by Olivier3001@[EMAIL PROTECTED]
Nov 6, 2007 at 01:56 PM
On Nov 6, 7:51 am, Paul Melis <paul.me...@[EMAIL PROTECTED]
> wrote:
> So, is rendering actually a full scanline of buckets further along
> than the framebuffer display shows in most renderers? Or are more
> advanced schemes used to produce final bucket output sooner, based
> only on neighoring sample results?
In 3Delight we simply enlarge the first row/column of buckets to
include the required samples. All following rows and columns are thus
offset wrt the buckets actually displayed. So you see everything
rendered except the 2-3 extra rows of pixels needed for filtering. I'm
fairly certain prman (and probably most implementations) do roughly
the same thing. Well except for aqsis which I think likes to sample
the extra pixels for all buckets.
Olivier
