hash memory test
A few days ago the question arose, in my case, whether it was better to perform a hash test on each chess engine that participated in a tournament and assign an amount according to the results, that is, each engine a different size. Or, on the contrary, test all the engines participating in the tournament and look for a size that (does not exceed 30% ram usage) is the same for all engines. Andrew from Openbench suggested to me that it was better to have the same size for everyone to avoid bias towards one chess engine or another. What do you suggest? Thanks in advance.
Jorge
hash memory test
hash memory test
I am not an expert in anything, I do not believe in anything, but everything is possible.
- Graham Banks
- Posts: 26995
- Joined: Sun Dec 18, 2005 5:47 pm
- Sign-up code: 0
- Location: Auckland, NZ
Re: hash memory test
In CCRL testing each engine uses the same amount of hash where possible.
256mb or 512mb per core.
256mb or 512mb per core.
- Gabor Szots
- Posts: 12890
- Joined: Sat Dec 09, 2006 6:30 am
- Sign-up code: 10159
- Location: Szentendre, Hungary
Re: hash memory test
For my blitz tests I always use 256 MB hash when possible.
Re: hash memory test
In CCRL all engines in a tournament or gauntlet compete under the same conditions, so a common hash size.
Some engines do use a little bit more than what is set.
Some engines do use a little bit more than what is set.
-
- Posts: 1983
- Joined: Thu Aug 05, 2021 2:35 pm
- Sign-up code: 10159
- Location: Cavite, Philippines
- Contact:
Re: hash memory test
I use 256mb, and I already did a test about it that 256mb performs well than 128 and 512 in our time control 2+1 or 40/2
CCRL Testing Group
Re: hash memory test
I use 512MB for the Chess324 list at 2+1. Being a new list I could choose what I wanted.