Author: Robert Hyatt
Date: 16:22:29 09/02/02
Go up one level in this thread
On September 02, 2002 at 17:24:00, Jesper Antonsson wrote: >On September 02, 2002 at 13:09:17, Robert Hyatt wrote: >>The question becomes how large can the jump table be? The answer is, it can >>be as large as you are willing to bear. It costs memory. In the "old days" >>the table had to be pretty small, but today, you could easily have values >>like 0, 100000 if you wanted to optimize for speed rather than for memory. > >Well, nowadays, you also have to consider branch prediction and instruction >cache issues. I wouldn't be sure that a big jump table would be as efficient if >you just have a few interesting values, if I hadn't tested it. A cache line is only 32 bytes long (64 on new PIV systems). That is either 8 jump addresses or 16. I don't think that the jump table will cause any adverse issues because of that...
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.