[Fluxus] fbo pixels commits

Dave Griffiths dave at pawfal.org
Thu Jan 29 10:47:15 PST 2009


On Thu, 2009-01-29 at 16:55 +0100, gabor papp wrote:
> > This is tricky isn't it... I haven't looked too closely at the new code,
> > but are you intending to kick off a new render for each pixel primitive?
> > You should be able to sequence things so the pixel primitives all get
> > rendered first, followed by the main scene - or are you doing this
> > already?
> i changed the Scenegraph RenderWalk function and added the render target
> as a state variable, the binding was done in the state apply method, and
> made an unapply function that unbinds the target. cameras seem a bit
> complicated, although i understand that they are attached to renderers.
> so each pixel primitives should create a new camera there, shouln't it?

Yes... I think we need to do something different here. The cameras in
the renderer are designed for rendering multiple views of the same
scene, what we really want is the pixel primitive to have it's own
completely separate scene - is that right? 

If so then they need their own scenegraphs as well as cameras, basically
they need to own a renderer object.

This would be recursive too - you could have a pixel primitive with it's
scene being rendered in another pixel primitive which is in the top
level scene. Does this work as far as OpenGL is concerned? What happens
if its a circular dependency? Feedback? :)

> i personally don't mind if pixel primitives are not rendered first
> automatically. it depends on the order of creation, isn't it?
> 
> i imagine that the camera of the pixel primitive is automatically set
> when the primitive is bound. maybe a new (with-pixels) macro could open
> a new state, bind the fbo, set the pixel camera at the beginning and set
> it back to the main camera at the end.

This is much nicer...

cheers,

dave




More information about the Fluxus mailing list