Recently, I tried running several instances of a very large render on
several different computers.  Where the images were spliced together, I was
left with lines in the shadowed areas.

I'm guessing that this has to do with different Radiosity calculations being
done on the different computers.  This must be a common problem, but I
didn't see anything in the documentation about it.

Is there a way around this?

-- 
Jeremy

Post a reply to this message

"Jeremy M. Praay" wrote:
> 
> Recently, I tried running several instances of a very large render on
> several different computers.  Where the images were spliced together, I was
> left with lines in the shadowed areas.
> 
> I'm guessing that this has to do with different Radiosity calculations being
> done on the different computers.  This must be a common problem, but I
> didn't see anything in the documentation about it.
> 
> Is there a way around this?

Don't use radiosity.

-- 
Ken Tyler

Post a reply to this message

On Tue, 25 Nov 2003 17:23:58 -0500, "Jeremy M. Praay"
<jer### [at] questsoftwarecom> wrote:

>I'm guessing that this has to do with different Radiosity calculations being
>done on the different computers.  This must be a common problem, but I
>didn't see anything in the documentation about it.

I think the docs mention it but maybe not the workaround.

The problem is that the radiosity cache (octree) is different on the
different machines because they render different sections of the
image. Hence the different lighting.

>Is there a way around this?

Yes. Have all the machines share the same radiosity data. This is
accomplished in two passes. In the first pass, for which you should
use the fastest machine you have, render the whole image (maybe in a
lower resolution and without antialiasing) and use always_sample on
and save_file. In the second pass, which you split across different
machines, use pretrace_start 1.0 pretrace_end 1.0 always_sample off
and load_file. This will load the same radiosity data on all nodes and
will minimize the amount of new samples they take, so you should end
up with properly lineable image slices.


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

"Peter Popov" <pet### [at] vipbg> wrote in message
news:aop8svg88fo6kpqeotvqil7li64dq1boqb@4ax.com...
> On Tue, 25 Nov 2003 17:23:58 -0500, "Jeremy M. Praay"
> <jer### [at] questsoftwarecom> wrote:
...
> >Is there a way around this?
>
> Yes. Have all the machines share the same radiosity data. This is
> accomplished in two passes. In the first pass, for which you should
> use the fastest machine you have, render the whole image (maybe in a
> lower resolution and without antialiasing) and use always_sample on
> and save_file. In the second pass, which you split across different
> machines, use pretrace_start 1.0 pretrace_end 1.0 always_sample off
> and load_file. This will load the same radiosity data on all nodes and
> will minimize the amount of new samples they take, so you should end
> up with properly lineable image slices.
>

Thanks!  I'll try that when I get a chance.

The effect wasn't very noticeable (luckily), since it was a
daylight+radiosity scene, but it was something that I noticed and I was a
bit disheartened, having spent 2 weeks running various pieces of the scene
on 3 different machines.

--
Jeremy

Post a reply to this message