Author: Robert Hyatt
Date: 12:42:26 09/18/99
Go up one level in this thread
On September 18, 1999 at 07:13:01, Inmann Werner wrote: >On September 17, 1999 at 16:16:39, Robert Hyatt wrote: > > > >> >>That is not the way to do it. There are two good ways to do this. >> >>One is to take each byte and use that as an index into a table that >>gives the first 1 bit set in a byte with that particular subscript. >>The other is to use the assembly code that is included with crafty so >>you can access the BSF/BSR hardware instructions that are _very_ fast. > >I tested the different routines (vcinline.c) > >AMD K62-450 (100 Million tests) >BSF/BSR 11.640 sec >Lookup 8.900 sec >My Rout. 14.720 sec > >Is the AMD BSF/BSR routine slower then the PII one? > >Werner your lookup test is badly skewed... if you just run it over and over, you end up in cache. If you run it normally, with all the other pieces of the engine, you don't get near as many cache hits and BSF blows it away...
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.