AIGLX on Radeon Mobility

[Owen Taylor]

On Thu, 2006-02-23 at 10:57 -0500, Adam Jackson wrote:
> On Thursday 23 February 2006 06:08, Ross Burton wrote:
> > Hi,
> >
> > I've got a Thinkpad T40p with a Radeon chip:
> >
> > VGA compatible controller: ATI Technologies Inc Radeon R250 Lf [FireGL
> > 9000] (rev 02)
> >
> > If I use Xglx and compiz the effects are pretty smooth, desktop spinning
> > is jerky if the system is busy but the shadows, fades and so on are
> > great.
> >
> > However, if I try AIGLX and CVS HEAD Metacity, it is noticably slower:
> > dragging windows is lagging by a good quarter of a second.
> >
> > Is there anyway to determine the cause of this.  I've a feeling its
> > actually still using software indirect rendering, is there any way to
> > verify the hardware is being used?
> 
> Not possible; the aiglx branch in its current form will never dispatch into 
> libGLcore.
> 
> I suspect if you were to run sysprof or similar you'd find all your time being 
> spent in pushing texture images to the card.  Right now the aiglx branch is 
> pretty much limited to the bus write bandwidth, which is lame.  What we 
> really need is the ability to texture directly from offscreen-but-on-card 
> pixmaps, rather than what we do right now which is render them in host memory 
> and then blast across.
> 
> I'll try to hack up something to do that in the next few days if I get a spare 
> moment.

But updating textures should be irrelevant when dragging windows around,
right?

Could damage bugs be causing excessive texture updates? luminocity
needed some hacks to the X server to work around excessive exposes being
generated by DAMAGE.

Regards,
						Owen