Hardware limits in the WCCC

[Rémi Coulom]

Hi,

This topic is for open discussion of hardware limits in the World Computer Chess Championship.

For reference, I attach the letters of David Levy to this message:

• Times.they.are.a.changing.pdf

  (40.2 KiB) Downloaded 639 times

• 8 Cores.pdf

  (58.16 KiB) Downloaded 635 times

• Third message of David Levy, inviting discussion:

  The ICGA would like to invite all chess programmers and members of their
  teams to discuss the "8-cores" issue in the ICGA forum.
  
  The URL is http://www.grappa.univ-lille3.fr/icga/phpBB3/
  
  Rémi Coulom, who is the Programmers' Representative on the ICGA executive,
  has started the discussion with the following topic:
  
  http://www.grappa.univ-lille3.fr/icga/p ... ?f=15&t=66
  
  We hope for a lively exchange of views on the 8-cores issue, preparing the
  way for a face-to-face meeting of all interested parties during the 2009
  World Computer Chess Championship (May 11th-18th).
  
  In regard to the precise wording of the rules for 2009, the ICGA would
  encourage you to discuss your thoughts on the details of the rule. This is
  NOT an invitation to support or oppose the idea of 8-cores for 2009, since
  that decision has alreday been made. It is invitation to help refine the
  rule in a pracical and fair manner.
  
  David Levy

I plan to edit this message as discussion progresses in order to summarize arguments and points of view:

• Are your for or against the hardware limit ?
• How can the limit be enforced in practice ?
• Should remote play be allowed ?
• What change in the WCCC can bring more participants ?

I am proposing to organize this discussion in my role of programmers representative for the ICGA. The opinion of non-programmers is welcome too, but please, if you are not a programmer, say so in your reply. If you are a programmer, but not a member of the ICGA, then your opinion is very valuable, since one of the objectives of this discussion is to attract as many new participants as possible.

Rémi

Summary

Programmers who wish to be counted as being openly against the hardware limit (either by saying it here, or by private e-mail to me):

• Vasik Rajlich
• Vincent Diepeveen
• Gian-Carlo Pascutto
• Zach Wegner
• Pradu
• Munjong Kolss
• Richard Pijl
• Gerd Isenberg

Programmers who wish to be counted as being openly in favor of the hardware limit (either by saying it here, or by private e-mail to me):

• Mark Uniacke

Arguments against the hardware limit

• It has been a long tradition in the WCCC to have no limit, since its first edition in 1974. With hardware limits, the tournament should be called WMCCC.
• The history of the WCCC is full of victories of microcomputers over supercomputers. It is extremely difficult to actually take advantage of non-conventional computing power.
• An 8-core limit requires hardware inspection. It cannot be done with remote play.
• It won't make hardware even anyways. Top over-clocked 8-core machines cost more than small clusters with more than 8 cores. The only solution for fair hardware would be uniform platform provided by the organizers.
• Announcing a change in the rules 5 months before the tournament is too late, especially since some participants have already started to invest into making a cluster version of their programs.
• Some spectators find it more interesting to watch a tournament with non-conventional hardware
• It makes it impossible for interesting machines such as FPGA (Hydra) to participate

Arguments in favor of the hadware limit:

• One should not be able to buy the title with superior hardware.
• Chess programs are already stronger than the strongest humans.
• Some spectators find it more interesting to watch a tournament on even hardware
• Customers of commercial software want to see the program run on the machine they have at home

Proposals for running a tournament with a hardware limit

• Trust participants with remote hardware.
• Force local play, and hardware inspection.
• What about FPGA, graphic cards, hyperthreading, Playstations ?

Proposals for attracting more participants:

• A shorter tournament lasting a few days. Some suggest more rounds per day, others prefer slow games with less rounds.
• Live internet coverage of the games
• No entry fees, or at least the same entry fee for commercial/amateur competitors.
• Some tournaments held in North America
• More travel incentives as previously used before, but announced well in advance in order for people to make arrangements.
• Hold the tournament during the summer holidays
• Not give incentives to operators in order to encourage the programmers to show up.

Rémi

[zwegner]

First off, thanks Remi for creating this forum. I hope all concerned programmers stop by to give their opinions.

are your for or against the hardware limit ?

Strongly against.

How can the limit be enforced in practice ?

The only real way is to have all hardware on-site, and not allow any internet connections. This is particularly bad for those who live far away, and they would have to ship their hardware.

Should remote play be allowed ?

For the 8-core rule to be taken seriously at all, remote play cannot be allowed. However, this has many other negative consequences.

What change in the WCCC can bring more participants ?

• No hardware limits.
• No entry fees, or at least the same entry fee for commercial/amateur competitors.
• A shorter tournament lasting a few days, with more rounds per day.
• Some tournaments held in North America
• More travel incentives as previously used before, but announced well in advance in order for people to make arrangements.
• Related to the above is a point brought up by Vincent Diepeveen, which is to not give incentives to operators in order to encourage the programmers to show up.

[Pradu]

are your for or against the hardware limit?

