Author: Robert Hyatt
Date: 13:00:26 03/27/03
Go up one level in this thread
On March 27, 2003 at 13:17:22, Gerd Isenberg wrote: >On March 27, 2003 at 10:06:47, Dezhi Zhao wrote: > >>On March 27, 2003 at 05:35:10, Uri Blass wrote: >> >>>If I want to get rid of the least significant 1 of a BitBoard >>>I do for a bitboard a >>>a&=a-1 >>> >>>What is the fastest way to get rid of the most significant 1 of >>>a bitboard? >>> >>>Is it slower than a&=a-1 and how much slower? >>>(it cannot be faster otherwise popcount of crafty could use it). >>> >>>Uri >> >>To my best knowlege, there is no such well known trick for the most significant >>bit. >> >>Maybe we can send a wish list to Intel and AMD for some new instructions. >>To seed the list: >> >>1. bsfr : bit scan forward and reset >>2. bsrr : bit scan backward and reset >> >>requirements: NOT in slow microcode or latency <= 2 cycle:) >> >>dzhao > >Agreed - at least bsf, bsr, btr reg64,reg64 <= 2 cycles each ;-) > >Gerd I'd be happy with bsr reg32, reg64 myself. :) And not to forget popcnt reg32, reg64 :)
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.