I've been working on a scene that's about ready for final render, but
one of the things in the scene is a CSG of about 8000 small objects, and
the autobounding is a LITTLE more liberal than it needs to be.. say, two
to three times the actual size of the CSG. If I'm not mistaken, this
means that for every ray that hits that box, it'll have to test against
all of those 8000 objects, even though most of the rays will never hit
any of them. In any case, the middle third of my render just crawls..
anyone have any suggestions for speeding this up? Would manual bounding
help? Is there some clever trick I can use?

-Xplo

Post a reply to this message

In article <39F33AC4.4A238A2E@unforgettable.com>, 
inq### [at] unforgettablecom wrote:

> anyone have any suggestions for speeding this up? Would manual bounding
> help? Is there some clever trick I can use?

The bounding of some objects, differences and intersections for example, 
is not perfect. If you have a complex CSG(I would consider 8000 objects 
"complex"), manual bounding is often a good idea.
Also, with some arrangements of large numbers of objects, you can make a 
heirarchial bounding "tree" by using nested unions of objects, this can 
eliminate a lot of bounding box testing. Someone else might have more 
details on this...

-- 
Christopher James Huff
Personal: chr### [at] maccom, http://homepage.mac.com/chrishuff/
TAG: chr### [at] tagpovrayorg, http://tag.povray.org/

<><

Post a reply to this message

"Xplo Eristotle" <inq### [at] unforgettablecom> wrote in message
news:39F33AC4.4A238A2E@unforgettable.com...
In any case, the middle third of my render just crawls..
> anyone have any suggestions for speeding this up? Would manual bounding
> help? Is there some clever trick I can use?
>
> -Xplo

I wish I had an answer for you - I spent a long time trying to figure this
out myself (back on 5/20/00, to be exact...) and Peter Popov and Mike
Williams and a couple of other folks provided some good advice. I was never
able to achieve acceptable results when adding radiosity and/or media to the
scene, however, even with manual bounding etc. Good luck, Xplo.

Regards,
Dave

Post a reply to this message

Xplo Eristotle wrote:
> 
> I've been working on a scene that's about ready for final render, but
> one of the things in the scene is a CSG of about 8000 small objects, and
> the autobounding is a LITTLE more liberal than it needs to be.. say, two
> to three times the actual size of the CSG. If I'm not mistaken, this
> means that for every ray that hits that box, it'll have to test against
> all of those 8000 objects, even though most of the rays will never hit
> any of them. In any case, the middle third of my render just crawls..
> anyone have any suggestions for speeding this up? Would manual bounding
> help? Is there some clever trick I can use?
> 
> -Xplo

If you are using any truncated cone (like cone{<0,0,0>1,<0,1,0>,0.5} )
be aware that there is a bug on the bounding box size (too big)
(for the previous cone, the box has a height of 2, not 1, and if
instead of 0.5 it is 0.95, the height is about 20 !!)

This bug is present in the official 3.1g
(and has been reported in p.b, so no more action needed until 3.5 !)

Post a reply to this message

Xplo Eristotle <inq### [at] unforgettablecom> wrote:
: Would manual bounding help?

  Try it.
  You can also bound smaller groups of object inside the whole CSG (besides
the global bound).

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):_;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

"J. Grimbert" wrote:

> If you are using any truncated cone (like cone{<0,0,0>1,<0,1,0>,0.5} )
> be aware that there is a bug on the bounding box size (too big)
> (for the previous cone, the box has a height of 2, not 1, and if
> instead of 0.5 it is 0.95, the height is about 20 !!)
>
> This bug is present in the official 3.1g
> (and has been reported in p.b, so no more action needed until 3.5 !)

The height is multiplied by 1/(1-rad2/rad1)? Strange.

--
David Fontaine  <dav### [at] faricynet>  ICQ 55354965
My raytracing gallery:  http://davidf.faricy.net/

Post a reply to this message

David Fontaine wrote:
> 
> "J. Grimbert" wrote:
> 
> > If you are using any truncated cone (like cone{<0,0,0>1,<0,1,0>,0.5} )
> > be aware that there is a bug on the bounding box size (too big)
> > (for the previous cone, the box has a height of 2, not 1, and if
> > instead of 0.5 it is 0.95, the height is about 20 !!)
> >
> > This bug is present in the official 3.1g
> > (and has been reported in p.b, so no more action needed until 3.5 !)
> 
> The height is multiplied by 1/(1-rad2/rad1)? Strange.
> 

No, normal : the bounding code was made before it was
possible to have a cone truncated, so it always included
the sharp extremety of the cone.

And the bounding code was not updated 
when the second radius has been introduced for the cone.

But as I said, the bug was found, and a correction made.
A shame it was discovered lately, but in the end, all will be ok.

Post a reply to this message

"J. Grimbert" wrote:

> > The height is multiplied by 1/(1-rad2/rad1)? Strange.
> >
>
> No, normal : the bounding code was made before it was
> possible to have a cone truncated, so it always included
> the sharp extremety of the cone.

Oh, that makes sense. I wasn't thinking that POV internally extended the cone.

--
David Fontaine  <dav### [at] faricynet>  ICQ 55354965
My raytracing gallery:  http://davidf.faricy.net/

Post a reply to this message