Author: Dave Gomboc
Date: 09:45:01 12/19/98
Go up one level in this thread
On December 18, 1998 at 11:38:49, blass uri wrote: > >On December 18, 1998 at 10:58:49, Harald Faber wrote: > >>On December 18, 1998 at 10:47:07, blass uri wrote: >> >>>>>In order to change the size of the hash tables I looked at c/program files >>>>>/chessmaster6000/cm and changes ttable_size but I am not sure if I did the right >>>>>thing because I found that there is ttable_size also in Cmdefault >>>>>and ttable_size is 20 in this file. >>>>> >>>>>should I change cm or cmdefault? >> >>>>Change the CM.INI >>> >>>I looked for the CM.INI and the computer gave me 1 file with the name cm but I >>>am not sure if I did the right thing because I did not find a difference in a >>>test position that I gave the computer when I change the hash tables >>> >>>here is the position: >>>r4rk1/2p3pp/p7/1p1pq3/8/2P2N2/PPQ2KPP/R1B5 w - - 0 1 >>> >>>How much time does it take chessmaster to find that Kg1 is losing(evaluation >>>-2.08) with ttable_size=26 and with ttablesize=10? >> >>Sorry, I am in office and can't check but there is no need to be 4x faster. >> >>>I found that in both cases the result is similiar >>>15 minutes and 4 seconds with ttable_size=26 and less than 18 minutes with >>>ttable_size=10(I do not know the exact time bexause I did other things in the >>>computer at the same time) >> >>15min<->18min is 20%. I admit I haven't investigated much the differences >>between the sizes of hash. Explanations and experiences for the size at which >>time control are well known. >>But 20% seems OK for me. Another point is that faster engines would profit more >>from larger hash tables. > >1)I am sure something is not allright because I changedthe ttable_size to -100 >and expected an error message from the computer but I found that chessmaster can >analyze > Finding an invalid setting in the configuration file, CM probably chose to ignore it and proceed with a default size. This is much better than an error message! Did you check the file afterward to see if it had overrode the illegal setting in the file as well? >2)I did not say that there is a factor of 20%(I did other things in the computer >at the same time in the test when ttable_size=10 > >3)I expect a bigger factor than 20% because ttable_size=10 is only one Kbytes >hash tables when ttable_size=26 is 64 Mbytes The difference you report sounds reasonable to me. You'd really have to check a suite of posiitons with both hash table settings to determine what speedup factor is realistic. In these circumstances, ~20% doesn't sound surprisingly low to me though. It may even be that 64 megs was higher than the optimal amount for the length of search being performed, and the same speedup factor would have been achieveable with only 16 or 32 megs. >Uri Dave Gomboc
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.