more on tearing with xpress 200m
Alex Deucher
alexdeucher at gmail.com
Thu Jul 7 19:23:39 PDT 2011
On Thu, Jul 7, 2011 at 8:02 PM, Andres Cimmarusti <acimmarusti at gmail.com> wrote:
> I have also noticed this tearing. 2D Responsiveness can be
> incredibly bad.
>
> I found out sometime ago that it could be compositing that was messing
> everything up. But I thought it was from the Desktop Environment point
> of view, that is, for example, gnome's 2 metacity trying to put some
> nice transitions when starting my computer (which sometimes worked,
> but almost always didn't). I tried using the low resource option in
> gconf, but that didn't help. So I switched to xfce 4.8 (Debian
> testing). All desktop effects are off by default. And the problem
> remained!
>
> So finally after reading more stuff, I decided to disable composite
> and aiglx in the xserver. I put a file:
> /etc/X11/xorg.conf.d/composite.conf with the following:
>
> Section "Extensions"
> Option "Composite" "Disable"
> EndSection
>
> Section "ServerFlags"
> Option "AIGLX" "off"
> EndSection
>
> After booting up again, performance was instantly better. I would say
> this is a better solution than giving up performance using Option
> "EXAVSync" "True". Desktop is very responsive and videos playing without
> tearing (both vlc and mplayer). Sadly the xorg log still shows aiglx
> is enabled, even though the same log also acknowledges my
> configuration file and disables it early on.
>
> So far, performance has been good and no problems, but have I truly
> disabled compositing? or is it still lurking around refusing to be shut
> off?
AIGLX has nothing to do with composite. You've disabled composite by
disabling the extension. You can also disable composite in your
window manager (e.g., metacity). EXAVSync will hurt performance since
all updates to the front buffer will stall the engine until scanout is
past the destination region. It's not really recommended. Ideally,
you'd use a GL compositor with vsync enabled.
Alex
More information about the xorg-driver-ati
mailing list