I just added a custom bounding box to a CSG union of mine. The union was a
few large, flat, rectangular boxes, piled on top of each other, each one
smaller than the one below it (it's supposed to be a platform with steps
going up to it from all sides). I bound it with one single box as wide as
the bottom step, and as tall as the top step. POV-Ray came up with a
message: "Parse warning: CSG union unnecessarily bounded". Nevertheless,
from two different camera views, I got a noticably faster render (9 seconds
faster at 2 min. 24 sec., and 10 seconds faster at 1 min. 36. sec. from a
different view). Is this a mistake by POV-Ray (3.6.1c) or are the speed
differences just coincidental? (I tried to the keep the computer doing
nothing else on my test renders.) What bounding box was used before I added
mine?

Post a reply to this message

If your steps ended up parallel to the axes, then I'd expect the
computed bounding slab to be identical to the bounding box you created.

If the steps ended up at an angle, then the computed bounding slab
should be the axis-aligned box that just contains it, which will be
larger than your bounding box.

You can draw the automatic bounding slab of your STEPS object like this:

box {min_extent(STEPS),max_extent(STEPS) pigment {rgbt <1,0,0,0.5>}}


Early versions of POV didn't have automatic bounding, so to achieve
reasonable render times people had to write their own bounding shapes.

When efficient automatic bounding was introduced, the automatic bounds
was often much more efficient than the manual bounds, so a warning was
added to advise people that it might possibly be faster without the
manual bounds. If your manual bounds are faster then feel free to ignore
that advice. POV can't know for sure if your bounds are better without
rendering both versions.

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message