[Intel-gfx] [ANNOUNCE] xf86-video-intel 2.6.99.903

[Jesse Barnes]

On Thu, 16 Apr 2009 18:59:50 +0100
Barry Scott <barry.scott at onelan.co.uk> wrote:

> Carl Worth wrote:
> > On Thu, 2009-04-16 at 17:56 +0100, Barry Scott wrote:
> >   
> >> The code you have to prevent tearing will only work without
> >> compositing. 
> >
> > Yes, I just verified that if I simply start xcompmgr then I get
> > tearing once again.
> >
> >   
> >> What is needed is a change to the compositing app to avoid the
> >> tearing. Which I think you might have exposed the hooks to with
> >> the sync to scan line stuff, so long as that is independent of XV.
> >>     
> >
> > Actually, I don't think there's anything exposed yet here that the
> > application can get at. So we'll need to think a bit about how to
> > get arbitrary applications to ask for and get tear-free behavior.
> >
> > But yes, the scanline-based waiting will be very useful, so we just
> > need to find a good way to let applications get at this.
> >
> >   
> The other feature you have that we might used to solve this problem
> is the "swap buffers"
> stuff.

I think we decided to use the scanline based delay in our
I830DRI2CopyRegion hook; that's what DRI2 uses to implement Mesa's
copysubbuffer extension, which is used by compositing managers to
update the front buffer.

Full screen updates and applications can still be done with
glXSwapBuffers, and that should be made tear free by my swapbuffers
implementation.

-- 
Jesse Barnes, Intel Open Source Technology Center