Author: Robert Hyatt
Date: 06:54:04 07/29/01
Go up one level in this thread
On July 29, 2001 at 01:46:46, pavel wrote:
>On July 28, 2001 at 23:42:00, Robert Hyatt wrote:
>
>>On July 28, 2001 at 19:55:54, Roy Eassa wrote:
>>
>>>Dann, you keep insisting that Dr. Hyatt does NOT endorse Ponder=ON for
>>>program-versus-program matches on a SINGLE computer with a SINGLE CPU. Yet
>>>that's what he says. Have you seen his responses to your statements?
>>>
>>>I must admit that the idea seems odd to me -- who's to say how much CPU power
>>>each app might use when it's not the active/front one? I've heard that CM8000
>>>totally hogs the CPU even when it's in the "background".
>>>
>>>On the other hand, I would not take Dr. Hyatt's opinion lightly. As you two
>>>clearly represent the opposite ends of the spectrum on this one (with Dann
>>>saying "you'd have to be crazy" to use Ponder=ON), I'd kind of like the two of
>>>you to discuss this and see if either one can convince the other.
>>
>>
>>OK... let me first explain my test again. I used two machines. But the
>>reason I used two machines was to handicap crafty enough to make the match
>>interesting. I used a single cpu 200mhz pentium pro for crafty, a 500mhz
>>PIII for an older gnuchess. With ponder=on, crafty played well and won
>>in spite of the time handicap... WIth ponder=off, crafty lost badly as it
>>was running out of time (the time control was something in 40 moves, like
>>40 moves in 15 minutes, etc... and near the end of each time control it was
>>running into time trouble.
>>
>>My conclusion... ponder=off was worse for crafty as its time management was
>>not well designed nor tested for this mode.
>>
>>As a result, I have always maintained that it makes more sense to play ponder-on
>>matches, even on a single machine. Both machines should compute all the time,
>>and they should get 50% of the cpu (they do under linux for example). Ponder=on
>>eliminates the time allocation problems in crafty, with no bad side effects at
>>all.
>>
>>In a utopian world, either should work equally well. But in my case, since I
>>do _all_ my testing with ponder=on, that is the best way to run the engine,
>>whether you use one cpu or two.
>
>
>so in simple words ponder==off effects its time management, not its playing
>strength.
>
>am I right?
Yes and no. :) ponder=off does affect strength. But if both have it turned
off, then the effect should be similar assuming both implement ponder=on in a
reasonable way.
Also, time management is an important feature of an engine... and it definitely
affects strength overall. If you get into time trouble, then you have to play
weaker to avoid losing on time, and vice-versa.
>
>pavs,
>
>
>ps, cant get crafty.rc to work, I put hash 200m on the first line to change its
>default hash setting. doesnt seem to help, but on top of it crafty doesnt move
>on first try, it sends this.
That is most commonly caused by a missing CR/LF on the end of the last line
in the crafty.rc file. Just make sure the last line of that file is blank
and you will be ok...
>
>EPD Kit revision date: 1996.04.21
>unable to open book file [./books.bin].
>found computer opening book file [./bookc.bin].
>
>Crafty v18.10
>
>White(1): e2e4
>Illegal move: -- _í_2e4
>White(1): e2e4
> time used: 1:17
> clearing hash tables
> book moves {c5, e6, d6, Nf6, c6, e5}
> book 0.0s 4% d6 d4
>
>Black(1): d6
> time used: 0.43
>learning position, wtm=0 value=0
>White(2): d4 [pondering]
> book moves {g6, f5, Nf6}
> book 0.0s 15% g6 Nc3
>White(2):
>
>
>
>
>whats wrong?
>
>I have 512mb memory
>
>pavs...
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.