|
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hi everyone,
I've got a question regarding bounding boxes. I've just started using
POV-Ray again (the last time I did was a long time ago, back when programs
like DKBTrace and Vivid 1.0 were still around), and I decided to render a
(very small) version of the "Last Guardian" scene by Johnny Yip (which won
POVComp 2004) as a test.
This was successful, too, but I noticed two things: one, the render took
very long (about 8:20 hours on a P4/2.66 GHz/512 MB RAM, for a 120x160
picture); and two, the intersection stats that POV-Ray dumps at the end of
the render show that the success rates for many of the shapes are extremely
low, and that there are far more intersection tests with the actual shapes
than bounding box tests - about 35 billion vs. only 500 million (see
below).
Is this to be expected, or does it hint at possible bugs in POV-Ray's
automatic bounding box generation? Thanks. :)
--- Intersection test stats ---
Blob 2453 97 3.95
Blob Component 910 420 46.15
Blob Bound 12265 910 7.42
Box 147470815 35655239 24.18
Cone/Cylinder 1063684672 2824318 0.27
CSG Intersection 98485645 2242437 2.28
CSG Union 1332219932 968874 0.07
Height Field 16208116 52465 0.32
Height Field Box 16208116 258793 1.60
Height Field Triangle 254001 52833 20.80
Height Field Block 514866 153635 29.84
Height Field Cell 2996414 140677 4.69
Isosurface 13916160 1245999 8.95
Isosurface Container 32828694974 13926910 0.04
Isosurface Cache 235795 45256 19.19
Lathe 1252 296 23.64
Lathe Bound 1252 717 57.27
Plane 22451753 15033221 66.96
Prism 112560 37668 33.46
Prism Bound 725170 432584 59.65
Sphere 24797823 3543791 14.29
Sphere Sweep 205080375 268461 0.13
Torus 49374207 152711 0.31
Torus Bound 49374207 193932 0.39
Bounding Box 533615869 127430453 23.88
Light Buffer 2709788124 1107570558 40.87
Vista Buffer 16121912 13135844 81.48
--- Intersection test stats ---
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Is this to be expected, or does it hint at possible bugs in POV-Ray's
> automatic bounding box generation? Thanks. :)
As far as I know, that scene used a lot of CSG and blobs and such.
Especially CSG is hard to bound automatically. It's not really a bug, but a
known draw-back of using bounding boxes. This has become a much larger issue
lately, now that PC's can handle a lot more than in the early days, and
bounding boxes aren't always efficient in terms of actually bounding an
object (as mentioned above, especially when lots of CSG is involved). Have
you tried +ud in the commandline? If I'm not mistaken, that shows the
bounding boxes, and you could have a look at the troubles for yourself. :-)
--
"Tim Nikias v2.0"
Homepage: <http://www.nolights.de>
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Tim Nikias" <JUSTTHELOWERCASE:timISNOTnikias(at)gmx.netWARE> wrote:
> Have
> you tried +ud in the commandline? If I'm not mistaken, that shows the
> bounding boxes, and you could have a look at the troubles for yourself. :-)
I'll give that a try, thanks! :)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|