Author: Gordon Rattray
Date: 13:37:21 06/10/01
Go up one level in this thread
On June 10, 2001 at 02:32:25, Ian Osgood wrote: >On June 09, 2001 at 20:08:11, Gordon Rattray wrote: > >> >>Thanks! This is an improvement over PocketChess to say the least! ;-) >> >>I've not been able to access the features.txt yet. I've got 64Mb on my Ipaq to >>play around with, so should I aim for large hash table settings? Or is the >>processor too slow to take advantage of these? >> >>Gordon > >Good question. I haven't used much other than the default 3M size, except to >test the hash option code. I know that Christopher Theron believes that 48K >tables are big enough for Tiger on the palm (about 500 nps). PocketCrafty >obtains about 5000 nps on an iPAQ, so maybe around 480K is more reasonable. > >This would be an area you could test and report on if you like! > >Ian Ok, I've now performed some testing as follows. Due to limited time I've only tested against one position: [D] 1r2kb1r/1p1R1ppp/4p3/q2pP1B1/p2Q4/P2B4/1P3PPP/2R3K1 w - - Test setup: Compaq Iqaq, 64Mb PocketCrafty v1.0 using Crafty 18.6 Analysed position for 10 minutes at each setting. Results (times are mins:secs): Hash table size / Time to find Rxd7 / Time to display depth "9 ->" / Nodes per second 48k 1:35 9:18 6562 96k 1:48 8:53 6580 192k 1:48 7:43 6761 384k 1:29 7:31 6634 768k 1:23 6:56 6753 1536k 1:20 6:49 6801 3072k 1:22 6:37 6800 6144k 1:21 6:34 6827 12888k 1:21 6:20 6756 It is not possible to configure PocketCrafty to use anything above 12888k. So, for longer time controls there is a definite benefit, but such time controls probably won't get used much on a PDA. Gordon
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.