At last!
I suppose I've probably spent more time thinking about doing this than
actually doing it, but here it is at last.  Only took me a couple days to
put it all together from scratch, but then my hard drive is littered with
failed attempts...
I posted a 2-d version of this a while ago and haven't worked on the project
at all since then, but this 3-d version is a big improvement in many ways. 
This is a 40x80x40 grid and only takes a matter of ten seconds or so for
each simulation step.  (well, this animation is only every fourth
simulation step so it isn't so slow, but it could be with much bigger time
steps.)  I didn't really feel like waiting, so there is no scattering here,
just emission and absorption since it takes longer to render than to
simulate.  It's entirely based on the Visual Simulation of Smoke paper
found at http://graphics.ucsd.edu/~henrik/papers/smoke/ .  This version
adds has three dimensions of course, a much quicker generalized
perconditioned conjugate gradient solver (so it shouldn't be too hard to
add viscosity and diffusion), and although it has the beginnings of
boundary conditions built in, I haven't set it up to use that at all.  This
time I might actually post the code, since more experience means I'm
starting to be more proud of the code than ashamed (have mercy--this really
is the only c++ project I've ever worked on! [except for some stupid little
cloth thing]).  That said, there sure isn't a GUI and changing any
parameters means recompiling.  Maybe if I get it a little more complete and
comment the heck out of it, someone else could help out.  Well, bedtime for
me.

 - Ricky

Post a reply to this message

Attachments:
Download 'fluid.mpg' (185 KB)