Author: Russell Reagan
Date: 14:27:39 10/17/03
Go up one level in this thread
On October 17, 2003 at 17:12:35, Russell Reagan wrote: >On October 17, 2003 at 15:52:25, Russell Reagan wrote: > >>On October 17, 2003 at 06:24:37, Tord Romstad wrote: >> >>>If you bother to continue the e-mail exchange with the author, please >>>point out that all comparisons with "Zillions of Games" are completely >>>worthless. >> >>Don't worry, I have. Now he wants to have a little contest where I write an >>array based move generator for his game, and compare it with his. He claims to >>be able to generate 140 million moves per second on unknown hardware. I suspect >>he isn't actually converting the attack bitboards into moves, and thinking that >>the attack board comprises a "move list" (which it could, but you're only >>pushing the extra work outside of the "move generator"). Or he's doing something >>else which allows him to juice up his numbers. > >More from him... > >"The hardware was a 2.0 Ghz Pentium III. > >We have a routine that spawns positions from attack bitboards which are fed >through the rest of the program. This occurs at lightning speed." > > >Apparently he's doing some pretty nifty overclocking, because there are no 2GHz >PIIIs (the highest overclocked PII I could find was a 1.76GHz), and he has some >magic routine that "spawns positions" from attack bitboards, whatever that >means. Assuming he even has a 2GHz cpu, to get 140 million moves per second, >that's 14 cycles per move generated, doing 64-bit bitwise operations, on 32-bit >hardware. Of course, he's doing 80-bit operations on 32-bit hardware, not 64-bit. That means no memory accesses, everything in L1 cache (doubtful with bitboards), no mispredicted branches...
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.