SPPM renderer (CPU-only)

[Dade]

There is something I don't understand :

http://src.luxrender.net/lux/file/1e957 ... el.cpp#l36

BSDF_ALL | BSDF_DIFFUSE is the same as BSDF_ALL, so what is the need for BSDF_DIFFUSE ?

Does BSDF_REFLECTION | BSDF_TRANSMISSON | BSDF_DIFFUSE is more correct ?

Did i miss something ?:

BSDF_ALL = BSDF_REFLECTION | BSDF_TRANSMISSON
(just check the source)

BSDF_ALL | BSDF_DIFFUSE means all transmitted and reflected rays of type diffuse.

BSDF_ALL alone means nothing. You have to combine with on or more of the followings: BSDF_DIFFUSE, BSDF_GLOSSY, BSDF_SPECULAR.

This come from that changset dcaca6612b.

It appears (because of this explicit changset) that we want to ignore glossy component when a photon falls on a hitpoint radius, but I'm wondering when this component is taken into account ?

BSDF_GLOSSY is handled by the eye pass.

so :

a) Am I right thinking that BSDF_ALL and BSDF_ALL | BSDF_DIFFUSE is the same ?

Nope.

b) Is the correct behavior BSDF_ALL or BSDF_REFLECTION | BSDF_TRANSMISSON | BSDF_DIFFUSE ?

The correct behaviour is BSDF_ALL | BSDF_DIFFUSE.

P.S. wait to patch LuxBlend25 (or apply the patch only on a branch) because we need to release 0.8 first.

[J the Ninja]

P.S. wait to patch LuxBlend25 (or apply the patch only on a branch) because we need to release 0.8 first.

I know, I did this mainly for my own screwing around with it, figured some other people might want it too. (also, I want to find out what I did wrong on the surface integrator panel)

[jeanphi]

BSDF_ALL = BSDF_REFLECTION | BSDF_TRANSMISSON
(just check the source)

Guibou is right, BSDF_ALL = BSDF_ALL_REFLECTION | BSDF_ALL_TRANSMISSION, which does include BSDF_DIFFUSE, BSDF_GLOSSY and BSDF_SPECULAR.

Jeanphi

[J the Ninja]

Remember this?

SPPM has the same problem:

Car paint (Polaris Silber preset):


Glossy:


Both of those were with startradius=2.0, alpha=0.7, 50k photons per pass. Both rendered for about 5 mins. The car paint one wasn't clearing up at all though, it was just stuck like that the entire time.

[guibou]

Remember this?

SPPM has the same problem:

Car paint (Polaris Silber preset):


Glossy:


Both of those were with startradius=2.0, alpha=0.7, 50k photons per pass. Both rendered for about 5 mins. The car paint one wasn't clearing up at all though, it was just stuck like that the entire time.

It may come from the fact that carpaint have a glossy and a diffuse component, so there is hitpoint on glossy surface. (In fact I have exacly the same pattern when I force storing hitpoints on glossy surfaces)

Let me take the metro and check that @work.

[guibou]

Ok, I have found something, thank to jeanphi who suggest that carpaint may generate lot of NaN values.

Some photons, after interactions with the carpaint surface, have Not A Number flux value. Currently I set a little fix inside hitpoints

the diff for thoses who can't wait for carpaint. Please note that this fix may (does) introduce bias.

Code: Select all

# HG changeset patch
# User Guillaume Bouchard <guillaume.bouchard@liris.cnrs.fr>
# Parent 1e957ff77c761ca502e998f870fd029fe91876a1
Temporary fix for carpaint (and other ?) material.

This fix is a ugly hack, but it removes NaN values which may comes from
interaction between carpaint and photons in SPPM.

diff -r 1e957ff77c76 renderers/sppm/lookupaccel.cpp
--- a/renderers/sppm/lookupaccel.cpp   Tue Mar 08 03:41:39 2011 +0100
+++ b/renderers/sppm/lookupaccel.cpp   Fri Mar 11 18:12:05 2011 +0100
@@ -40,6 +40,12 @@
      XYZColor flux = XYZColor(sw, photonFlux * f * hp->eyeThroughput);
      // TODO: it should be more something like:
      //XYZColor flux = XYZColor(sw, photonFlux * f) * XYZColor(hp->sample->swl, hp->eyeThroughput);
-      osAtomicInc(&hp->lightGroupData[light_group].accumPhotonCount);
-      XYZColorAtomicAdd(hp->lightGroupData[light_group].accumReflectedFlux, flux);
+
+      // check for nan, especially from carpaint, TODO: this is an ugly fix,
+      // and the issue must be handled elsewhere
+      if(!flux.IsNaN())
+      {
+         osAtomicInc(&hp->lightGroupData[light_group].accumPhotonCount);
+         XYZColorAtomicAdd(hp->lightGroupData[light_group].accumReflectedFlux, flux);
+      }
}


You can see from the attached image that there is still lot of noise in the ball and that the color is a bit weird.

[guibou]

See the attached image with the fix I proposed on top of the page

Color are much better (sorry, I totally forgot to set tonemapping the same for the three images)

patch :

Code: Select all

# HG changeset patch
# User Guillaume Bouchard <guillaume.bouchard@liris.cnrs.fr>
# Parent e9bbd187259575997eb95efb167c97484521ab01
Does not account for glossy reflection on hitpoints when measuring BSDF.

diff -r e9bbd1872595 renderers/sppm/lookupaccel.cpp
--- a/renderers/sppm/lookupaccel.cpp   Fri Mar 11 18:02:54 2011 +0100
+++ b/renderers/sppm/lookupaccel.cpp   Fri Mar 11 18:11:51 2011 +0100
@@ -33,7 +33,7 @@
      if ((dist2 >  hp->accumPhotonRadius2))
         return;

-      SWCSpectrum f = hp->bsdf->F(sw, hp->wo, wi, false, BxDFType(BSDF_ALL | BSDF_DIFFUSE));
+      SWCSpectrum f = hp->bsdf->F(sw, hp->wo, wi, false, BxDFType(BSDF_TRANSMISSION | BSDF_REFLECTION | BSDF_DIFFUSE));
      if (f.Black())
         return;


[binarycortex]

See the attached image with the fix I proposed on top of the page

Color are much better (sorry, I totally forgot to set tonemapping the same for the three images)

That looks great guibou, but I see a weird black shadow (ring) around the gray luxball base (stand).

[J the Ninja]

See the attached image with the fix I proposed on top of the page

Color are much better (sorry, I totally forgot to set tonemapping the same for the three images)

That looks great guibou, but I see a weird black shadow (ring) around the gray luxball base (stand).

It's visible in the first patch image too.

[binarycortex]

See the attached image with the fix I proposed on top of the page

Color are much better (sorry, I totally forgot to set tonemapping the same for the three images)

That looks great guibou, but I see a weird black shadow (ring) around the gray luxball base (stand).

It's visible in the first patch image too.

Yes it is.

Guibou, I don't suppose you would like to take a look at EXPM now, since it exhibits the same behavior and is already biased.