[Bf-committers] OpenGL accelerated Nodes

Timothy Baldridge tbaldridge at gmail.com
Sun Sep 9 03:59:59 CEST 2007


True, although I'd have to ask why on earth you'd be creating a mask
in this way that would have to be so sensitive.

But what you said about previews is right. I'm in the process of
writing a video editing program, and the plan I settled on, is to have
two pipelines. A "Performance" OpenGL pipeline for 90% of situations
and for previews, and a "Quality" pipeline that is mathematically
perfect (I.E, software 16bit floating point per pixel).

If the compositor used C++, you could create a software renderer
class, and then the OpenGL class could subclass that, and override any
functions it wanted to support, leaving the rest for software. In this
way, with a simple sub classing of the software renderer class,
support could be added in the future for Cg, CUDA, CELL, and Altivec.

Timothy



-- 
If you put tomfoolery into a computer, nothing comes out but
tomfoolery. But this tomfoolery, having passed through a very
expensive machine, is somehow ennobled and no one dares criticize it.
(Pierre Gallois)


More information about the Bf-committers mailing list