Micha Riser wrote:
> Rune wrote:
>> Maybe there's still hope. With only a 200% increase
>> in parse time we could calculate each point three
>> times, each with a different edge being the w edge.
>> Then average those three points.
>>
>> What? What's wrong with that??
>>
>> ...
>
> Try it, I could not bring this idea working. I do
> not really understand what causes the different
> sizes of the triangles.

I think I have a solution, and this time it might even work correctly...
;)

I did some #debug work and I found out that the w value is indeed
"skewed" and that changing the epsilon1 value to 1.0 makes it
"straight".

However, when u, v and w are all straight, the patch does not join up
smoothly with other patches. Making any one of the u, v or w value
skewed makes it smooth again, and it doesn't matter which of them it is.

So instead of having one of them completely skewed, I tried making all
of them 1/3 skewed and 2/3 straight. It seems to work. The patch seems
to be smooth, and the triangles are distributed evenly. Also, there is
practically no slow-down, as the averaging part is done to the u, v and
w values before the actual points are calculated.

I'll post the code when it's cleaned up.

Rune
--
3D images and anims, include files, tutorials and more:
rune|vision:  http://runevision.com (updated July 12)
POV-Ray Ring: http://webring.povray.co.uk

Post a reply to this message