0000942: Mix materials in hybrid make whole scene blown out

(0002810)
jeanphi (administrator)
2011-03-23 06:00
edited on: 2011-03-23 06:02

That issue might be fixed after my last commit. I've actually tested my latest fix with such a use case. Some shading issues remain but they are unrelated to the original issue.

(0002811)
Dobz (reporter)
2011-03-23 13:58

I'll test later today.

(0002823)
jeanphi (administrator)
2011-03-24 14:06

There's a remaining issue related to the handled of pass through materials, but I don't think it is related to the mix material.

(0002826)
Dobz (reporter)
2011-03-25 00:34

Issue is still present, the scene fades into the sky if any single object has a mix material.

(0002829)
Dade (developer)
2011-03-25 03:10
edited on: 2011-03-25 03:12

I was looking into the MiS code today and I have found a line that looks suspicious:

line 453 at http://src.luxrender.net/lux/file/57e02219618e/integrators/path.cpp [^]

Le *= PowerHeuristic(1, state->bouncePdf, 1, pdf * DistanceSquared(state->pathRay.o, ibsdf->dgShading.p) / AbsDot(state->pathRay.d, ibsdf->ng));

Isn't "DistanceSquared(state->pathRay.o, ibsdf->dgShading.p)" just INFINITY/MAX_FLT and totally screwing up MiS ?

I can only browse the sources from where I'm at the the moment so I can not check if the problem is really there.

(0002830)
jeanphi (administrator)
2011-03-25 03:37

No, that's not the issue. ibsdf->dgShading.p is a point on the scene bounding sphere or a disc tangent to this sphere.
I tested the issue with an object that had a mix with null material, if I remove the special casing for pass through materials, the result looks much more reasonable, but is still incorrect. I might have an idea why, but need to check: it might be that direct lighting is applied several times to the last non pass through point when going through a pass through surface.

(0002889)
jeanphi (administrator)
2011-04-05 09:46

Same issue than 941, it is now fixed.

Issue History
Date Modified	Username	Field	Change
2011-02-16 05:15	Dobz	New Issue
2011-02-16 05:15	Dobz	File Added: hybrid.7z
2011-02-22 05:36	jeanphi	Target Version	=> 0.8RC2
2011-02-24 02:21	Dobz	Description Updated	View Revisions
2011-03-18 08:29	Lord Crc	Relationship added	related to 0000941
2011-03-19 05:59	jeanphi	Target Version	0.8RC2 => 0.8RC3
2011-03-23 03:54	Dade	Assigned To	=> Dade
2011-03-23 03:54	Dade	Status	new => assigned
2011-03-23 06:00	jeanphi	Note Added: 0002810
2011-03-23 06:02	jeanphi	Note Edited: 0002810	View Revisions
2011-03-23 13:58	Dobz	Note Added: 0002811
2011-03-24 14:06	jeanphi	Note Added: 0002823
2011-03-25 00:34	Dobz	Note Added: 0002826
2011-03-25 03:10	Dade	Note Added: 0002829
2011-03-25 03:11	Dade	Note Edited: 0002829	View Revisions
2011-03-25 03:12	Dade	Note Edited: 0002829	View Revisions
2011-03-25 03:37	jeanphi	Note Added: 0002830
2011-04-05 09:46	jeanphi	Note Added: 0002889
2011-04-05 09:46	jeanphi	Status	assigned => resolved
2011-04-05 09:46	jeanphi	Fixed in Version	=> 0.8RC3
2011-04-05 09:46	jeanphi	Resolution	open => fixed
2011-06-06 08:20	jeanphi	Status	resolved => closed

Notes
(0002810) jeanphi (administrator) 2011-03-23 06:00 edited on: 2011-03-23 06:02	That issue might be fixed after my last commit. I've actually tested my latest fix with such a use case. Some shading issues remain but they are unrelated to the original issue.

(0002811) Dobz (reporter) 2011-03-23 13:58	I'll test later today.

(0002823) jeanphi (administrator) 2011-03-24 14:06	There's a remaining issue related to the handled of pass through materials, but I don't think it is related to the mix material.

(0002826) Dobz (reporter) 2011-03-25 00:34	Issue is still present, the scene fades into the sky if any single object has a mix material.

(0002829) Dade (developer) 2011-03-25 03:10 edited on: 2011-03-25 03:12	I was looking into the MiS code today and I have found a line that looks suspicious: line 453 at http://src.luxrender.net/lux/file/57e02219618e/integrators/path.cpp [^] Le = PowerHeuristic(1, state->bouncePdf, 1, pdf DistanceSquared(state->pathRay.o, ibsdf->dgShading.p) / AbsDot(state->pathRay.d, ibsdf->ng)); Isn't "DistanceSquared(state->pathRay.o, ibsdf->dgShading.p)" just INFINITY/MAX_FLT and totally screwing up MiS ? I can only browse the sources from where I'm at the the moment so I can not check if the problem is really there.

(0002830) jeanphi (administrator) 2011-03-25 03:37	No, that's not the issue. ibsdf->dgShading.p is a point on the scene bounding sphere or a disc tangent to this sphere. I tested the issue with an object that had a mix with null material, if I remove the special casing for pass through materials, the result looks much more reasonable, but is still incorrect. I might have an idea why, but need to check: it might be that direct lighting is applied several times to the last non pass through point when going through a pass through surface.

(0002889) jeanphi (administrator) 2011-04-05 09:46	Same issue than 941, it is now fixed.