Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Bitscanning on the Opteron

Author: Gerd Isenberg

Date: 04:20:37 12/05/03

Go up one level in this thread


On December 05, 2003 at 00:22:11, Russell Reagan wrote:

<snip>

>Bob said, "I suspect my approach is actually faster when you use it in some
>real code, as it gets inlined very compactly..." and he is probably right.

Yes, probably the shortest code will best. May be, for lsb with reset 64-bit de
Bruijn mul may be competitive, due to the yet slow "bsf/bsr" 9-cycle vector path
implementation, but fast 4-cycle direct path 64-bit mul on Opteron.

>Cramming functions into loops like this isn't always the best way to benchmark
>something.

For sure, even more with predictable pattern.
Huge lookup tables may pollute cache in a real program a bit more.
Was it your intention to mix lsb- with msb-routines?

Gerd

<snip>




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.