Hybrid Back/Forward path tracing (aka BiDir without MIS)

[PetrT]

Thanks all works

[bnzs1]

B.Y.O.B.
i downloaded those branch and it works - i ask scene - to see settings and check boxes...

[acasta69]

As much I see you got my point. I don’t mean to simulate real optics system. In a math sim this is not required because you do not require to correct chromatic aberrations, because dispersion can be avoided per definition. In math sim there is no limitation but fp precision for the shape of the lens and the sensor, i.e. the benefits exceed in my opinion the calculation costs. And I guess the implementation is just a small extension of the current algorithms, exact the same for all bounces between camera and light. Extra computation should be less than increasing depth by 2, including ‚perfect‘ depth of field.

Actually, a benefit of simulating real lens systems could be a realistic lens flare effect
However, for this purpose, it would probably be much more computationally efficient to implement the camera by sequential ray tracing, i.e. geometric raytracing using matrix calculations.

I admit that I have no exact idea how much this kind of "camera realism" can affect performance, I imagine that different kind of additions may have different cost...
We are often discussing about having max possible speed and also about our resources being so limited... and I don't see a realistic camera in the feature poll done some time ago, so I guess there is not so much interest around this.
(In the old LuxRender forum I once proposed the introduction of tilted focus plane... a niche feature for sure, ok, but interest was close to nil.)

Personally I'm a fan of having realistic physical simulation in rendering systems, and especially in the camera: the rare times I render an image it is usually to obtain some kind of photographic-like result.
I had a look at this, just as an example:
https://rmanwiki.pixar.com/display/REN22/PxrCamera
A nice set of features, if you ask me, but again I have no idea about computation cost or implementation complexity, let alone how many users could be interested.

[CodeHD]

A nice set of features, if you ask me, but again I have no idea about computation cost or implementation complexity, let alone how many users could be interested.

Don't get me wrong, I don't believe either that it will be worth the effort to implement all this.
I've been trying to get lens flares with a simple elns system today and even failed at that wth the current engines, image formation seems difficult enough for Metropolis, the weaker interreflections seem to be just lost.
I just always also like to think about what is missing, or where the limitations are

[epilectrolytics]

Hybrid does not work well with rough glass.
Apparently the path tracer renders caustics too and messes up the result with fireflies:

testhybr_LuxCore.zip

(477.84 KiB) Downloaded 176 times

.
While I tested this Blender crashed two times during rendering.
And when trying to render the light pass I got very inconsistent behavior when changing the sampler:
Sometimes a black screen, sometimes a normal result, sometimes way too bright caustics.

So I exported to LuxCoreUI, there I got no crashes but (when changing between Sobol and Random) some times black or too bright caustics:

Win10Pro 1903, LuxCore not latest build but from July 7. or 8.

[Dade]

Hybrid does not work well with rough glass.
Apparently the path tracer renders caustics too and messes up the result with fireflies:

I should have fixed this problem.

So I exported to LuxCoreUI, there I got no crashes but (when changing between Sobol and Random) some times black or too bright caustics:

I should have fixed this problem too.

[epilectrolytics]

I should have fixed this problem.

Thanks for fixing this!

But just having downloaded the latest build for Windows once it was finished (20190723.3), the "path traced caustics mixed in" issue still persists when rendering from Blender, my results look like in the first image two posts above.

Is it possible that Azure messed up latest builds again?
I see that there was a second manual build run and waited for it to complete.
But I don't know how to identify the actual build version from a zip file with windows binaries.

[Dade]

Yes, the deployment has already failed 2 times (https://dev.azure.com/LuxCoreRender/Lux ... e-progress), I scheduled another try but it is pending there without any additional information. Azure isn't exactly very reliable.

[Dade]

The last deployment seems to have worked.

[epilectrolytics]

The last deployment seems to have worked.

With the latest build for Windows downloaded in the morning of July 24. indeed the caustic mix issue is solved and hybrid render works as expected with rough glass shader.
But in LuxcoreUI I still get the "too bright caustics" problem:

No idea if the bug is still there or if Azure does not provide the really latest build.

Interesting:
With rough glass LightCPU engine renders a complete solution including SDS-paths, albeit slowly.
I wanted to provide a screenshot but LuxCoreUI crashed when I tried to open the render engine editor window.

Because of the current heatwave in Europe this week I will not do extensive texting on my PC which is getting too hot but once a day try latest builds and report if they work.