Feedback from xf86-video-intel exa branch

[Wang Zhenyu]

On 2006.07.19 14:34:57 +0000, Daniel Kasak wrote:
> I've been testing out the exa branch of the xf86-video-intel driver (
> against xorg-server-1.1.1 ).

Thanks for testing.

> 
> Firstly, a bug report ...
> 
> If I *don't* enable exa, I get visual corruption in many windows ( but
> not all ) when I scroll, move things over them, etc. Examples of apps
> that demonstrate this behaviour are firefox, thunderbird, vmware. An app
> that *doesn't* seem to be affected is Eterm. If I enable the composite
> extension and start xcompmgr, this problem is fixed! Also of note, is
> that enabling the composite extension doesn't seem to be *that* bad on
> performance ... with exa disabled. The corruption occurs whether the
> composite extension is enabled or not.

Do you see this corrupt happen on other branch of intel driver? or just
from exa branch? 

> 
> Next ... exa. If I enable exa and the composite extension, performance
> is horrible. It feels like the out-of-memory problems I used to have on
> a radeon with exa if I opened too many windows. To be fair, this is an
> i845G with something like 12MB of memory ( shared too ), so I suppose
> this is to be expected :( Anyway, compositing works amazingly well
> *without* exa. The above visual corruptions are fixed with exa enabled (
> with *or* without the composite extension being enabled ).
> 

Current exa render implement is still very experimental, many area needs 
to be fixed, e.g exa mem problem, you may refer
https://bugs.freedesktop.org/show_bug.cgi?id=7524, as Eric pointed some
future improvement.