POV-Ray : Newsgroups : povray.binaries.animations : wripple (124k bup) : Re: wripple (124k bup) Server Time
20 Jul 2024 11:15:36 EDT (-0400)
  Re: wripple (124k bup)  
From: Remco de Korte
Date: 2 Feb 2001 05:22:39
Message: <3A7A893D.1E61F9DA@xs4all.nl>
Marc-Hendrik Bremer wrote:
> 
> Remco de Korte schrieb in Nachricht <3A79D232.8C7C423C@xs4all.nl>...
> 
> >Thanks for the hint. I didn't think it would matter much but I expect if
> you
> >think about it would.
> 
> Well if I understand it right, the function must be calculated for every ray
> hitting the container object. Okay, there may be (no, there are) some cache
> strategies, but if you make a picture which is only partly occupied by a
> isosurface you can see that rendering slows down, when it comes to the
> container. Since you don't need the empty parts of the container, it's only
> wasted time.
> 
> > I was already worried about the speed.
> 
> I did not try it myself but maybe it would help to precompute some more
> constant terms in your final function. If you precompute things like w*pi/ph
> or add  #declare f1 = f1*pi/ph, less has to be computed at rendertime. I
> might be wrong with this but I don't think that POV really recognizes those
> constant terms in the function (but I would be glad if someone would tell me
> that I'm wrong).
> As far as I can see, the '1+' for the sin and cos part is not nessecary too,
> because it only levitates the Iso a bit. It would be intresting to test if a
> translated Iso with a simpler function is faster than a more difficult one,
> which is at the right place allready. I bet it is.
> 
> OTOH all this will not make that much difference IMHO, and there were some
> rumours that Isos will be 10 times faster in the next official version.
> Hopefully who ever is working on that is succesfull!
> 
> BTW: What are the rendertimes for this? How long took one frame?
> 
> Marc-Hendrik

Thanks again for the hints. I have tried some optimizing but with tweaking the
function things got messed up a bit and I didn't bother any more. I don't know
the eaxct render time because whenever I have such scenes that take longer then
a minute to render I start doing other things on the same system or go to bed to
see the results in the morning. In this case it was a combination, but I
remember something like half an hour per frame (320x240). That is, on a P233,
with lots of other stuff going on and render priority set at lower. Not very
useful information...
I may use this in a larger scene (animated) so I'm sure I'll optimize it a bit
in that case, but I'm also considering using heightfields.

Remco


Post a reply to this message

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