Author: Anthony Cozzie
Date: 06:57:12 05/06/04
Go up one level in this thread
On May 06, 2004 at 07:17:51, Vasik Rajlich wrote: >On May 05, 2004 at 17:14:20, Anthony Cozzie wrote: > >>I experimented with attack tables on the recommendation of Vincent last fall. >>The following code is based on his move generator. I eventually decided that on >>32 bit machines, attack tables were a clear win over bitboards, but at 64 bits >>it would be about a break-even deal, and since I already had a bitboard engine, >>the choice was easy ;) >> >>anthony >> > >I'm curious, why do you think bitboards are break-even at 64 bits, but a loss at >32? > >I have noticed that Rybka's NPS relative to Fritz varies greatly based on the >architecture. > >Aside from 32 vs 64 bits, L2 cache is probably very important. Rybka's >precomputed bitboard masks take just under 200 K (this # should be more or less >the same for any rotated bitboard engine), but I don't know how this behaves in >the cache. > >Vas It was more of a guess. Zappa gets a _lot_ faster on 64-bit machines, though. It is something like 3x as fast on my Opteron 1.8 compared to Gunther Simon's P4-2.4. Of course, some of that is due to other architectural superiorities, but still . . . anthony
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.