[cairo] Cairo GLES2.0 backend

[Alexandros Frantzis]

On Tue, Nov 16, 2010 at 05:47:45PM +0800, Eric Anholt wrote:
> > 
> > 1. Is someone currently actively working on the GL backend that is in
> >    master? What is the state of the backend?
> 
> I'm not really working on it, having too many other things on my plate.
> 
> I've mostly used it as a testbed for "is GL fast enough to do 2D
> drawing?"  The answer so far has been: the GL driver is slower than it
> should be, but after much work, it's much better than Xlib.

Yes, I have seen those results, they were quite impressive! I expect
(and hope) that the gains will be even better for systems with ES2.0
(which tend not to have optimized X drivers).

> 
> On my Ironlake graphics, it's failing 72/242 testcases.  Many those are
> due to an Ironlake GL bug with texture border color, some are due to
> radial gradients being flipped vertically (see my recent fix for
> linear), and some are due to filtering differences that the testcases is
> just too picky about.  So, it doesn't look too serious.

That is good to know, Thanks!

> >    However, at this point there is a lot of fixed-function code in the
> >    GL backend. Do you think it is necessary to maintain the current GL
> >    1.x compatibility?
> 
> No, I don't think so.  I wrote it initially while trying to get cairo-gl
> up quickly, and it was a mistake to have ever done so.  It's a major
> impediment to improving cairo-gl.  People with fixed-function hardware
> will be better served by the image backend regardless, I'm sure.

Those are my thoughts, too. Not having to deal with both GL 1.x and 2.x
will make the code much more clean and maintainable and improve the
chances of getting it optimized properly.

Thanks,
Alexandros