Wasn't it Slime who wrote:
>> At the moment I can only generate rectangular panels of these greebles.
>> I tried generating a rectangular panel and then using an intersection on
>> it to create a triangular patch, but the render time went through the
>> roof, presumably because it loses the bounding boxes on the little
>> pieces, so it has to test each ray against thousands of greeble units.
>> This image rendered in 1m 53s, but a similar model made from triangular
>> patches would take several days.
>
>Try intersecting each individual greeble with the sides of the triangle,
>instead of making one huge intersection.

Unfortunately, with this system, I don't keep a record of where the
individual greeble units are going to end up being placed. They're
created deep inside a recursive routine and can form part of a union
that is translated many levels higher in the recursion hierarchy.

Several copies of the same greeble are translated to different
locations, allowing the construction of repetitive patterns, so I can't
pass down the information about the translations from the higher levels
to the lower levels.


I think the way to go might be to look for a way to subdivide an
arbitrary triangle into a set of rectangular panels and then fill in the
bits round the edge with triangular greebles. Don't want to use
triangular subdivision everywhere, because that would end up with the
overall pattern having an odd triangular look to it. Don't want to just
slice the arbitrary triangle into lots of tiny squares, because I want
to end up with a range of greeble sizes.

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message