Page 12 of 17

Re: Test scences GPU Path vs CPU BiDir

Posted: Mon Aug 27, 2018 6:41 am
by FarbigeWelt
B.Y.O.B. wrote:
Sun Aug 26, 2018 6:39 pm
If you use an intersectible light source (e.g. an area light), Path engines will also be able to render the caustics (with much worse performance of course, compared to Bidir).
You are right. It is possible. Result depends on the size of the area light. Too large or too small does not work well.
GPU Path 300 Samples Area Light
GPU Path 300 Samples Area Light

Re: Test scences GPU Path vs CPU BiDir

Posted: Mon Aug 27, 2018 6:51 am
by FarbigeWelt
wasd wrote:
Sun Aug 26, 2018 7:16 pm
FarbigeWelt wrote:
Sun Aug 26, 2018 12:19 am
Glossy translucent is a rather complex material.
It's not complex. It just isn't translucent. The only tricky part in it is how with rather thin volume and absolutely smooth surface it scatters light so evenly.
Here be the glossy translucent material transmitting 97% light on the left and glass on the right.
Both has heterogeneous volume with 0.001 scattering scale.
I am not sure yet if glossy translucent depends on inner material. Texture applies at all surfaces, there is the option missing apply once and use transparent if uv map is out of mesh.
Glossy translucent
Glossy translucent

Glossy transluscent 2. Study

Posted: Mon Aug 27, 2018 10:22 pm
by FarbigeWelt
Glossy transluscent does not depend on inner material, at least it does not on clean volume and its IOR. And in contrast to other materials you can have an inner material and a textbox for IOR on the main node together.
One part of the texture is made from the image and and another from alpha only. The uv meshes are so aranged that only one side shows the image the others show just an empty, white face. This means alpha is not required here. Due to the very thin cuboid one cannot see much of the four faces. The thicker the cuboid the fuzzier the image is on the backside. This material behaves as one expects from a transluscent body. There is the mirrored image at the backside. What I miss is the setting for the specular light and the scattering.
Glossy Tranluscent LuxCoreRender Quad View
Glossy Tranluscent LuxCoreRender Quad View
Quad view does not like openCL, just kidding, openCL is not liking my glossy translucent material in view port but in final render it does. :o
Glossy Tranluscent LuxCoreRender Quad View openCL
Glossy Tranluscent LuxCoreRender Quad View openCL

Re: Test scences GPU Path vs CPU BiDir

Posted: Mon Aug 27, 2018 11:02 pm
by wasd
FarbigeWelt wrote:
Mon Aug 27, 2018 10:22 pm
This material behaves as one expects from a transluscent body.
You have very peculiar translucent materials around you. I can hardly find one around me that will scatter light as much as glossy translucent does.
I can only do cpu rendering, and at least in that case, glossy translucent is not translucent at all. When it says it's 97% transmissive, that means that it in fact 3% reflective and absorbs everything else. Abandon all hope, photons who enter here! Glossy translucent mercilessly devours all photons brave enough to travel beyond surface, then with all it almighty surface violently emits violent photons of evil in every direction.

Re: Test scences GPU Path vs CPU BiDir

Posted: Tue Aug 28, 2018 6:10 am
by FarbigeWelt
wasd wrote:
Mon Aug 27, 2018 11:02 pm
FarbigeWelt wrote:
Mon Aug 27, 2018 10:22 pm
This material behaves as one expects from a transluscent body.
You have very peculiar translucent materials around you. I can hardly find one around me that will scatter light as much as glossy translucent does.
I can only do cpu rendering, and at least in that case, glossy translucent is not translucent at all. When it says it's 97% transmissive, that means that it in fact 3% reflective and absorbs everything else. Abandon all hope, photons who enter here! Glossy translucent mercilessly devours all photons brave enough to travel beyond surface, then with all it almighty surface violently emits violent photons of evil in every direction.
As said, the cuboid in the picture is very thin and I miss scattering settings. You may need a model made from two 3D meshes for your purpose, one made from a very thin Glossy transluscent and the other made from Glass, both with clear volume and the same IOR. Heterogenous volume has a big influence on Glossy transluscent.
Glossy Tranluscent LuxCoreRender CPU BiDir Heterogen Front
Glossy Tranluscent LuxCoreRender CPU BiDir Heterogen Front
Glossy Tranluscent LuxCoreRender CPU BiDir Heterogen Back
Glossy Tranluscent LuxCoreRender CPU BiDir Heterogen Back
In the picture above cuboid's depth is 1.0. In contrast to the earlier post, alpha channel is used for five faces and heterogenous volume as inner volume.

