"Light path"-"is Singular Ray" node Affects material visibility trough a glass material different way in two Blender versions #50805

New Issue

Ximo Blanes · 2017-02-27T20:19:05+01:00

Ximo Blanes commented

2017-02-27 20:19:05 +01:00

System Information
OS: Windows10 (x64)
CPU: Intel(R) Core(TN) i7-4790 CPU 3.60GHz
RAM: 16GB
GPU: NVIDIA Quadro K2000
Drivers: (NVIDIA) 376.84

Blender Version
Broken: 2.78c e9f235283
Worked: 2.78a e8299c8

Short description of error
There is a plane with an image that is not possible to view trough a glass object in version 2.78c but is viewable with 2.78.a version of Blender using the same file for rendering.

Exact steps for others to reproduce the error
There is a "Light path" node in the image material setted as "is Singular Ray" and connected in the "factor" input of a "Mix node" where are mixed the image with a "Diffuse BSDF". When the "Filter Glossy" option (into the "Render"/"Light paths") menu has any value (minimal or maximal 0.0001 or 10) the image is visible trough the Glass material object when rendering in 2.78a version but in the same condition it isn't visible when rendering in 2.78c version.
There is the same issue when I try doing the same in another PC with different CPU and GPU.
I'm not sure if it's a bug but there is a difference, then I tell you. It is quick to see it in a render in the screen preview.Transp-Glass Shader issue_x.blend

**System Information** OS: Windows10 (x64) CPU: Intel(R) Core(TN) i7-4790 CPU 3.60GHz RAM: 16GB GPU: NVIDIA Quadro K2000 Drivers: (NVIDIA) 376.84 **Blender Version** Broken: 2.78c e9f235283 Worked: 2.78a e8299c8 **Short description of error** There is a plane with an image that is not possible to view trough a glass object in version 2.78c but is viewable with 2.78.a version of Blender using the same file for rendering. **Exact steps for others to reproduce the error** There is a "Light path" node in the image material setted as "is Singular Ray" and connected in the "factor" input of a "Mix node" where are mixed the image with a "Diffuse BSDF". When the "Filter Glossy" option (into the "Render"/"Light paths") menu has any value (minimal or maximal 0.0001 or 10) the image is visible trough the Glass material object when rendering in 2.78a version but in the same condition it isn't visible when rendering in 2.78c version. There is the same issue when I try doing the same in another PC with different CPU and GPU. I'm not sure if it's a bug but there is a difference, then I tell you. It is quick to see it in a render in the screen preview.[Transp-Glass Shader issue_x.blend](https://archive.blender.org/developer/F498118/Transp-Glass_Shader_issue_x.blend)

Ximo Blanes commented

2017-02-27 20:19:05 +01:00

Changed status to: 'Open'

Ximo Blanes commented

2017-02-27 20:19:05 +01:00

Added subscriber: @ximobl

Ray molenkamp commented

2017-02-27 20:25:01 +01:00

Added subscribers: @Sergey, @LazyDodo

Ray molenkamp commented

2017-02-27 20:25:01 +01:00

@Sergey mind taking a peek here? can confirm 2.78a works, but 2.78c and master don't.

Ximo Blanes commented

2017-02-27 20:34:38 +01:00

Version 2.78a works fine with the nodes configuration as i said. Is the version 2.78c with the same config not working.
The 2.78c version was downloaded in http://download.blender.org/release/Blender2.78/
These are the two versions I've tested with the file I uploaded

Version 2.78a works fine with the nodes configuration as i said. Is the version 2.78c with the same config not working. The 2.78c version was downloaded in http://download.blender.org/release/Blender2.78/ These are the two versions I've tested with the file I uploaded

Sergey Sharybin commented

2017-02-28 09:54:56 +01:00

Short reply:

Think the real issue here is a lack of explicit mention of this change in the logs, which i did do now: https://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.78/Performance_update#Cycles

Full version:

This was a deliberate change in c4d6fd3. The reasoning here goes as:

Mathematically speaking, GGX/Beckmann with 0 roughness is a singular ray, since there is no distortion happening for the normal.
Some parts of Cycles were replacing GGX/Beckmann with Sharp distribution, when filter glossy was not used. This was causing different behavior of scenes after tweaking filter glossy or changing distribution. Such a change in behavior is really tricky to nail down in a real production movie shot scene.
We were preparing for further changes, such as D2416 which required unification of ray visibility across closures as much as possible.

I would really prefer to keep behavior of Sharp and GGX/Beckmann unified so we can apply various optimization techniques both on Cycles and scene and yet keep things more consistent for artists.

So this is something we would live with. You can quite easily work that around by setting GGX distribution to a low non-zero value like 0.001.

Thanks for the report, closing it now.

Short reply: Think the real issue here is a lack of explicit mention of this change in the logs, which i did do now: https://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.78/Performance_update#Cycles Full version: This was a deliberate change in c4d6fd3. The reasoning here goes as: 1. Mathematically speaking, GGX/Beckmann with 0 roughness is a singular ray, since there is no distortion happening for the normal. 2. Some parts of Cycles were replacing GGX/Beckmann with Sharp distribution, when filter glossy was not used. This was causing different behavior of scenes after tweaking filter glossy or changing distribution. Such a change in behavior is really tricky to nail down in a real production movie shot scene. 3. We were preparing for further changes, such as [D2416](https://archive.blender.org/developer/D2416) which required unification of ray visibility across closures as much as possible. I would really prefer to keep behavior of Sharp and GGX/Beckmann unified so we can apply various optimization techniques both on Cycles and scene and yet keep things more consistent for artists. So this is something we would live with. You can quite easily work that around by setting GGX distribution to a low non-zero value like 0.001. Thanks for the report, closing it now.

Sergey Sharybin commented

2017-02-28 09:55:03 +01:00

Changed status from 'Open' to: 'Archived'

Sergey Sharybin closed this issue

2017-02-28 09:55:03 +01:00

Sergey Sharybin self-assigned this 2017-02-28 09:55:03 +01:00

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

"Light path"-"is Singular Ray" node Affects material visibility trough a glass material different way in two Blender versions #50805