Re: [AD] SF.net SVN: alleg:[13312] allegro/branches/4.9

[Thomas Fjellstrom <tfjellstrom@xxxxxxxxxx>]

On April 24, 2010, Peter Wang wrote:
> On 25 April 2010 11:40, Thomas Fjellstrom <tfjellstrom@xxxxxxxxxx> wrote:
> > On April 24, 2010, Peter Wang wrote:
> >> On 25 April 2010 10:39, Thomas Fjellstrom <tfjellstrom@xxxxxxxxxx> wrote:
> >> > I'm not sure I can see how xrandr info fetching can cause any
> >> > serious cpu use, and even then audio just shouldn't stutter these
> >> > days (least I find it hard to get audio to stutter even when both
> >> > my cores are pegged, amarok and X/KDE can be locked up solid, but
> >> > the audio will still play).
> >> 
> >> I think it has to do with analogue VGA connectors.  IIRC the
> >> flickering happens when the analogue outputs are used (even as
> >> secondary monitors). Maybe the information it needs is not readily
> >> available so it uses some brute force probing technique on those
> >> monitors.
> > 
> > I haven't noticed such a problem on any of my systems (intel gma vs
> > radeon vs gforce). So I'm not sure what's going on.
> 
> It doesn't seem to be always reproduceable (e.g. I can't reproduce it
> today!). If you search for "XRRGetScreenResources flicker" you'll find
> quite a lot of reports about it.
> 
> Here is a relevant thread and quote:
> http://thread.gmane.org/gmane.comp.freedesktop.xorg/30567
> 
>     "XRRGetScreenResources is an expensive call.  Always.  If gtk is
> calling it on every app startup they're absolutely insane."
> 
> although patches to mitigate the problem are proposed later in the
> thread. I don't know if something like that is in the latest xorg.
> But...
> 
> Looking at GTK, it looks like they switched to relying on xinerama
> information for a while, then restored randr _1.3_ support, using a call
> to
> XRRGetScreenResourcesCurrent instead of XRRGetScreenResources.
> http://svn.gnome.org/viewvc/gtk%2B/trunk/gdk/x11/gdkscreen-x11.c?view=log
> 
> Indeed, although I can't reproduce the flicking/stuttering today,
> with XRRGetScreenResourcesCurrent Allegro programs start up quicker.
> With XRRGetScreenResources I at least get a delay of maybe half a second.

That I do remember seeing. mmon init code would add at least 100ms to the 
startup time on my laptop.

> So *Current looks to be the solution.

Looks like a 1.3+ specific feature. Going to have to check to see if it 
works with the nvidia binary driver. Theoretically since Xorg has been auto 
upgrading all drivers to the latest XRandR protocol, and emulating the newer 
stuff on drivers that don't support it, it may be available.

> Peter
> 
> -------------------------------------------------------------------------
> -----


-- 
Thomas Fjellstrom
tfjellstrom@xxxxxxxxxx