Florian Brucker <tor### [at] torfboldcom> wrote:
> Not sure about the wet sand, but those stones are fantastic! The
> polished ones are 100% real, some of the others look a bit to smooth for
> their texture, but they're great nevertheless.

Ok, I cheated very slightly with the smoothness, by rendering with
 +W1200 +H900 +A0.2 +AM2 +R2
then using the good old NetPBM package to scale up by 250%, then blur with a
simple 3x3 average convolution, then scale down to 1024 x 768.

> Now I'm just curious: Why did you use one single isosurface?

Because I could. :) Seriously, this started as an exercise in making large
sets of balls using a single isosurface for efficiency. The stone textures
& then the shapes just followed as a natural progression, since I'd just
been tidying up my stones reference chart.

Now the isosurface isn't generally one of POV's faster shapes, but it's very
efficient in terms of memory usage for complex objects. Using the isosurface
like this to make multiple things from one object, it can be more efficient
to trace than tracing a heap of separate objects, and it even gets
more efficient the more things there are! I just did some quick tests
earlier tonight (using +W256 +H256 -F -D), varying the W parameter, and
here are the run times:

 1/W  Time
  10: 28s
  20: 26s
  40: 23s
  80: 19s
 160: 17s

Remember, the number of stones is 1/(W*W), so 25,600 (tiny) stones took 17
seconds, compared to 100 stones in 28 seconds! I stupidly forgot to record
the momory usage. I'll leave that as 'an exercise for the reader'. :)

At the moment I'm rendering a scene with 4 overlapping rotated copies of the
isostone grid. It looks better than I expected, although the stones still
need more perturbing. I've added a bit of y-randomness in the positioning,
and I'm seriously thinking about a bit of randomness in the orientation,
but I just don't want them to start intersecting with each other.

Post a reply to this message