Xgl and direct rendering

[Owen Taylor]

On Tue, 2006-02-28 at 19:23 +0100, Matthias Hopf wrote:
> On Feb 27, 06 15:18:34 -0500, Owen Taylor wrote:
> > It may be that the only interesting case is full-screen video, which
> > is the easiest case to design, since the compositing manager basically
> > just has to get out of the way.
> 
> It definitively isn't.

I didn't mean it's the only interesting case of video; rather it may be
the only case where the performance and quality of using overlays,
when available, outweighs their disadvantages.

> > This "get-out-of-the-way-I'm-going-fullscreen" protocol is also needed
> > for fullscreen 3D games with direct rendering.
> 
> This shouldn't be much of a problem, the composition manager should know
> when a single nontransparent window is shown fullscreen.

It's not a *complicated* protocol, but there are a couple of bits of
information that need to be transmitted both ways: the client needs to
know that it can't use the Xvideo for smaller windows, but it can
use it when going fullscreen. And since most WM's will display other
windows on top of fullscreen windows in some circumstances, the WM/CM
needs to know that there is something special about the fullscreen
window: that windows can't be put on top of it, or that windows can
be put on top but can't be alpha blended. (You don't want a fringe
of chroma-key around the edge.)

> [...]
> > With Xegl it gets even harder ... you'd need a GL extension with similar
> > capabilities to Xvideo or something like that.
> 
> Talked with many vendors and programmes, and all agreed that using pixel
> shaders is the way to go. Eventually, native support for YUV textures
> might go away in the future as well.

Certainly with newer and high-end cards, overlays become pretty
irrelevant.

						Owen