[cairo] Re: [Mesa3d-dev] Points in Cairo Glitz paper

[Peter Nilsson]

Ok, this is what went down with glitz. When we first started the work
behind glitz we wanted to see if we could implement a library after
Render semantics that is fairly well accelerated by hardware, using
OpenGL. We have tried to let people know what's been going on with the
development process, but no real opinions about the solution have
reached us up until now.

We shouldn't view Glitz as a being a final solution right now, more like
a playground for trying to find the best possible way to go. Just
because the paper is written about the current implementation doesn't in
any way mean that design is frozen. Some changes to Cairo has been made
recently to make it work better with glitz. More changes could be made
if they are motivated by better performance / hw-support without
comprimizing the fundamental ideas behind Cairo's design.  

This is what we have to play with. All creative criticism and new ideas
are much appreciated, and we will keep trying to make the best possible
OpenGL support for Cairo within the rules of the game. This offcourse
cannot be done without experts like all of you guys now raising your
voices, so thanks, keep the good stuff coming. :)

- Peter

--
Peter Nilsson ( c99pnn at cs.umu.se ) 



On Thu, 2004-04-22 at 19:31, Jon Smirl wrote:
> Don't take this as a criticism of Cairo. Cairo is an excellent piece of work. 
> 
> The existence of the 3D hardware pipeline is just a fact of life that all of us
> have to live with. It is entrenched enough that it won't be significantly
> changing in the foreseeable future. I'm not suggesting that Cairo follow the
> OpenGL API but that it should take this hardware into account. The fact is we
> only have one accelerated hardware model and Cairo simply needs to decide to
> make full use of it or not. 
> 
> For example, I'd much rather describe surfaces as mathematical equations, but
> you can't accelerate that. Instead I'm forced to use the much more complicated
> system of triangle meshes since they can be accelerated.
> 
> --- Carl Worth <cworth at east.isi.edu> wrote:
> > On Thu, 22 Apr 2004 08:41:28 -0700 (PDT), Jon Smirl wrote:
> > > Should we step back and reconsider the API and see if the need for pbuffers
> > can
> > > be eliminated?
> > 
> > API suggestions are always welcome. All cairo API decisions have come
> > about through discussion on the cairo mailing list.
> > 
> > If someone can point at something in the public cairo API that is
> > intrinsically "hard" for OpenGL, then we'll be glad to take a closer
> > look at it.
> > 
> > > My tendency would be to start from the 3D hardware model and work
> > > backwards to the 2D API. I agree that this will probably result in a more
> > > complex 2D API but the performance gains from doing this are worth it.
> > 
> > That's a possible approach. But that's not cairo.
> > 
> > -Carl
> 
> =====
> Jon Smirl
> jonsmirl at yahoo.com
> 
> 
> 	
> 		
> __________________________________
> Do you Yahoo!?
> Yahoo! Photos: High-quality 4x6 digital prints for 25
> http://photos.yahoo.com/ph/print_splash
> 
> _______________________________________________
> cairo mailing list
> cairo at cairographics.org
> http://cairographics.org/cgi-bin/mailman/listinfo/cairo