Against. Writing a good shared-memory search and/or a cluster search is very difficult and effort in this area should be rewarded for good speedup on a large number of cores. It is not really clear how much programs like really Rybka benefit from just adding more computers (from the PVs produced in the last ACCA tournament... it might not be very much at all). Rybka has a good chance and perhaps even a better chance to win with a hardware limit. Even when applying a hardware limit there are different levels of hardware components that meet the 8-core hardware limit (e.g. "Extreme or Black Edition" of a processor and a low-end version that costs perhaps less than a third of the high-end version). If a hardware limit is ever to be enforced, it should be done with uniform hardware, but this again limits innovation on non-standard hardware.

How can the limit be enforced in practice?

-

Should remote play be allowed?

Yes. It allows the use of hard-to-move hardware.

What change in the WCCC can bring more participants?

A tournament closer to the majority of chess programmers should bring more participants. I like Zach's idea of having a tournament in North America. I know that there are a lot of chess programmers in Europe as well. After giving people a "first-time experience" I'm sure that they'll be more willing to join the tournaments further away. I myself haven't attended an ICGA event yet and attended my first computer chess event last year but I hope I will be able to attend more in the future.

As for weight of opinions, people who have attended previous ICGA tournaments should have greater weight attached to their opinions.

[Rémi Coulom]

As for weight of opinions, people who have attended previous ICGA tournaments should have greater weight attached to their opinions.

Yes, maybe. Right now, we are not organizing a poll in order to make a decision. So there is no vote-counting, just open discussion.

Also, I think the main question is not about hardware limits. The main question is: how can we attract more participants in the WCCC ? So, the opinion of programmers who have not yet participated is very valuable. Thanks for expressing yours.

Rémi

[Computer Chess Fan.]

Programmers against the 8-core hardware limit

• Zach Wegner (ZCT)
• Pradu (Buzz)

Programmers in favor of the 8-core hardware limit: none so far.

I will summarize ideas later, when more programmers participate in the discussion.

You are a member of CCC so it's easy to see that most other programmers are _clearly_ against the hardware limitation and also which are them. So you should put them also. I'm not aware of any other programmer that it is in favor of the rule.
I don't think that posting here makes their opinion stronger in order to put them on a category. If you want i can provide links with their opinion.

My personal opinion is that i believe the rule change is _completely_ stupid for several reasons and i would not bother to explain any as i believe ICGA also believes this. If this rule change survives and be applied to WCCC i(at least) will not watch it at all.
Better look at CEGT games where they at least play 800-1000 games.

The main reasons i was seeing WCCCs was to see new researches and methods in Computer Chess, very big hardware(bigger than usual) with great moves, new approaches etc. Now they remove some of these. So they make it uninteresting for me.

And one question i would like an ICGA official to answer: If Hydra asked for participation what would be the answer? Yes or No?

[Rémi Coulom]

You are a member of CCC so it's easy to see that most other programmers are _clearly_ against the hardware limitation and also which are them. So you should put them also. I'm not aware of any other programmer that it is in favor of the rule.
I don't think that posting here makes their opinion stronger in order to put them on a category. If you want i can provide links with their opinion.

I considered doing that in the beginning, but I refrained.

My idea in creating this forum is that the ICGA needs an official online ICGA place for communicating about these issues. Programmers can rant as much as they wish in all the forums they want, it will have no effect because ICGA officials don't read them. David Levy and Jaap van den Herik are a little disconnected from the programmer community. It has been a complaint of programmers for a very long time. It is for this reason that the position of ICGA programmers representative was created in 1997. I don't expect that David Levy and Jaap van den Herik will participate in this forum either, but if many programmers participate in the discussion, and we come up with a summary of the discussion that every programmer agrees with, then as a programmers representative, I can communicate this summary to David Levy and hope that he will take it into consideration.

I don't want to include in this summary opinions that were expressed elsewhere and that can be found by searching the web. If I did so, I would risk missing opinions that were expressed somewhere and that I was not aware of. That would be unfair.

I have read many messages in the CCC and in other forums. So far, I am not aware of a single chess programmer that is in favor of the rule. In his letter, David Levy mentions that, according to the emails he received, "exactly half are in favour of the idea and half are against". I would like to encourage programmers who are in favour of the idea to speak up.

An official email will be sent very soon to ICGA members, so I hope that many programmers will join here to discuss these important issues.

Rémi

[MjKolss]

Hello Rémi, hello all,

as a co-programmer of Ikarus and three-time participant in WCCC tournaments, I would like to express my humble opinion. I have also e-mailed David Levy upon his earlier request to do so. Many of my views match those of other programmers in this and the CCC forum, so I will try and keep it short...

Are your for or against the hardware limit ?

I am against a hardware limit. Such a limit will most likely not significantly change the prospects of any entering program, in particular not those of amateurs. There are thus no potential benefits, and these zero benefits have to be weighed against the significant costs of reduced interest by media and spectators and the discouragement of trying out new approaches / solutions to hardware usage (and in fact excluding several potential participants!). The balance turns out negative.
By the way, if "the ICGA" feels the need to have a tournament with limited hardware, why not at least take the newly installed chess section from the olympiad, while keeping the WCCC as it is and should be?!

