I hope that Always do a version Engine for ONE CPU CORE

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

Moderators: hgm, Dann Corbit, Harvey Williamson

Karmazen & Oliver
Posts: 374
Joined: Sat Mar 10, 2007 12:34 am

Re: I hope that Always do a version Engine for ONE CPU CORE

Post by Karmazen & Oliver »

bob wrote:
Karmazen & Oliver wrote:
Ovyron wrote:
Karmazen & Oliver wrote:maybe it´s better if THE ENGINE can select How MANY % of CPU and number of that use...
Multi CPU engines already have the option to only use one core.
are you sure that ALL engines MULTICORES have that options ?

deepfritz ?
deepschereder ?

multi rybja OK...

but the problem is that THAT engines are optimized for > CPU...

what is better, the version MONO CORE ? or a version multicore that ONLY use ONE CPU ?

i think that the version optimized for ONE CPU are better that versions "deep" ... castrated in ONE CORE...

bye.
Your concept of parallel search engines is badly garbled. They are not "optimized for more than one core". They are optimized for one core, in fact, but will use more than one if possible...
I´m not sure about that coment ? after all, it is not supposed that rybka lost the party in mexico, because zappa this better one optimized for 8 cores...

In other theme... I´m sure that THE version for multicores, are "diferents" that versions in SIGLE mode... the reason of this is that in parallel processes, acces cache, RAM for each processor... remember that same engines using RAM x2.. if have two cores... etc...

on the other hand... programing for ONE CPU, have not these problems ... the algoritms can are more "LINEAL"...

The question is:

¿ Is it better a ENGINE program on SIGLE CPU ?

or

¿ Is it better a ENGINE multicore, that we force to using ONE CPU core ?

I think that the microcode for engines Deep. "MP", They make more slow to these if we only using ONE core...

sorry me... but I don´t have more "words" ... for explain it... ;-)

bye. from spain. Oliver
Karmazen & Oliver
Posts: 374
Joined: Sat Mar 10, 2007 12:34 am

Re: I hope that Always do a version Engine for ONE CPU CORE

Post by Karmazen & Oliver »

M ANSARI wrote:... but a MP engine can and should be able to use a single core.
YES ¡... but is it similar results that the engine single core...?

I think that that engine have more program code that is compensated when using but processors, if alone we use a processor we are in an unfavourable situation... with regard to an unique design...
M ANSARI wrote: By the way today's MP engines would also run fine on an old 486 machine.
Are you sure ? some engines Deep. not run on my "P200MMX", because need other registers, cache, etc...

bye from spain. Oliver
Karmazen & Oliver
Posts: 374
Joined: Sat Mar 10, 2007 12:34 am

Re: I hope that Always do a version Engine for ONE CPU CORE

Post by Karmazen & Oliver »

Spock wrote:If I can quote Vas from the Rybka forum:
http://rybkaforum.net/cgi-bin/rybkaforu ... id=2519#fp
Actually, single-processor Rybka sales far outnumber multi-processor, by something like 4:1. For Fritz, this ratio might be even a little bit higher...
Vas
OK ¡¡ now this is a logical pronostic ¡¡¡ ;-) because MP are more expensive, and there a lot of people with have ONE CORE... but the reason given for VAS not is a good new:
... The average person is just not sitting at the top of the hardware curve.
Vas
OK... he attributes that situation, because a lot of people not are in "hardware curve"...

OK. I hope or I wait that this it is not the only reason... ¡¡¡ read that fantastic resume to URIC:
Oliver explained that using single processor engine is not limited to people
who have only single processor and even people who have more than single processor
may want to use only one processor for analysis when the second processor does a different task at the same time.

Uri
also ... for that in a time future, the hardware curve opens up overcome my pocket money, (bankruptcy?) , and I will continue needing an engine SINGLE ;-)

bye. from spain. Oliver

Spock wrote:That really surprises me
OK. not for me... ¡ ;-)