LuxCoreRender Forums

New LuxCoreRender v2.0alpha4 release is available at: https://luxcorerender.org/download

v2.0 release notes are available here: https://wiki.luxcorerender.org/LuxCoreR ... Notes_v2.0

The complete list of changes between alpha3 and alpha4 in Luxcore API is available here (19 commits): https://github.com/LuxCoreRender/LuxCor ... v2.0alpha4

The complete list of changes between alpha3 and alpha4 in BlendLuxCore is available here (16 commits): https://github.com/LuxCoreRender/BlendL ... v2.0alpha4

This version includes the fix of random crashes and the new visibility driven sampling of environmental lights: viewtopic.php?f=5&t=93

Thanks already on the track !

Hi,

just had a bit of time to test the new version. Some feedback:

1) The new sobol sampler rocks

2) Clamping evaluation is still much to high in my case, it suggest 134 000 000 for a scene with water made of architectural glass and some sand.

3) It would also be awesome for early testers, if the geometry caching could be implemented, even in a simple maner. It would help a lot to test on real scenes. At the moment, I have 2min40 of export time on every render for about 30sec of rendering. With even hacky caching, it would make adjusting settings much easier.

4) How can I test the random sampler?

Somehow, the fireflies are visible in F12 render but not in viewport... but I couldn't find why, it doesn't seem to be clamping. Would be great to have the same clean image on final render.

Edit: nvm, it's just due to some parameter changes not updating the viewport.

blibli wrote: ↑Sat Feb 10, 2018 8:11 am 3) It would also be awesome for early testers, if the geometry caching could be implemented, even in a simple maner. It would help a lot to test on real scenes. At the moment, I have 2min40 of export time on every render for about 30sec of rendering. With even hacky caching, it would make adjusting settings much easier.

Can you send me the scene?

blibli wrote: ↑Sat Feb 10, 2018 8:16 am Edit: nvm, it's just due to some parameter changes not updating the viewport.

Which parameters exactly don't update the viewport?

at 4): the random sampler is not available in BlendLuxCore because Sobol seems better in every situation. Is my assumption false?

the default cube with 2 subdiv modifiers at 6 will show the export times at each render. http://pasteall.org/blend/index.php?id=48896
Edit: Actually, on Linux this scene even crashes because it's way to big, sorry. With one subdiv at 5 and the other at 6, it renders and show the long export times on every render.

Regarding the random sampler, it had many commits in alpha3, so I thought it may have some advantages Dade is working on it?

blibli wrote: ↑Sat Feb 10, 2018 12:55 pm Regarding the random sampler, it had many commits in alpha3, so I thought it may have some advantages Dade is working on it?

Now it has adaptive sampling like Sobol but it should be inferior in any case. It is more there as test/reference to check if something is going wrong with Sobol or Metropolis.

ok, thanks for the clarification. Did you measure how much faster it renders with adaptive sampling? At least in my test using CPU, render time and result where the same using adaptive at 0.0 and 0.7.

@B.Y.O.B., the long export time seem to only happen with OpenCL. So it could be due to kernel recompile, but it also happen when just changing the clamping value...
Regarding viewport, changing the adaptive sampling value has no effect.

If you apply the modifiers it should export much quicker.
But yes, I should look into geometry caching again sometime.

Adaptive sampling is not used in viewport, that's why changing the parameter does not trigger an update.
Viewport uses RT Path CPU which requires a special RT Path sampler.

The viewport render is intended for checking shadows, lighting, materials, reflections etc. but not for render settings finetuning.

blibli wrote: ↑Sat Feb 10, 2018 1:16 pm ok, thanks for the clarification. Did you measure how much faster it renders with adaptive sampling? At least in my test using CPU, render time and result where the same using adaptive at 0.0 and 0.7.

What do you mean with "where the same" ? Or better, how did you measured they were the same ? This viewtopic.php?f=5&t=83 includes an example of results ("The race" section).