Author: Christophe Theron
Date: 18:51:50 04/28/00
Go up one level in this thread
On April 28, 2000 at 20:44:04, Robert Hyatt wrote:
>On April 28, 2000 at 15:55:14, Dann Corbit wrote:
>
>>On April 28, 2000 at 15:36:59, Christophe Theron wrote:
>>[snip]
>>>While I think the ponder=off experiment is a valid one (at least to see how
>>>program strength varies with time controls), I have doubts about ponder=on on
>>>one CPU.
>>>
>>>If you can, I would suggest you use 2 computers. But if you do not have 2
>>>identical computers, ponder=on on one CPU will maybe tell us something too...
>>
>>Here I agree. Which program dominates will depend largely on how effectively a
>>program grabs all of the CPU. If a program had an occasional
>>yield()/sleep()/whatever in it so that other programs could get a chip of CPU
>>from time to time or at least the mouse moved freely, it might crush that
>>program in that setting.
>>
>
>
>Clue: You will _not_ find sched_yield() or sleep() in _any_ chess engine
>'core'.
>
>no chance whatsoever.
>
>:)
Under Windows, polling the keyboard too often (at each node for example) would
do that. Above some thresold, Windows thinks the program is idle, waiting for
user input!
I doubt it is the case here. It would happen only if the program was written in
DOS of console mode.
Christophe
>>I don't think the machines have to be equal, as long as you alternate and play
>>at least 4 games. E.g. with Winboard, use /mg 4.
>>
>>However, if the machines are totally lopsided (e.g. p5 against AMD Athlon) then
>>that would harm results also. But as long as the machines are within 2x of each
>>other, I think you would still get data of some value.
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.