On Tue, 28 Jan 2003 15:05:30 -0800 Ian Romanick <[EMAIL PROTECTED]> wrote:
> > I'll see if I can clean this up a bit. On the mesa-4-0-4 branch this > > code is disabled anyway, so there is not really a hurry to apply my > > stupid little patch. About this reading backward thing, where is that > > documented. I have an AMD Athlon optimization guide from February 2002 > > which doesn't mention it. > > I've seen a reference posted to dri-devel a couple times. Here's a > couple references the Dieter posted on 09-Jan-2003: > > http://marc.theaimsgroup.com/?l=linux-kernel&m=103548024914815&w=2 > http://208.15.46.63/events/gdc2002.htm Ok. It's this block prefetch which is also documented in the optimization guide. The reading backwards thing is not faster in itself. But it uses the same register as counter and index, when the index gets 0 you're done. So basically you save one increment or decrement operation per cycle. They also substitute prefetching with pre-reading which cannot be ignored. I think this has the biggest impact here. But according to AMD it is only useful for memory copying. For arithmetics prefetch works better. 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