Dear Vesa, I can follow your notation very well ;-) / I will take care to produce a new scene at the week end, so I can make some new control points, not more nurbspoints to the nurbscurves and try to reduce it, I will see the result and perhaps I can change the rendering times. But I wonder why some other scenes with nurbscurves were rendered very well, not so easy to understand that, but I will learn with the further projects :-) Thank you very much and see you again, best regards, Frank Brübach
Von: [email protected]
Gesendet: 18.03.06 12:34:53
An:
Betreff: Re: Rendering times and nurbscurve picture ;-)
Hi,
One note about NURBS curve rendering: actually a low point density can make
rendering slower! A curve is split into segments before rendering. In the
usual cubic case, each segment includes 4 points, from which the actual
smooth shape is interpolated.
If the control polygon is very heavily bent to all directions, those 4 point
base segments can become quite large in all dimensions. That makes raytrace
optimizations harder (more curve pieces potentially travel across an
examined volume of space). If the point density is sufficient, the 4 point
NURBS segments become quite straight and 'thin' in some dimension(s).
This trick of increasing the point density works only when the original
curve has very low point density. For example, cubic NURBS circles (or
revolutions of a helix curve), which have only the minimal 4 points per 360
degrees are potentially slower than, say, 8 point circles. I would estimate
that even higher point counts ( over 10 points/revolution) no longer help
but start slowing down the rendering.
In the linear case, the lowest point count is naturally the fastest.
Best regards,
Vesa
|
|
