"Tony[B]" <ben### [at] panamac-comnet> wrote...
> >   On the other hand, calculating radiosity only where it's needed and
not
> > everywhere saves rendering time and memory.
>
> What if one were willing to invest the time and memory for calculating the
> radiosity of the whole scene? Wouldn't that be cool? I mean, you process
> once, and then you can do a flythrough throught any point of the scene as
> many times as you want to, without having to reprocess. I would do it. :)

How would this save any calculation time?  Let's say we go with the current
implementation, and set some options to make sure it doesn't take extra
samples if it doesn't need them.  We do one fly-through of the scene,
keeping the data along the way.  After that, we know that we've got all the
info we need for another fly-through, and not any extra.  We can then do
another fly-through very quickly.

Also, if we wanted to do a fly-through backwards at that point, we could do
that, too.  We might end up having to do more calculations to get the
back-sides of objects.  But think about the situation where you never do the
backwards fly-through.  If you had pre-calculated everything, you would have
done work that didn't need to be done.  You would have "wasted" computation
time, and the whole process would have taken _longer_ (assuming you're using
the same algorithm in both cases).

-Nathan

Post a reply to this message