Re: [Algorithms] Fast Floating-Point compares?

[Jonathan B. <jo...@nu...>]

Also it's worth pointing out that if what you're really writing is
something like Max() or Min(), or "use index a if x < y, else
use index b", you can do all that without branching pretty
easily, if branching throws you into a tizzy.  (and without
using newer assembly instructions like cmov).  So if you want
the Max() of two floating-point numbers each of which you
know is non-negative, it all boils down to a small number of
integer math instructions only.

----- Original Message ----- 
From: "Conor Stokes" <cs...@tp...>
To: <gda...@li...>
Sent: Friday, August 15, 2003 11:31 PM
Subject: Re: [Algorithms] Fast Floating-Point compares?


> > Compares are always slow because filling branch delay slots is not
> > always possible and aggressive prediction and speculative execution are
> > very hard problems.  You'd be better off spending your time figuring out
> > a better algorithm (gee, that's what this list is for!) and doing fewer
> > compares in total.
> 
> Compares aren't always slow (especially on the x86) - branches give them a
> bad name.
> 
> Conor Stokes
> 
> 
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email sponsored by: Free pre-built ASP.NET sites including
> Data Reports, E-commerce, Portals, and Forums are available now.
> Download today and enter to win an XBOX or Visual Studio .NET.
> http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
> _______________________________________________
> GDAlgorithms-list mailing list
> GDA...@li...
> https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_id=6188