POV-Ray : Newsgroups : povray.binaries.images : ivy on a tree WIP : Re: ivy on a tree WIP Server Time
7 Aug 2024 23:19:24 EDT (-0400)
  Re: ivy on a tree WIP  
From: Alain
Date: 7 Feb 2006 20:00:23
Message: <43e942a7$1@news.povray.org>
Kenneth nous apporta ses lumieres en ce 06/02/2006 06:01:
> "Kenneth" <kdw### [at] earthlinknet> wrote:
> 
>>"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> 
> 
>>
>>I should have made a note of that, but didn't. Total render time (using
>>cylinders for vines) was between 3 and 4 hours on my 400MHZ  Pentium II,
>>using AA 0.2.  Parse time was a small fraction of that--perhaps 5-to-10
>>minutes at most.
> 
> 
> I have more accurate results now: Parse time is only about 2 minutes, using
> (sixty) vines of either type. But the scene with sphere_sweeps for vines
> takes an astoundingly long time to render--after 19 hours of a *partial*
> render, I did a quick calculation, and realized it would consume about 148
> hours total!!  But I also discovered that the reason wasn't just the
> sphere_sweeps...it was due to two "area" spotlights in the scene. Changing
> those to regular spotlights reduced the sphere_sweep render time
> drastically, to a mere five hours. Area lights of any kind do slow down a
> render, of course, but I don't really understand this area
> light/sphere_sweep interaction, why it's so MUCH slower. (Thinking it might
> be a bounding-box problem, I tried manually bounding each vine; but that
> didn't improve the render time at all.)
> 
> Ken
> 
> 
You may try choping the sphere_sweep into smaller pieces. A sphere_sweep have a prety
large bounding 
box that is mostly empty. As you have 60 sweeps in the same area, all bounding boxes
pile up leading 
to a very ineficient situation. By subdividing it, you get smaller, tighter, bounding
boxes that 
overlap less with each other.
When you test to see if you are in the shadow, you first check for the bounding box,
you hit 60 of 
those for EACH test. Then, for each bounding box you encounter, you need to evaluate
the whole sweep 
to see if it does cast a shadow on that particular point. In the case of area_light,
you have to 
multiply that by the number of samples that you need to take. "adaptive" is a great
time saver, 
start at adaptive 0 and increase by 1 as needed.

-- 
Alain
-------------------------------------------------
For THIS I bought a computer?


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.