Clarification of the 8-cores rule for the WCCC

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

diep
Posts: 1822
Joined: Thu Mar 09, 2006 11:54 pm
Location: The Netherlands

Re: Clarification of the 8-cores rule for the WCCC

Post by diep »

bob wrote:
diep wrote:
bob wrote:
Yar wrote:Hello,

Clarification by David Levy is here:
http://www.grappa.univ-lille3.fr/icga/n ... .php?id=40

With best regards,
Yar
Ah, yes. That _completely_ clears things up.

Err... what in the hell is a "logical core"??? What if I were to use 16 illogical cores???

stupid rule, stupid decision... Reminds me of Forrest Gump, "stupid is as stupid does..."
A logical core is something intel only.
Other processors do not have this technology,
these technologies of other processors obviously are not forbidden.
like the SMT in Niagara (i wonder anyway who would want to run on a processor that can do like 1 instruction a cycle).

There is a huge difference between intel and other manufacturers here. Intel CLAIMS they have 2 times the number of cores than it has when i count. I'd argue if they claim it is a logical core then it is a core.

In case of Sun,IBM,Alpha (21464 never released though) and some embedded processors they have SMT to allow more threads, yet they do not claim that is additional cores, so it is not extra cores, so that hardware is allowed if you don't use more than 8 cores. As simple as that.

If you do a claim (intel in this case) it gets awarded of course, even though HT in intel doesn't help much for computerchess anyway for speedup as we know, and all participants will turn it off anyway.

It is a definition question, not something practical :)
My point was that the "wording" is stupid. A 4-core processor with hyper-threading is technically 8 "logical processors". If I were using one, it would have hyper-threading disabled and would be running with 4 physical cores. Could I run on a dual-I7 which has 8 real cores or potentially 16 logical cores?

Even a little bit of thought would have shown that this rule and discussion is so far beyond ridiculous, it takes sunlight 6 months to get from ridiculous to the discussion taking place.

And the sad thing is, David doesn't seem to "get this..."
The claim from intel is that they have copied some logics in the processor, in short Nehalem i7 has 8 REAL logical cores. This in contradiction to Sun/IBM/Alpha where a core does NOT split itself into more real cores, but basically tries to 'fill up' the existing core with more instructions, which we call SMT.

Now you can argue it is a definition question. Intel indeed has this logics and therefore 8 real logical cores. What the speedup of those logical cores is or isn't is none of my business to be honest, what matters is that intel claims the above and i'd argue: "let's count it as a core therefore".

Vincent
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Clarification of the 8-cores rule for the WCCC

Post by bob »

diep wrote:
bob wrote:
diep wrote:
bob wrote:
Yar wrote:Hello,

Clarification by David Levy is here:
http://www.grappa.univ-lille3.fr/icga/n ... .php?id=40

With best regards,
Yar
Ah, yes. That _completely_ clears things up.

Err... what in the hell is a "logical core"??? What if I were to use 16 illogical cores???

stupid rule, stupid decision... Reminds me of Forrest Gump, "stupid is as stupid does..."
A logical core is something intel only.
Other processors do not have this technology,
these technologies of other processors obviously are not forbidden.
like the SMT in Niagara (i wonder anyway who would want to run on a processor that can do like 1 instruction a cycle).

There is a huge difference between intel and other manufacturers here. Intel CLAIMS they have 2 times the number of cores than it has when i count. I'd argue if they claim it is a logical core then it is a core.

In case of Sun,IBM,Alpha (21464 never released though) and some embedded processors they have SMT to allow more threads, yet they do not claim that is additional cores, so it is not extra cores, so that hardware is allowed if you don't use more than 8 cores. As simple as that.

If you do a claim (intel in this case) it gets awarded of course, even though HT in intel doesn't help much for computerchess anyway for speedup as we know, and all participants will turn it off anyway.

It is a definition question, not something practical :)
My point was that the "wording" is stupid. A 4-core processor with hyper-threading is technically 8 "logical processors". If I were using one, it would have hyper-threading disabled and would be running with 4 physical cores. Could I run on a dual-I7 which has 8 real cores or potentially 16 logical cores?

Even a little bit of thought would have shown that this rule and discussion is so far beyond ridiculous, it takes sunlight 6 months to get from ridiculous to the discussion taking place.

And the sad thing is, David doesn't seem to "get this..."
The claim from intel is that they have copied some logics in the processor, in short Nehalem i7 has 8 REAL logical cores. This in contradiction to Sun/IBM/Alpha where a core does NOT split itself into more real cores, but basically tries to 'fill up' the existing core with more instructions, which we call SMT.

Now you can argue it is a definition question. Intel indeed has this logics and therefore 8 real logical cores. What the speedup of those logical cores is or isn't is none of my business to be honest, what matters is that intel claims the above and i'd argue: "let's count it as a core therefore".

Vincent
My problem is that I have run on one, and it is _not_ the same as 8 real cores.