In the new beta the behavior of the maximum trace level has been changed
a bit. Now the trace level is *not* incremented for objects which do not
have reflection nor ior (but which may be transparent or semi-transparent).

  This means that now things like media containers and other transparent
objects (which do not have reflection nor ior) do not require a larger
max_trace level to work. This allows having numerous transparent containers
(and other such objects) and reflecting/refracting objects using a smaller
max_trace_level, usually making the rendering faster. There's basically
no limit on how many transparent objects (w/o reflection/ior) you can have
one in front of the other.

  As this is an experimental new feature it would require extensive testing.
I would appreciate if the community would help me testing this thoroughly.

-- 
                                                          - Warp

Post a reply to this message

Warp nous apporta ses lumieres en ce 2007/08/31 06:57:
>   In the new beta the behavior of the maximum trace level has been changed
> a bit. Now the trace level is *not* incremented for objects which do not
> have reflection nor ior (but which may be transparent or semi-transparent).
> 
>   This means that now things like media containers and other transparent
> objects (which do not have reflection nor ior) do not require a larger
> max_trace level to work. This allows having numerous transparent containers
> (and other such objects) and reflecting/refracting objects using a smaller
> max_trace_level, usually making the rendering faster. There's basically
> no limit on how many transparent objects (w/o reflection/ior) you can have
> one in front of the other.
> 
>   As this is an experimental new feature it would require extensive testing.
> I would appreciate if the community would help me testing this thoroughly.
> 
Test going: 2500 transparent spheres in a cluster.
No global_settings section. So far so good. Some areas have more that 200 
spheres to cross.
Specular highlights apears on surfaces separated from the light by over 20 
surfaces, seen trough about as many other surfaces.

-- 
Alain
-------------------------------------------------
Important letters that contain no errors will develop errors in the mail.

Post a reply to this message

I ran some render time tests on 10000 spheres
with various transmit values.
http://s201.photobucket.com/albums/aa280/Zakardis/?action=view&current=beta_test.jpg

It looks like the new default follows the curve for
simply increased max_trace_level.  But maybe that
is different with finishes.

The shadows looked speckled a bit, probably
from coincident surfaces in the test spheres.

Post a reply to this message

Actually my intention was not to ask "test with tons of objects put one in
front of the other, as many as you can, to see that it indeed works", but to
test that the feature didn't break any existing scenes (eg. by crashing the
program or causing visual artifacts). :)

-- 
                                                          - Warp

Post a reply to this message

>  Actually my intention was not to ask "test with tons of objects put one 
> in
> front of the other, as many as you can, to see that it indeed works", but 
> to
> test that the feature didn't break any existing scenes (eg. by crashing 
> the
> program or causing visual artifacts). :)

Well, yes. It seems not to have any problems crashing.

It does seem to make some speckling of shadows in some
cases where scenes in 3.6 did not.  Specifically
in 3.6 at mid transmit values the group of spheres
would appear to have shadows that were darker
and unspeckled, but in beta 22 there were speckles
that looked like coincident surface behavior.

http://s201.photobucket.com/albums/aa280/Zakardis/?action=view&current=beta22_36.png
vrs
http://s201.photobucket.com/albums/aa280/Zakardis/?action=view&current=beta22_37.png

Performance was up to seven times slower than 3.6 default.
It was better than twice as fast as compared to 3.6 with
max_trace_level increased.  Most interesting to me, performance
wise, is that beta 22 default and beta 22 with max_trace_level
increased performed almost nearly identical to each other.
Maybe the default max_trace_level could be raised a bit
without impacting performance?

Misc. editor issues
-------------------
In WinXP with 3.7 (with render window | get focus on
creation checked),
the render window now only gets focus on creation of the
render window,  successive renders do not get focus,
this is a change in behavior from earlier versions...
perhaps this deserves an extra checkbox? get focus?

Also if render window | keep above main is checked
then if the render window is minimized, (or if the render
window was minimized before "keep above main" was
checked) then it is impossible to return focus to the render
window, it's forced under the main window, instead of over it.
You need to uncheck keep on top to get the render window
back.  In 3.6 if "keep above main" was checked then if
the render window was minimized it docked to the editor
window instead of to the windows taskbar (that way it
was still over the editor window).

Post a reply to this message

Tim Attwood <tim### [at] comcastnet> wrote:
> It does seem to make some speckling of shadows in some
> cases where scenes in 3.6 did not.

  Do the speckes disappear if you specify an ior or reflection for the objects?

  Could you build a minimal scene which demonstrates the problem?

-- 
                                                          - Warp

Post a reply to this message

Warp wrote:
>  In the new beta the behavior of the maximum trace level has been
> changed a bit. Now the trace level is *not* incremented for objects
> which do not have reflection nor ior (but which may be transparent or
> semi-transparent).

Will there be a new keyword for specifying the max desired number of 
transparent surfaces passed through (default=infinite)? If not, how can it 
be avoided that scenes that rendered fast before (with acceptable artifacts) 
will render extremely slow with the new behavior?

Rune
-- 
http://runevision.com

Post a reply to this message

>  Do the speckes disappear if you specify an ior or reflection for the 
> objects?

It looks like the speckles are there even with ior or reflection or with
max_trace_level boosted.

>  Could you build a minimal scene which demonstrates the problem?

Sure, this shadow looks banded by coincident surfaces in 3.7
but not in 3.6

#declare testmat = material { texture {
      pigment {color <1,0,0> transmit 0.5}
      // finish { reflection 0.1 }
   }
   // interior{ior 1.1}
};
camera { location  <1.0, 2.0, -4.0>
   direction 1.5*z
   right     x*image_width/image_height
   look_at   <1.0, -0.2,  0.0>
   angle 30 }
light_source {<-10, 10, -5> color rgb <1, 1, 1>}
plane {y,-1 pigment{rgb <1,1,1>}}
union {
   sphere {<0.0,0,0>,0.5 material{testmat}}
   sphere {<0.1,0,0>,0.5 material{testmat}}
   sphere {<0.2,0,0>,0.5 material{testmat}}
   sphere {<0.3,0,0>,0.5 material{testmat}}
   sphere {<0.4,0,0>,0.5 material{testmat}}
   sphere {<0.5,0,0>,0.5 material{testmat}}
}

Post a reply to this message

Warp wrote:
>   Actually my intention was not to ask "test with tons of objects put one in
> front of the other, as many as you can, to see that it indeed works", but to
> test that the feature didn't break any existing scenes (eg. by crashing the
> program or causing visual artifacts). :)
> 

You should know that posting "Please test this" in these newsgroups is 
tantamount to saying "Please try to break this" :)

-- 
...Ben Chambers
www.pacificwebguy.com

Post a reply to this message

Tim Attwood <tim### [at] comcastnet> wrote:
> >  Could you build a minimal scene which demonstrates the problem?

> Sure, this shadow looks banded by coincident surfaces in 3.7
> but not in 3.6

  It doesn't seem to be related to the max trace patch because disabling it
doesn't change the problem.

  You should make a proper bug report.

-- 
                                                          - Warp

Post a reply to this message