In article <3899a328@news.povray.org>, "TonyB" 
<ben### [at] panamac-comnet> wrote:

> It's too slow for any practical use... at least on my computer. It chugs
> away at very slooow speeds, like the old media. Can't you do something to
> speed it up? I haven't been able to finish any of the demo files included.

What processor are you on? I have a 266 MHz PowerPC G3, and although it 
was slow, it wasn't too slow to be practical.
The demo scenes had a very high samples value, a lower value can lead to 
graininess but is much faster. Also, there are 3 different methods of 
shooting the rays: one is very slow for points far from the target 
object, but is the most accurate. Another one is faster, but doesn't 
give an even distribution of samples. The third one switches between the 
other two depending on where the point being evaluated is. It is faster 
outside the object bounding box and more accurate inside, but can 
sometimes leave artifacts around the boundary between the two types.

And I can't really do anything about the speed...it figures out how far 
it is from the parts of the object by shooting a bunch of rays at it. 
For simple things like spheres, this isn't a good method(spherical works 
fine), and many other objects could be approximated with function 
patterns, but it is needed for complex CSG's and other objects.
Someone else who knows more advanced math might be able to find a better 
way...

Something I just thought of-you could use the trace function to place 
blob components for my blob pattern on the surface of the object. If you 
get the placement right, you could get a pretty good approximation of 
the proximity pattern. And although the blob pattern is still 
un-optimized, it would be faster.

-- 
Chris Huff
e-mail: chr### [at] yahoocom
Web page: http://chrishuff.dhs.org/

Post a reply to this message