Re: Test scences GPU Path vs CPU BiDir

Posted: Tue Aug 28, 2018 2:34 pm
by wasd
FarbigeWelt wrote:
Tue Aug 28, 2018 6:10 am
alpha channel is used for five faces
And that's the problem. Opacity is not physical property. And look at lasers, they are not bent. It is not glossy translucent you're testing, but opacity.

Re: Test scences GPU Path vs CPU BiDir

Posted: Tue Aug 28, 2018 3:47 pm
by FarbigeWelt
wasd wrote:
Tue Aug 28, 2018 2:34 pm
FarbigeWelt wrote:
Tue Aug 28, 2018 6:10 am
alpha channel is used for five faces
And that's the problem. Opacity is not physical property. And look at lasers, they are not bent. It is not glossy translucent you're testing, but opacity.
Good point. One cannot see the volume effect if one uses glossy translucent. Opacity 100% removes the surface completely, concluding there cannot be any refraction. Is there any refraction with glossy material expected? It is glossy not glassy.

Re: Test scences GPU Path vs CPU BiDir

Posted: Tue Aug 28, 2018 4:53 pm
by wasd
FarbigeWelt wrote:
Tue Aug 28, 2018 3:47 pm
One cannot see the volume effect if one uses glossy translucent.
You can see volume effect, but you can't see through, because of the name glossy translucent only first part is true.

Volume effect: they are now both glossy translucent. Left is green, scattering scale 0.01, absorption depth 1 cm; right is red, scattering scale 10.0, absorption depth also 1 cm. Both has diffuse color #333333
glotra-6.png
FarbigeWelt wrote:
Tue Aug 28, 2018 3:47 pm
Is there any refraction with glossy material expected? It is glossy not glassy.
Of course refraction is expected with translucent material which IOR is different from its surrounding. Btw, glass is also glossy.

Re: Test scences GPU Path vs CPU BiDir

Posted: Tue Aug 28, 2018 5:46 pm
by B.Y.O.B.
AFAIK, the glossy translucent material scatters light uniformly, no refraction happens.
wasd wrote:
Tue Aug 28, 2018 4:53 pm
Btw, glass is also glossy.
Not sure what you mean by that.
The technical term is "specular" (perfectly smooth surface leads to perfect reflection/refraction).
In LuxCore, materials are divided in three categories: diffuse, glossy and specular.
Rough glass is a "glossy" material.
Metal is also a "glossy" material.
Matte would be a "diffuse" material.

If you look at the source code of the glossy translucent material, the transmission event is "DIFFUSE | TRANSMIT", while the reflection event is "GLOSSY | REFLECT".

Re: Test scences GPU Path vs CPU BiDir

Posted: Wed Aug 29, 2018 5:39 am
by FarbigeWelt
wasd wrote:
Tue Aug 28, 2018 4:53 pm
FarbigeWelt wrote:
Tue Aug 28, 2018 3:47 pm
One cannot see the volume effect if one uses glossy translucent.
You can see volume effect,

Volume effect: they are now both glossy translucent. Left is green, scattering scale 0.01, absorption depth 1 cm; right is red, scattering scale 10.0, absorption depth also 1 cm. Both has diffuse color #333333
You are right with this one. I was not precise enough.