Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Neverending story with incomplete tablebases

Author: Sune Fischer

Date: 03:47:15 08/18/03

Go up one level in this thread


On August 18, 2003 at 06:36:14, Uri Blass wrote:
>>>Well Sune, if it was that easy to just take 'em...
>>>Starting with SMP complicates your program by a factor x and introduces new
>>>fresh bugs. That goes for every little improvement and change you are going to
>>>do. You first have to think about how it affects your SMP logic even if it's
>>>"foolproof" written with good design. You always have to keep it ín mind and
>>>your debugging is even harder than before. More complicated and more bug
>>>intensive = 50 ELO. Hmmm I think I'll wait. If it was >51 I would go for it ...
>>>:-)
>>>
>>>/Peter
>>
>>That might be true, I don't have much experience with it, but I think it should
>>be manageable.
>>
>>Probably the primary argument not to do it, is that not many users have duals,
>>so from commercial point of view perhaps indeed it is better to spend the time
>>on something else which will benefit all.
>>
>>However, for the die hard fan who just wants the strongest kit (HW+SW), dual is
>>the way to go.
>>
>>-S.
>
>I am not sure.
>
>If you believe that at the same time that you implement dual you may implement
>instaed of it some better search algorithm than dual is not the way to go.

You arugment seem to require a never ending stream of improvements to the search
algorithm, and at the same rate as it would take to make the program parallel.

While this is no doubt true for most amateur engines for quite a while to come,
I'm not sure it is the case for the pros who has already been undergoing
optimizing procedures for decades.

>Remember also that dual also may make slower the progress after you implement it
>because you may need to care about more possible bugs.

I think it is matter of habit, you can get used to thinking parallel like you
can get used to so many other things. Weird pruning rules can certainly also
make life misserable ;)

Anyway, if the program is designed from the beginning to go parallel, I think it
is doable, just look at Crafty and Amy :)

-S.
>Uri



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.