Author: Vincent Diepeveen
Date: 05:04:02 12/29/02
Go up one level in this thread
On December 29, 2002 at 06:12:30, Dave Gomboc wrote:
>On December 29, 2002 at 04:15:41, Alessandro Damiani wrote:
>
>>On December 29, 2002 at 01:17:14, Dave Gomboc wrote:
>>
>>>On December 28, 2002 at 12:31:01, Alessandro Damiani wrote:
>>>
>>>>My new variant of ABC uses depth reductions instead of extensions. So I am
>>>>looking at the difference between Ed's reductions and mine. I don't use
>>>>conventional null-move yet.
>>>>
>>>>There is a doctoral thesis by Thomas Barth which describes how depth reductions
>>>>work fine with a hashtable. His work is from 1988.
>>>>
>>>>Alessandro
>>>
>>>Could you please post a URL to the location of the thesis?
>>>
>>>Dave
>>
>>I guess you have to go to your university library because I have got a paper
>>copy from the library here in Zurich. If you can't find it in your local library
>>you may ask there for a distance rent from another library. I don't know if
>>there is a digital copy of this thesis.
>>
>>Here are the details:
>>Thomas Barth, Dissertation, Neue Varianten von Suchverfahren und
>>Stellungsbewertungen im Computerschach, Technische Universität Wien, 1988
>>
>>The thesis is written in German.
>>
>>Alessandro
>
>German, huh? Darn. ;-)
>
>Dave
I read german better than english, no problem :)
However around 1997 i already fixed the bug i had with non-recursive
FHR reductions in hashtable. For Ed's original Reduction-1, to be
called RED1 later on here, this is trivially the same thing to do.
Of course FHR is highly buggy whereas Ed's RED1 because of the clearly
defined conditions for it and the high values, is more likely to work.
Especially knowing that Ed won a world title with his clever search
techniques and the good evaluation of Rebel, this against giant opponents
in a time that the number of Mhz of Ed's processors weren't big yet.
Fix, to implement RED1 correctly. Let's assume you have a variable
RED1 which is put to 0 if no reduction has been done and to 1 if
a reduction has been done.
Now we add to our hashtable an extra bit which describes RED1 too.
Because in pseudo-code you can waste so much more bits and cycles:
hash.red1 it is in hashtables.
Further we have the normal transposition code which also uses values
like depthleft.
So the depth comparision which currently should read:
if( hash.depthleft >= depthleft ) {
.. further normal transposition code
}
It should get now:
if( hash.depthleft > depthleft
|| (hash.depthleft == depthleft && (!RED1 || hash.red)) ) {
.. further normal transposition code
}
Best Regards,
Vincent
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.