Sorry... I'm just really enamoured with using radiosity on this particular
scene.

When I added the spheres, render time went up from around 5 mins, to over 8
hours! I've no idea why the huge difference... any ideas? The spheres are
just a simple union of nested #while..#end loops.

Andy Cocker

---------------------------------------------------------------
www.mariner9.net

..... for my music and graphics.
---------------------------------------------------------------
'I spilled spot remover on my dog. He's gone now. '
'I went to a restaurant that serves "breakfast at any time."
So I ordered french toast during the Renaissance. '

 - Steven Wright.

Post a reply to this message

Attachments:
Download 'radballs.jpg' (20 KB)

Preview of image 'radballs.jpg'

In article <3a21f759@news.povray.org>, "Andy Cocker" 
<big### [at] mariner9fsnetcouk> wrote:

> When I added the spheres, render time went up from around 5 mins, to 
> over 8 hours! I've no idea why the huge difference... any ideas? The 
> spheres are just a simple union of nested #while..#end loops.

Well, it takes a lot more computational work to check against hundreds 
of spheres than to check against a couple boxes or planes. You might be 
able to help with this by using a nested bounding structure, you know 
what a quad-tree is? Basically, divide each wall of spheres into four 
unions, bounded individually. Divide those unions further into four 
each, bounded individually...etc. This will allow POV to skip testing 
the bounding boxes for most of the spheres. You will have to tell POV 
not to ignore your bounding, though, and you probably won't be able to 
use while loops to generate the walls, though they would be easy to do 
with a recursive macro.

-- 
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

Andy Cocker wrote:

> Sorry... I'm just really enamoured with using radiosity on this particular
> scene.

I for one "hate photorealism," but I too am so enamored by radiosity-using
scenes that I wonder what all those boring photorealistic scenes would look
like with a heavy dose of radiosity.

Post a reply to this message

Chris Huff wrote:
> 
> Well, it takes a lot more computational work to check against hundreds
> of spheres than to check against a couple boxes or planes. You might be
> able to help with this by using a nested bounding structure, you know
> what a quad-tree is? Basically, divide each wall of spheres into four
> unions, bounded individually. Divide those unions further into four
> each, bounded individually...etc. This will allow POV to skip testing
> the bounding boxes for most of the spheres. You will have to tell POV
> not to ignore your bounding, though, and you probably won't be able to
> use while loops to generate the walls, though they would be easy to do
> with a recursive macro.
> 

Sounds like an interesting idea, i also had such problems in my recent
superellipsoid scene, i just wonder how much faster it would be (anyone
experience with such structures?)

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
Homepage: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Christoph Hormann wrote:
> 
> Sounds like an interesting idea, i also had such problems in my recent
> superellipsoid scene, i just wonder how much faster it would be
	Lots, I already got a tenfold speed increase simply by adding an
individual bounding box for each line in a brick wall...


-- 

* Abandon the search for truth, * mailto:ber### [at] inamecom
* Settle for a good fantasy.    * http://www.enst.fr/~jberger
*********************************

Post a reply to this message

> 
>         Lots, I already got a tenfold speed increase simply by adding an
> individual bounding box for each line in a brick wall...
> 


That sounds promising, i will probably give it a try.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
Homepage: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

"Andy Cocker" <big### [at] mariner9fsnetcouk> wrote...
> Sorry... I'm just really enamoured with using radiosity on this particular
> scene.
>
> When I added the spheres, render time went up from around 5 mins, to over
8
> hours! I've no idea why the huge difference... any ideas? The spheres are
> just a simple union of nested #while..#end loops.

The primary reason is most likely that POV can't re-use as many radiosity
sample points.  It will have to take samples at more points because of the
curvature of the surfaces.  The "quality" of a sample point is determined by
how similar that point is to the current spot, looking at the position and
surface normal, among other factors.  If the point is not good enough for
the error_bound value, then more samples are taken, which slows things down.
When the surface normal changes a lot (like on a sphere), more samples have
to be taken.

More samples are also taken inside crevices, where diffuse lighting can
change drastically over a short distance.

This sample-reuse works best on flat surfaces, where the amount of diffuse
illumination doesn't change much from one point to the next.

-Nathan

Post a reply to this message