Tina S. schrieb in Nachricht <3ab10eed.32032954@news.povray.org>...
>On Thu, 15 Mar 2001 19:40:16 +0100, "Marc-Hendrik Bremer"
><Mar### [at] t-onlinede> wrote:
>
>>It seems to be an accuracy-problem for me. Sometimes you have to go really
>>low with this value (something like 0.000001). Your contained_by{}-box is
>>way to big and you get better results if it's fits better. And it will be
>>faster. (Throw that unnecessary '+0' out of the function, it slows down
the
>>iso, too - well a little bit :-))
>
>Hmm, I had the +0 in there because I stole that formula from a web
>page, but I'll happily discard it if it'll speed things up :).

Well, it is reall yjust a little speed up, something like 4:36 to 4:38 in my
testrender. But I have found out that it is good to precompute as much as
you can if it comes to isosurfaces. If you can't calculate the value with
the same precision as Pov does, use a variable, which is as not much slower
than a constant value. The function has to be evaluated over and over while
the iso is rendered, so make the computation as short as possible.
BTW x*x is faster than x^2.

>Turns out it was the contained_by box; thanks for pointing that out. I had
>changed it earlier to increase it since I had thought maybe it was
>getting cut off but I didn't try decreasing the size, which is what
>did the trick.

I don't know why it makes such a difference, since there isn't something
like samples, which are distributed over the container. But it's a good idea
to make it as small as possible for speed reasons.

>I still don't really know what I'm doing with these things, so I'm
>rarely sure why things work the way they do. I think I need to learn
>more math.

Math is useful for isos, but it does not take that much, if you don't want
to make those amazing shapes Tor Olav Kristensen, you will be lucky with
some easy shapes and an idea how to combine them.
Do you know the isosurface manual of the Smellenberghs already? If not, look
at http://users.skynet.be/smellenbergh/. It's a good source and a good
reference.

Marc-Hendrik

Post a reply to this message