Radeon 9000 Pro corruption problem on PPC when DRI is enabled

Ari Entlich lmage11 at twcny.rr.com
Thu Aug 24 22:58:18 PDT 2006


On Thu, 2006-08-24 at 12:39 +0200, Michel Dänzer wrote: 
> On Thu, 2006-08-24 at 03:12 -0400, Ari Entlich wrote:
> > On Thu, 2006-08-24 at 02:55 -0400, Ari Entlich wrote:
> > > On Wed, 2006-08-23 at 18:34 +0200, Michel Dänzer wrote: 
> > > > On Wed, 2006-08-23 at 02:24 -0400, Ari Entlich wrote:
> > > > 
> > > > > > However, Composite doesn't work as well as I'd hoped (see
> > > > > > http://gentoo-wiki.com/Xorg_X11_and_Transparency#ATi_.2B_Xorg ).
> > > > > > Composite was one of the main reasons I switched to Xorg 7.1 (another
> > > > > > reason was staying up-to-date, of course), so I'd be a bit
> > > > > > disappointing
> > > > > > if I can't get it running at any better than non-accelerated speeds
> > > > > > (which is what I'm getting) :-/. Is this a known issue...?
> > > > 
> > > > Depends. You'll want to not enable backing store 
> 
> Have you tried that?

Yup.

> 
> > > > and use Option "AccelDFS". If that doesn't help, you can try playing with the
> > > > "MigrationHeuristic" and "FBTexPercent" options, and if all of that
> > > > doesn't help, you could finally try the exa-damagetrack xserver git
> > > > branch.
> > > 
> > > Hmm... The only one of those options that my system understands is
> > > "FBTexPercent", 
> 
> You'll need xf86-video-ati git for AccelDFS, but MigrationHeuristic was
> already in xorg-server 1.1. How exactly did you try it?
> 

Oh, sorry, since MigrationHeuristic isn't documented I assumed it wasn't
supported. I tried setting it to "always" and... Wow!!! It made a HUGE
difference! Opacity changes are WAY faster now. However, I'm now getting
some corruption... See
http://home.twcny.rr.com/fandaswebsite/decoration_corruption.jpg (it's
supposed to look like
http://home.twcny.rr.com/fandaswebsite/decoration_ok.jpg ). And here's
some more: http://home.twcny.rr.com/fandaswebsite/more_corruption.jpg .

> > > and setting it to 0 didn't help... I think this problem
> > > might be deeper than setting a few options. I'm not just getting
> > > sub-optimal accelerated performance, I'm getting completely
> > > non-accelerated performance. 
> 
> How can you tell?
> 
> > > I can disable DRI completely and get exactly the same speed (or an 
> > > unnoticeable amount slower).
> 
> Disabling the DRI doesn't disable 2D acceleration.
> 

Oops! Sorry 'bout that... I assumed composite uses the DRI :-/. Turning
RenderAccel off made it a lot slower :).

> 
> > And I'm also having the "OpenGL stuff in one window appears in front of
> > another window even if the OpenGL window is in the background"
> > problem...
> 
> That's because DRI rendering isn't redirected yet.

Hmm... Is this a driver-level issue? I have a friend who's using Xorg
7.0 with the proprietary nvidia drivers, and this doesn't happen for
him....

Thanks,
 Ari




More information about the xorg mailing list