How can the limit be enforced in practice ?

I agree with Zach that this is going to be difficult. The only way would be to force participants to have their hardware on site. That will make it especially deterrent to amateurs. I remember carrying along (in the train) a (quad) server for the 2006 tournament, and I can assure you that that was very little fun and that I am not going to do that again! Which sponsor is going to give me two servers, one for testing at home in advance and one just for the tournament on site?!

Should remote play be allowed ?

By all means, yes! For the reasoning, see above. I am not going to carry with me a server, and there are probably not that many sponsors around that would be willing to ship a server to the tournament site for a wimpy amateur program.
Of course, if a hardware limit were to be enforced, remote play must be disallowed, as otherwise cheating is impossible to prevent.

What change in the WCCC can bring more participants ?[/list]

This is a good question, one that maybe should have been asked long ago!
Here are my views, in descending order of importance:

1. Make the tournament shorter - if you are interested in amateurs' participation. It is very, very difficult to take more than five precious days of holidays per year for such a tournament, let alone convince the family that this is the best way to spend those holidays. It should absolutely be possible to finish the tournament within a week (plus weekends!). E.g., travel on Saturday, setup and first round on Sunday, two rounds each on Monday, Tuesday, Wednesday, one round and the speed tournament on Thursday, last (ninth) round plus possible play-offs on Friday, travel on Saturday, or something similar. There is no need to start or end the tournament mid-week!

2. Announce the tournament well in advance. With "well in advance", I do NOT mean something like three or four months before the start. If a tournament is to take place say in August 2010, the announcement date should be end of 2009 at the very latest!! Some people - especially amateurs, if they are of any interest - need to plan their holidays well; some have to sign up for their holidays during the preceding calendar year. Many amateurs have commitments at work; they may have family to consider. If in January, a tournament is announced for May, it depends on pure luck whether I happen to be somewhat free during that time; the odds are low, however. I am sure I am not the only one. Besides, logistic problems such as booking flights and hotels or arranging hardware are slightly easier to solve with a little more time.

3. Travel and accomodation support as used for Beijing is (and would be) nice. I suppose it might be restricted to amateurs; professionals can get some of their expenses back through their tax declaration, amateurs cannot. But that depends on the sponsors, of course. At any rate, I strongly agree with Vincent and Zach that absolutely no money should be wasted on operators. The programmers should be present in person, or otherwise pay someone else themselves.

4. I can see that it is difficult to find sponsors at all who are willing to support such an event. It would be nice if more active attempts could be made to find one in North America for a change. That might attract a few more participants from the US and Canada, although I would not expect total expenses for me (as a European) to differ at all whether the event be in Europe or America. It is probably more a psychological issue (which should not be neglected, though!).

Well, you can see that many of my views are through an amateur's eyes. However, I am positive that if you want to attract more than the handful of professional chess programmers, amateurs should not be entirely neglected.

I hope that someone finds my thoughts worthy of consideration and that they may contribute to a constructive discussion (and decision).

Thank you for your efforts and thanks to everyone actually reading through all this for their time !

Best regards - Munjong.

[pijl]

[*] Are your for or against the hardware limit ?

Against. Noteworthy is that I'm not using more than 8 CPU's (just 4 in fact) and am not planning to do so in any tournament this year. Main arguments against the limit:
- The limit is arbitrary, excludes participants from (recent) previous editions. Their work on getting their program to work on 8+ CPU cores is void.
- Hardware is not the biggest difference between programs. The major difference is probably the program itself, followed by the book.
- Hard to enforce the rule without putting further restrictions on the participants

[*] How can the limit be enforced in practice ?

Probably only in one way: by having the equipment onsite, without direct connection to the internet.

[*] Should remote play be allowed ?

Not allowing remote play is a serious problem for some participants. Carrying a proper tournament machine to the tournament location can be a major undertaking when the location is not within driving distance.

[*] What change in the WCCC can bring more participants ?[/list]

A three to five day tournament (also depending on time control, which determines whether 2 or 3 rounds a day are possible) should be enough, which limits the expense for staying in hotels, but also the number of days off from work will be less.
The time control could be a little faster, allowing for 3 rounds a day. 3-4 hours per round should be enough.
Richard.

[Gerd Isenberg]

Hi,
great to have an ICGA Forum! Happy new year to all members and participants.

If it is about the ICGA WCCC, I am absolutely against any hardware limit.
I have sympathies for playing with hardware on-site (with my shuttle box but I agree that this is not a practicable option for ambitious participants.

I also favor a shorter event. Four or five days for seven rounds is appropriate, imo. Still, I prefer no more than two rounds per day, but up to six hours per game. More events in the US or America, University hosts appreciated.

Regards,
Gerd

[Rémi Coulom]

Hi,

Thanks to all who have replied.

I try to keep the summary in my first post up to date with the replies I read here. Let me know if you'd like to see the summary edited in order to reflect your point of view better.

Rémi