Author: Russell Reagan
Date: 12:55:07 06/07/02
Go up one level in this thread
I think 64-bit will be a big step, and then the next big step will probably be 256-bit, which would allow an entire position to be stored in a single 256-bit value (64 squares at 4-bits each). I could forsee there being advantages to this. Maybe a new kind of bitboard, maybe a nibbleboard? Regardless of the name, you could implement some kind of scheme to where making a move would be as simple as doing an XOR operation. Eventually when computers have large enough word lengths, we will be able to do a lot of cool things I think. Maybe generating legal moves at a single CPU instruction (AND, OR, XOR, whatever) per move, making a move in a single instruction, unmaking a move with the SAME XOR operation (that sounds really cool to me, XOR it in, XOR it out), and so on. We will probably see (at least) 100GHz processors in our lifetime. If such a chess program were to be developed that made use of such simple CPU instructions, that would mean a program would crunch through about 2.8 billion NPS. A computer like that could search 40 plies ahead in about 6.5 minutes. I suppose by then we'll have bigger EGTB completed, and if you rigged up 32 of these processors on a machine, you could search pretty deep and maybe even find a mate from the opening position. Probably not, but it's interesting nonetheless. Russell
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.