WCCC 2009 - Limit of 8 cores

[bob]

Over the last 10 years or so I have argued about this subject. The name of the game is which is the best computer program or which computer program plays chess the best. It is not about who has enough money or connections to use a main frame or who has a 10 billion move opening book vs a 10 thousand book.
Once again the quest is for the best program, period !!!
Why can't people either understand that or at least remember what the game is about. In my opinion no one, I repeat no one will ever know which program is the worlds best until they all use identical hardware and opening books.
I still remember in CCT5 or CCT6 where one program beat another without ever leaving its opening book. The entire game had been played before and just by luck his opponent made the same losing moves. That, in my opinion is not chess because the program never had to make a move on its own. Who knows, maybe the program could not play legal chess at all.
Bill

That's fine. Give us _all_ 5 node 8 core clusters and that's equal hardware, right? If a program can't use the parallel search approach necessary to use both a cluster and SMP node, tough. I'll vote for _that_ any day... But not using _one_ core, which is an arbitrary number that nullifies all the work many of us have done to create efficient parallel searches...

You have a distorted view of the purpose of the WCCC. If it is to find the best "program" using equal hardware, why not have a meeting, take the best program from the CCRL or SSDF or whatever, which use _equal_ hardware. That's a far better way to find the best "program" and saves a lot of time, effort and expense.

The problem is, what you stated is _not_ the purpose of the WCCC. Never has been, and I hope it never will be...

[bob]

Gentlemen
I realize that a Swiss system or even a round robin is not enough to really determine which program is the strongest. The basis of my post is to point out the ways these contest are held right now.
Most of us admired Deep Blue but I also believe that a great many people today also believe that some of todays strong program running on quad cores or more might actually beat Deep Blue.
Bill

So what? Many of the programs from 2017 will stomp Rybka of today. What does that prove and who cares? Incentive drives computer chess. Equal platforms is _not_ any sort of incentive and would have set computer chess back at _least_ ten years from where it is today, probably far more... Fortunately those of us organizing past CC tournaments recognized this and did not fall for the poorly reasoned arguments. We had uniform platform tournaments years ago. They are no longer held. Care to guess why? Total lack of interest.

[bob]

Gentlemen
I realize that a Swiss system or even a round robin is not enough to really determine which program is the strongest. The basis of my post is to point out the ways these contest are held right now.
Most of us admired Deep Blue but I also believe that a great many people today also believe that some of todays strong program running on quad cores or more might actually beat Deep Blue.
Bill

Sure, but if a rejuvinated Deep Blue (or Hydra) asked to enter it would be turned away.
Too darn many CPUs.

That's not true ! :

Alongside the World Computer Chess Championship the ICGA organizes the annual Computer Olympiad, and when there are two or more contestants we will organize, as part of the Computer Olympiad, an open chess tournament (or match if there are just two entries) in which hardware systems with more than 8 cores may compete. This will NOT be given the status of a world championship event – it will be the Computer Olympiad Open Chess Championship.

We already know what would happen. Those of us actively competing would go to the "open" event. And, like the old uniform-platform tournaments of years past, the new-and-improved WCCC would die away. We don't need to keep going down a path that is already known to be a complete failure... those suggesting this are simply not thinking clearly...

[bob]

In my opinion, they have this exactly backwards.

The WCCC has since its inception in 1970 been the venue for open hardware. It should remain so.

If they wish to return to the days of segregation, then they should reinstitute the World Microcomputer Chess Championships. That would be an appropriate venue for hardware limits.

Book limits are stupid, in my opinion. Are there any regular chess matches that impose such a rule? Of course not.

Ian

Actually the first was 1974. The ICCA was formed at the 1977 WCCC in Toronto, where many of us had gathered for the second WCCC. ACM tournaments started in 1970 however...

