Nicolas Alvarez wrote:
> I doubt I'd understand *any* of the maths involved :)
> 
> (I've yet to solve t out of vlength(center - start+direction*t) = radius,
> what are my chances with radiosity?)

The issue isn't likely to be with the maths per se; that is written and working.
The problem is more to get it working properly with the multi-threading. This
involves two main thrusts: firstly, to get it running in parallel efficiently
(the pre-trace is a major problem here), and, secondly, to avoid the need to do
new samples during the actual render, since apart from problems with animations
(different render sequences of sub-blocks could lead to slightly different
lighting), we'd also have the nasty issue of efficient thread-safe access to the
tree holding the sample information ('efficient' being the key issue).

Christoph has done some investigation into this and we think we have a means of
avoid the run-time sampling, though due to man-power issues we've not gone into
radiosity much (and really we need to get other stuff working properly first).

-- Chris

Post a reply to this message