I compiled pvmega tonight and got some differences compared to normal
mega, look at the images (same
topic) in p.b.i. The first one is standard mega and the second one is
pvmega. 
I used the source from Kari Kivisalos post Realistic radiosity in
p.b.s-f
The image made by pvmega is brighter than the one from mega, most
noticable on the roof, but there
are also some more subtle differences, the small box in pvmega doesn't
have the same inderect lighting
on the forward facing side, it's actaullay darker, and has a sligh
reddish tint in the upper right
corner, mega has a little bit brighter lower left corner and darker
upper right then pvmega has.

There are also some slight differences in the shadow of the large box,
but it's not that noticable
but it's the shadow in the corner, that is, the shadow from the
innermost lefthand lower corner of 
the large box to the corner of the room, there is some differences. And
also on the large box the
lower corner closest to the camera, the red colorbleed is more noticable
than on the image by mega.

/Peter, who hopes anyone understands what he's trying to tell... *tired*

Post a reply to this message

Peter Toneby wrote:
> [about various differences]

The lighting on the small box and roof changed this way in megapov
when I changed pretrace_end, count, nearest_count and error_bound.
It seems that the pvmega image uses lower quality settings
(higher pretrace_end and error bound, lower count and nearest count).

K.K.

Post a reply to this message

Kari Kivisalo wrote:
> 
> Peter Toneby wrote:
> > [about various differences]
> 
> The lighting on the small box and roof changed this way in megapov
> when I changed pretrace_end, count, nearest_count and error_bound.
> It seems that the pvmega image uses lower quality settings
> (higher pretrace_end and error bound, lower count and nearest count).

I didn't pass any options other than +qr so it's all defaults, so you'd 
think that it should render somewhat more alike, unless pvmega uses an
old snapshot of mega, but I don't think so...

/Peter

Post a reply to this message

Peter Toneby wrote:
> 
> I compiled pvmega tonight and got some differences compared to normal
> mega, look at the images (same
> topic) in p.b.i. The first one is standard mega and the second one is
> pvmega.
> I used the source from Kari Kivisalos post Realistic radiosity in
> p.b.s-f
> The image made by pvmega is brighter than the one from mega, most
> noticable on the roof, but there
> are also some more subtle differences, the small box in pvmega doesn't
> have the same inderect lighting
> on the forward facing side, it's actaullay darker, and has a sligh
> reddish tint in the upper right
> corner, mega has a little bit brighter lower left corner and darker
> upper right then pvmega has.
> 
> There are also some slight differences in the shadow of the large box,
> but it's not that noticable
> but it's the shadow in the corner, that is, the shadow from the
> innermost lefthand lower corner of
> the large box to the corner of the room, there is some differences. And
> also on the large box the
> lower corner closest to the camera, the red colorbleed is more noticable
> than on the image by mega.

Sorry to answer that late. I'm just reading while being home for a day
or two in the middle of some nice holidays.

I haven't carried out some extensive testing, even about radiosity
parameters values. The main difference between mega and pvmega is that,
while the octrees should be identical after the preview passes, mega
still adds samples when necessary, while pvmega doesn't to avoid the
problems coming from having different octrees among nodes. This might
explain some differences. Some problems with parameters might come from
the fact that the splitting itself and the effect it has on sampling.
And you musn't forget that the master process does not parse the source
files, so it ignores the embedded ini settings.

I'll probably work on this in september, also trying to improve the
process management which is not decent at all for the moment.

Thanks for your post.

-- 

      __  __ __  __  _
|  | /  \  /  / |_  /  |/
\/\/ \__/ /_ /_ |__ \_ |\

Post a reply to this message