Author: Vincent Diepeveen
Date: 06:04:24 06/01/01
Go up one level in this thread
On May 31, 2001 at 19:59:27, Robert Pawlak wrote: >Although Bob has a good point about net bandwidth, I have been wondering for >some time whether something akin to SETI@home is worthwhile for a CAP-like >endeavour. Maybe someone has already proposed this, but I just missed it. > >This way memory bandwidth would not be an issue, since each PC would be working >on it's own set of positions. > >Bob Transposition table is the biggest speedup for a chessprogram by far. Of course nullmove gives a great speedup and it can work IN COMBINATION with transpositoin table. But if i had to chose whether to get rid of a) extensions b) nullmove c) hashtables d) evaluation Then i would chose of course evaluation latest. Now the time required to do an evaluation is nothing compared to latency issues of the network. Nullmove you always can do, network or no network. Extensions idem. Figuring out where to split is no big problem to communicate over the network, but... ...one thing you NEED to share somehow and that's tranposition table. Because big number of Ghz is going to give on paper huge depths, you simply CANNOT afford to lose hashtable info. Now there are solutions for it to do it with huge latency which involve big bandwidth, but basic problem is that you need to transport huge number of information. The speedup is SO huge. If you don't do it then the branching factor gets real shit, to use a big understatement. Also it's harder to figure out where to split etcetera. EVErYTHING goes harder.
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.