Im not too sure about bounding objects, but in the case of a difference from
a sphere it makes sense that a planar difference is more efficient than a
boxed difference.
This is because the plane only has 1 component; that is...plane exists if
y>0 and plane doesn't exist if y<0. So when cutting it out of a sphere only
one component of the sphere needs to be checked, in this case the y
component.
For a box; the sphere needs to be checked against all 3 components <x,y,z>.
Is this right?

387






Alain <aze### [at] qwertygov> wrote:
> Hughes, B. nous apporta ses lumieres ainsi en ce 15/06/2004 07:31... :
>
> >
> >I was going to concede about this until I decided to try a test of the two
> >ways.
> >
> >Yours parsed and rendered slower, than mine, (in version 3.6) over several
> >runs, rather consistently too. I put them into a while loop, making a
> >150X150 grid facing the camera. Output image resolution was 1280X1024 and
> >default AA used. I kept the number of objects down to prevent going to XP's
> >Page File, so it wouldn't interfere.
> >
> >plane = Parse: 1 to 2 seconds Render: 19 to 20 seconds, or averaging 21
> >seconds total.
> >
> >box = Parse: 2 to 4 seconds Render: 23 to 24 seconds, or averaging 27
> >seconds total.
> >
> >Perhaps imprecise, being so short-lived, but I believe it is showing the
> >introduction of the box is actually causing a loss of optimization or
> >something. I don't know the truth behind the numbers; it just looks like the
> >plane has an advantage, at least in this case.
> >
> >Bob H.
> >
> >
> >
> >
> OK. Aparently, when removing a plane, an infinite object, the bounding
> is limited to the original object, but when removing a box, a finite
> object, the bounding encompas both objects. Did you try both
> intersection and difference?
>
> Alain

Post a reply to this message