On Wed, 29 Jan 2003 03:00:00 +0100 (MET)
Peter Finderup Lund <[EMAIL PROTECTED]> wrote:
> On Tue, 28 Jan 2003, Felix Kühling wrote:
>
> > prefetching looks odd to me. In the first transform loop in
> > _mesa_3dnow_transform_normalize_normals memory is prefetched which is
> > never read but only written. This is obviously useless. Then in the
>
> No -- at least not *obviously* useless. Whether it *actually* is useless
> or not depends on the write allocate policy of the cache. On some caches
> it is a good idea to prefetch something you are going to write to because
> if it weren't in the case the write would have to trigger a read anyway so
> you might as well do that early on.
Yeah, I realized that. You never overwrite a whole cacheline at once.
Sorry for my harsh words.
>
> Can't remember the K6 details, though, so it might be unnecessary.
At least the Athlon optimization guide doesn't mention this specific
case. They just say if you are going to modify data you should PREFETCHW
it.
>
> -Peter
>
> ``Average programmers should be rounded up and placed in internment camps
> to keep them away from keyboards.''
>
Felix
__\|/__ ___ ___ ___
__Tschüß_______\_6 6_/___/__ \___/__ \___/___\___You can do anything,___
_____Felix_______\Ä/\ \_____\ \_____\ \______U___just not everything____
[EMAIL PROTECTED] >o<__/ \___/ \___/ at the same time!
-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel
