 28th place, behind Shredder,
Junior, Deep Fritz, Fritz, Deep Junior, Gandalf, Hiarcs, Pro Deo,
Chess Tiger, Chessmaster 9000, Gambit Tiger, Junior, Rebel, Hiarcs,
Ruffian, and Rebel Century. Clearly you are at a disadvatage on a
single CPU system. This is nothing to be ashaded about; many of the
programs that beat you can't even use a second processor, and it takes
more skill to write an efficient multithreaded program.

The WCCC, on the other hand, compares that subset of chess engines
that have enough financial backing to be able to go to Reykjavik,
running on the best and fastest computers they can afford, including
custom hardware (not that doing that has helped in recent years).

Under those conditions Deep Sjeng tied Shredder for 3rd place, behind
Zappa and Fruit. I believe (correct me if I am wrong) that Zappa and
Deep Sjeng ran on an AMD64 Dual Core at 2.2MHz, Fruit ran on a single
AMD64 at 2.4MHz, and Shredder ran on quad AMD64s at 2.6MHz.

SSDF and WCCC measure different things.

Your comment that "entities such as ChessBrain are always notably
absent from the competition" seems to imply that you think Deep
Sjeng can beat ChessBrain (if so, I tend to agree; I don't think
that any distributed chess engine has solved the intercommunication
problem well enough to beat a top-tier PC program). Why not challenge
it to a match and prove it?

In all aspects, a well functioning Chessbrain is more like Hydra - but
even more diffuse. It doesn't really compare to chessengines!

  #10   Report Post  
Old October 19th 05, 05:25 PM
Guy Macon
 
Posts: n/a
Default Distributed Computing for Chess



Guy Macon wrote:

...your false claim of "No actual 'information' on those sites,
only marketing-like hype." Would you like to retract that false
statement now? As I pointed out before, a web site that has the
source code for a distributed chess engine as well as informative
articles such as the one at http://www.frayn.net/beowulf/theory.html
can hardly be characterized as having "No actual 'information'
... only marketing-like hype."


I will take Gian-Carlo Pascutto's non-response to the above
as an admission that his claim was indeed a false claim.

Your comment that "entities such as ChessBrain are always notably
absent from the competition" seems to imply that you think Deep
Sjeng can beat ChessBrain (if so, I tend to agree; I don't think
that any distributed chess engine has solved the intercommunication
problem well enough to beat a top-tier PC program). Why not challenge
it to a match and prove it?


I am not sure how to take Gian-Carlo Pascutto's non-response to
the above. Does he think that Chessbrain can beat Deep Sjeng?
I really think that Deep Sjeng would come out on top of such a
match; distributed chess computers are a lot less refined than
top-tier PC chess programs are.

Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT +1. The time now is 05:42 AM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017