Rendering an object from the inside

@normen said: @zarch: the fill rate is only increased by the backface rendering, you always increase the fill rate also when the object in front is not transparent.

JME sorts to the opaque bucket front to back to eliminate overdraw. On many cards (nVidia seems to be particularly good at this) this will lessen fill rate since it can often trivially reject fully occluded spans. This is my theory for why vertex lighting is slower on my machine than regular per-pixel lighting. Because my scene is vertex heavy but overdraw light.

And note: even if the spans are not trivially rejected then at least their fragments are never processed… so that also affects the impact of fill.

Some cards seem to be really bad at this. (I’m looking at you ATI.)

Oh nice, sounds like we actually have some renderer issues solved ^^
The content of this post is meant to be read as a straight information or question without an implicit dismissive stance or interest in having the other party feel offended unless there’s emotes that hint otherwise or there’s an increased use of exclamation marks and all-capital words.

What @pspeed said, although I didn’t know about the occluded spans thing :slight_smile:

@zarch said: What @pspeed said, although I didn't know about the occluded spans thing :)

It was only a guess. Back when I wrote software 3D renderers, span buffers were “cutting edge” and relatively lightweight. I always sort of hope they are still around in some form. :slight_smile:

At any rate, the fragments themselves can be trivially rejected since Z is already known.