Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HyperThreading & Comp v Comp matches

Author: Robert Hyatt

Date: 13:37:57 11/24/04

Go up one level in this thread


On November 24, 2004 at 10:04:16, Tony Werten wrote:

>On November 24, 2004 at 09:38:37, Robert Hyatt wrote:
>
>>On November 24, 2004 at 07:53:15, Tony Werten wrote:
>>
>>>On November 23, 2004 at 23:08:42, Robert Hyatt wrote:
>>>
>>>>On November 23, 2004 at 01:11:10, Ray Banks wrote:
>>>>
>>>>>On November 22, 2004 at 23:20:58, Robert Hyatt wrote:
>>>>>
>>>>>>I use ponder-on all the time, one cpu or not...  If both engines ponder, each
>>>>>>gets 1/2 the processor, and that doesn't disrupt normal time allocation...
>>>>>
>>>>>Why does everyone use ponder off then on their tourneys?
>>>>
>>>>I don't have any idea...
>>>>
>>>>
>>>>>
>>>>>I've previously been told that with a single processor you can't guarantee each
>>>>>engine will get 50%, therefore to use ponder off.
>>>>
>>>>With a decent O/S I don't see how there could be any imbalance at all, except
>>>>for the case where one engine is to lazy to find a move to ponder, and it just
>>>>"sits"...  but that is an engine deficiency, not a problem with testing with
>>>>ponder=on...
>>>
>>>When engine A is out of book and engine B isn't, A is rewarded for engine B
>>>having a better book.
>>>
>>>Tony
>>
>>So?  There are just as many good things as bad things...  And it _does_ test the
>>engine the way it runs "normally"...
>
>That wasn't really the point. You can't garantee that each engine will get 50%.
>
>Besides that you are lowering the quality of play, since you waste processing
>cycles ( ie the bad pondermoves )

use 2x the time.


>
>But then again, the logic "Normally I run them with ponder on, because then I
>have them run as I normally have" is hard to beat.
>
>Still, I don't really see the advantages of running 2 engines on a single cpu
>with ponder on (rather than ponder off) except when you're testing pondering.
>
>Tony

I personally allocate time differently when pondering or not pondering.  The
"not pondering" time allocation doesn't get tested by me at all, and in the past
it has done poorly.  I've tweaked it from time to time, but only when someone
sends me logs showing a problem.  I tweak the normal time allocation anytime I
see something unusual, although I have not done so for a couple of years now...



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.