LuxCoreRender Forums

B.Y.O.B. wrote: ↑Sun Jun 23, 2019 6:35 pm I rendered a quick animation with it: 90 frames with 40 sec rendertime per frame.
https://youtu.be/MXKYdekMgSE

Cool!

Dade wrote: ↑Sun Jun 23, 2019 5:24 pm The solution is a fully functional BiDirVM, something that can now happen but not in v2.2, we need to stop and release v2.2 or is never going to be finished if we continue to add features.

Great news and I agree to stop adding more features.

Let's keep caustic cache as it is, no more wasting time on it and maybe hide it behind an "experimental feature" tab in LuxBlend.

Now for the hybrid backforwardtracer:
While it works already great out of the box with the Wallpaper scene, in my pooltest scene everything is black apart from the tiles.
Looks like infinite lights are not supported yet and unlike BiDir no caustics are rendered in the water.

epilectrolytics wrote: ↑Sun Jun 23, 2019 7:03 pm Looks like infinite lights are not supported yet and unlike BiDir no caustics are rendered in the water.

I have probably forgotten something, I will check.

B.Y.O.B. wrote: ↑Sun Jun 23, 2019 5:57 pm Dade, did you see my question on the last page?

B.Y.O.B. wrote: ↑Sun Jun 23, 2019 4:21 pm One question: does it make sense to use this for CPU-only rendering, or will Bidir still win there?

Hybrid is a LOT faster than BiDir in term of samples/sec (it traces a lot less rays). BiDir samples may have less noise. The answer is in the "may have": if "less noise" wins over "more samples/sec", BiDir is better otherwise Hybrid is better.

Short version: yes, it make sense to use this for CPU-only rendering.

So, environment cache won't make it into 2.2?

lacilaci wrote: ↑Sun Jun 23, 2019 7:31 pm So, environment cache won't make it into 2.2?

It is something i definetely want to see in 2.2. i'm rendering a lot of interior now without GPUs envcache + PGI do wonder on CPU. It is good to render so quickly on weak hardware.

I think people with laptop will appreciate that one. So having Env Cache on GPU will be crazy.

lacilaci wrote: ↑Sun Jun 23, 2019 7:31 pm So, environment cache won't make it into 2.2?

It will (if it works well).

Hi all,
Thanks for enabling this options.
It sounds totally reasonable to combine this two algorithms to speed up rendering.

Does the hybrid rendering concept keep the physics base?
Is energy conservation given during the rendering process?
Does the hybrid option allow light paths meeting somewhere on track between light source and camera (or vice versa)?
What is the multi importance sampling made for, if it works properly without it?
How Is the allowed number of light path and eye path defined?

BR

lighting_freak wrote: ↑Mon Jun 24, 2019 5:19 am Does the hybrid rendering concept keep the physics base?

Yes, it is unbiased and consistent. It is just a different type of Bidirectional Path tracing.

lighting_freak wrote: ↑Mon Jun 24, 2019 5:19 am Is energy conservation given during the rendering process?

Yes.

lighting_freak wrote: ↑Mon Jun 24, 2019 5:19 am Does the hybrid option allow light paths meeting somewhere on track between light source and camera (or vice versa)?

It connects the vertices of pure specular light paths to the camera (i.e. it has the only task to render caustics, everything else is done trough eye paths and classic path tracing).

lighting_freak wrote: ↑Mon Jun 24, 2019 5:19 am What is the multi importance sampling made for, if it works properly without it?

MIS is not used: the light paths compute samples that eye paths don't, so there is no overlap and no need to mix the results, they can be just added.

lighting_freak wrote: ↑Mon Jun 24, 2019 5:19 am How Is the allowed number of light path and eye path defined?

Yes, you can define the ratio between light paths and eye paths, to decide where to dedicated more computing resources.

Which depth (amount of bounces) is used for light paths?

B.Y.O.B. wrote: ↑Mon Jun 24, 2019 8:15 am Which depth (amount of bounces) is used for light paths?

The same of eye paths (so the max. "Eye <=> Light" path length is the same).