Author: Jeremiah Penery
Date: 23:22:09 01/31/00
Go up one level in this thread
On February 01, 2000 at 01:47:19, David Blackman wrote: >On January 31, 2000 at 12:32:28, Dan Andersson wrote: > >>The link http://www.eetimes.com/story/OEG20000128S0008 describes briefly a DSP >>on a memory chip. DSPs might still rule the world! If they wern't so dang hard >>to use. >> >>Regards Dan Andersson > >Has anyone written a chess program that spends 95% of it's time doing long >vector dot products? (Mine certainly doesn't.) If so it should really fly on a >DSP chip. Actually, if the low-latency lookup tables were RAM instead of the sin >and cos lookup ROM that most DSPs have, then i think you could just about do a >fast, smart static eval function on a DSP. > >Another piece of fast hardware that is already installed in many PCs is the 3D >graphics card. For a few hundred dollars you can get one that does many >GigaFlops, as long as what you want to do is render textured triangles. If you >design the textures really carefully, and line up multiple overlapping textured >triangles just right, maybe you could get it to do chess move generation and >eval really fast. But it wouldn't be easy. There are a bunch of other things these things can do: texture effects, bump mapping, z-buffering, bi-linear filtering, lighting/shading effects, etc. A lot of the new cards also have lots of memory (up to 64MB) - Hash tables! I think it would be really cool if someone could figure out how to make a chess program using this stuff. :) I suspect it would have to communicate directly with the main CPU to do some stuff, and it could also use the main system memory. Perhaps the 3D card could handle move generation and evaluation, and the main CPU could handle the rest... Jeremiah :))
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.