Le Tue, 02 Jun 2009 15:32:07 +0200
Roland Scheidegger <[email protected]> a écrit:

> >> If there are other chipsets that cannot do true separate alpha blend, we 
> >> can always use PIPE_CAP to let Mesa know, and Mesa can do the fragment 
> >> program alteration itself. (r3xx and r4xx also have this, so it might be 
> >> useful later on.)
> > 
> > I did more tests. I think I was wrong, in fact, when using
> > BlendEquationSeparate functions, no triangle rendering command is sent
> > to the gpu, so I suppose it's all software rendering. The shader
> > program is a single instruction to set current color, plus it resend
> > all tcl engine state.
> > 
> > But it's good news if a fragment program can do this instead, so adding
> > the PIPE_CAP should be useful, if other cards need it also.
> 
> How do you do this with a fragment shader?
> I think if you wanted to do this, you'd need to redirect all normal
> rendering as opaque to a temporary surface. Then, you could use a
> fragment shader for doing the blend (using the temporary surface and the
> destination surface as textures), outputting to another temporary
> surface. Then copy back the 2nd temporary to the original destination.
> Maybe...

Well, my last question is still valid though. If no fragment program
work can do it easily, is it a good idea to have a PIPE_CAP for it? so
Mesa knows how to deal with it, even falling back to software rendering?

> Not sure though why r3xx gets mentioned here, should easily handle this
> (even r200 can do it...)

-- 
Patrice Mandin
WWW: http://pmandin.atari.org/
Programmeur Linux, Atari
Spécialité: Développement, jeux



------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables unlimited
royalty-free distribution of the report engine for externally facing 
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Mesa3d-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

Reply via email to