Margus Ramst wrote:
> 
> OK, just showing off my progress here.
> Fur can now be added to any object, irrespective of shape/position etc.
> But, TINSTAFL. Parse times skyrocket: this one took >15 minutes to
> parse. I can probably reduce this, but the memory hit remains. This
> scene has 85000 cones (~17000 hairs at 5 cones/hair for smooth
> curvature). Unless you have at least 128M RAM - and I don't - it's a
> real pain.
> You can see that hairs still sprout out in a regular grid; I'll fix this
> soon. The "baldness" at the top is partly intentional - hairs get
> shorter towards the top. Just for fun.
> When all is done (and there's sufficient interest) I'll post the source,
> ASAP, for your inspection. Beware, it has a load of parameters and I
> ain't gonna reduce them by much. Also, it is restricted to the
> Superpatch, as it uses trace()
> 
> Oh, and TINSTAFL = There Is No Such Thing As a Free Lunch
> 
> Margus

I was hammering away at a different approach to the problem last night.
I ripped off some code out of Paul T. Dawsons tree.inc file and adapted
it to sphere/cone generation. It's a recursive while loop that spawns
branches. By constraining the child to one branch and adding control
to the angle of the spawn it offered some good control over the shape
of the hair. One problem that remains, and a big one at that, is the
distributions of the object(s) over a non uniform topography or surface.
You seem to have figured that part out but requires a patch or two
to make that work. 

I achieved a 220,000 object render last night. Instead of the hoped for
individual strands it looked nearly like a solid object. And yes there
was a very large memory hit for this. I have 128 megs installed and it
consumed close to 195 megs. My swap kicked in after about 1.2 minutes
and did not stop until the render was over. Parse time wasn't too bad
at around 8 minutes and the render time for the near solid objects was
much shorter than this. It took longer to release the memory than it
did to do the trace.

 Much more work to be done on this and I may abandon all hope.

  For those who are interested but don't wan't to waste time perfecting
their own proceedures for this there is hope in sight for you.

  Chris Colefax is also working on this issue and he has informed me that
he is very close to having it finished and ready for release to the public.
Most of his final work has been in the optimizing of the code to reduce
the memery requirements and reduce object count. He admits there are limits
and people with older computers may have to wait until they can afford a
faster machine, with enough memory to handle the large number of objects
required. That is if they want to get any kind of realistic covrerage for
the object needing hair growth.

-- 
Ken Tyler

mailto://tylereng@pacbell.net

Post a reply to this message