Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Don't have a native bit-population-count instruction? Here's why.

Author: Robert Hyatt

Date: 20:02:20 01/12/06

Go up one level in this thread


On January 11, 2006 at 19:13:11, Stuart Cracraft wrote:

>On January 11, 2006 at 09:44:57, Robert Hyatt wrote:
>
>>On January 10, 2006 at 18:10:15, Stuart Cracraft wrote:
>>
>>>Here's Butler Lampson, one of the inventors, of the Parc Alto
>>>talking about Seymour Cray and population-count units:
>>>
>>>http://research.microsoft.com/users/gbell/craytalk/sld051.htm
>>
>>Seymour said the same thing about the "leading zero" instruction, which was
>>about the same thing as "find first one bit (starting from left) since the
>>instruction just counting the leading zeros up to the first one.  The Crypto
>>folks used both.
>>
>>And yes, the 7600/Cyber-176 was a fast box in its day.  Chess 4.x used it to
>>drub other chess programs for a few of years until a certain program moved to
>>even faster hardware (the Cray-1).  :)
>
>Great stuff.

>
>Wish I had a population count instruction and find first one bit instruction!
>
>I hear that these exist on some chips but are not advertised.

Intel has BSF/BSR which find first/last set one bit quickly.  They work on
X86-64 machines too so that they are available for full 64 bit operands, just
like on a cray.  But no popcnt.

Apples have the leading zero type instruction as well.  There are a few
intrinsics used in Crafty to get to this if you search the source.  I didn't
write 'em however, someone contributed...



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.