[geots]

Over the last 10 years or so I have argued about this subject. The name of the game is which is the best computer program or which computer program plays chess the best. It is not about who has enough money or connections to use a main frame or who has a 10 billion move opening book vs a 10 thousand book.
Once again the quest is for the best program, period !!!
Why can't people either understand that or at least remember what the game is about. In my opinion no one, I repeat no one will ever know which program is the worlds best until they all use identical hardware and opening books.
I still remember in CCT5 or CCT6 where one program beat another without ever leaving its opening book. The entire game had been played before and just by luck his opponent made the same losing moves. That, in my opinion is not chess because the program never had to make a move on its own. Who knows, maybe the program could not play legal chess at all.
Bill

That's fine. Give us _all_ 5 node 8 core clusters and that's equal hardware, right? If a program can't use the parallel search approach necessary to use both a cluster and SMP node, tough. I'll vote for _that_ any day... But not using _one_ core, which is an arbitrary number that nullifies all the work many of us have done to create efficient parallel searches...

You have a distorted view of the purpose of the WCCC. If it is to find the best "program" using equal hardware, why not have a meeting, take the best program from the CCRL or SSDF or whatever, which use _equal_ hardware. That's a far better way to find the best "program" and saves a lot of time, effort and expense.

The problem is, what you stated is _not_ the purpose of the WCCC. Never has been, and I hope it never will be...

What Bob has said above- there is no possible argument against. He is 100% correct. For anyone- read his above thread again. It sums it up in the best way possible.


Best,

[Daniel Mehrmann]

Hi, I brought this topic in another post and I agree with the decision of having uniform hardware, but I would have just stated one-cpu and let the users decide on the one cpu, instead of core number. I believe this decision has a lot to do with Rybka having 40 cores certainly a ridiculous situation!

The main investment for participating is not the investment in the hardware, but the investment in the software. It is not a trivial task to make effective use of >2 cores and the task will be even more difficult on cluster type of configurations.
So if the competition is about the software, it is nonsense to limit the hardware.
Richard.

I must 100% agree with Richard.

[Spock]

The main investment for participating is not the investment in the hardware, but the investment in the software. It is not a trivial task to make effective use of >2 cores and the task will be even more difficult on cluster type of configurations.
So if the competition is about the software, it is nonsense to limit the hardware.
Richard.

I think that is the best post I've seen on this issue, totally brilliant, and brings me clarity now on the issue. I had thought the hardware limit was a good idea and supported it, but this post has caused me to make a complete u-turn. NO hardware limit is the way to go, no question in my mind now.

[playjunior]

I wonder who would provide all the engine authors who participate a 40-core machine?

This is not about removing incentives for implementing good parallel search, it just gives a chance to amateurs who don't have 40-core clusters or University chair and funding with 500 core machines to compete. There are 1 billion such regulations in sports like Formula1, so that the rich teams cannot just outspend the poorer ones in a way that the drivers won't matter at all. Here, cars translate to hardware and pilots are the software. Well almost.


And, I think that if you have a really good parallel search compared to competitors, 8 cores would give quite some advantage. Look at scaling difference of Zappa vs engines of the same time.

[Harvey Williamson]

I wonder who would provide all the engine authors who participate a 40-core machine?

This is not about removing incentives for implementing good parallel search, it just gives a chance to amateurs who don't have 40-core clusters or University chair and funding with 500 core machines to compete. There are 1 billion such regulations in sports like Formula1, so that the rich teams cannot just outspend the poorer ones in a way that the drivers won't matter at all. Here, cars translate to hardware and pilots are the software. Well almost.


And, I think that if you have a really good parallel search compared to competitors, 8 cores would give quite some advantage. Look at scaling difference of Zappa vs engines of the same time.

Good post.

[bob]

The main investment for participating is not the investment in the hardware, but the investment in the software. It is not a trivial task to make effective use of >2 cores and the task will be even more difficult on cluster type of configurations.
So if the competition is about the software, it is nonsense to limit the hardware.
Richard.

I think that is the best post I've seen on this issue, totally brilliant, and brings me clarity now on the issue. I had thought the hardware limit was a good idea and supported it, but this post has caused me to make a complete u-turn. NO hardware limit is the way to go, no question in my mind now.

I have written that same thing about ten times each and every time this concept is brought up.