Author: Vladimir Medvedev
Date: 06:54:34 10/17/03
Go up one level in this thread
>I fear your switch case with single isolated LS1B is not optimizable as a jump >table, but works like 64 nested "if then elses". Not sure about how the branch >prediction works here, but i guess the squares on the "black" side are a bit >handicapped ;-) > >I suggest a lookup-approach (similar to your popcount), Walter Faxon's genius >folding trick, Matt Taylor's de Bruijn approach or several assembler routines >using 32-bit bsf, all already posted here in CCC x times. When I switched to bitboards from 0x88 representation I run some tests (with Win32 version), trying to use lookup table for LSB too. The results were (for set of random 64-bit numbers generated by my Randon64): 1. lookup is faster than switch() for PopCount 2. switch() is faster than lookup for LSB But I can't remember exact digits now...
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.