[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [xougen] Arch specific optimizations [XAA]
From: |
emmanuel ALLAUD |
Subject: |
Re: [xougen] Arch specific optimizations [XAA] |
Date: |
Tue, 9 Sep 2003 22:34:42 +0200 (CEST) |
--- Alan Cox <address@hidden> a écrit : >
On Llu, 2003-09-08 at 16:44, emmanuel ALLAUD wrote:
> > So basically you think that the only real
> improvment
> > would be done by adding hardware acceleration
> support
> > to some core functions of XRender (I have seen
> > messages on XFree-devel from someone who is doing
> that
> > though I don't know how advanced he is now).
>
> Once you can do DMA fetch of tiles of video screen
> (reading them
> with the CPU is really slow) then every last clock
> you can ring
> out of the software renderer will be worth it for
> stuff you can't
> directly perform on the video card itself.
>
> Currently you should be able to use the vesa
> unaccelerated driver
> with shadowfb to write hand tuned Xrender functions
> for unaccel
> cards and that doesn't need to wait for other stuff.
> (In the
> vesa/shadowfb case your render buffer to read is
> main memory so
> the performance issue is gone and you can just fix
> the render
> code to be fast).
OK just to make sure my brain is catching up now ;-)
Either we use hardware to render things directly,
either we do it via software in main memory (and as
smart as possible to be speedy) because the big
bottleneck is _reading_ video memory (at least without
DMA). Is that the idea?
Bye
Manu
___________________________________________________________
Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com