here the scene
https://we.tl/t-0gsJXscmDi
DLSC test
Forum rules
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
Re: DLCS test
Don't use the synthetic Sky but HDRI/IBL.
Check more about it here: viewtopic.php?f=9&t=472&start=10#p4808
Re: DLCS test
Well your own test there actually shows that it is not sky or hdri but a combination of sun and sky or hdri or flat color.kintuX wrote: ↑Sun Sep 30, 2018 9:00 pmDon't use the synthetic Sky but HDRI/IBL.
Check more about it here: viewtopic.php?f=9&t=472&start=10#p4808
For example I rarely use HDRI cause they only make sense in some cases. For example in most archviz I do use sun and sky system, so what do I do if I want to use sun&sky and many interior lights that could benefit from DLCS?
Re: DLCS test
Well I did further tests after, with alpha3 on interior, and it appeared that, when using HDRI with Sun, those splotches were gone. True, I didn't post any results yet (because i didn't thoroughly test it further), but here's a quickie:lacilaci wrote: ↑Mon Oct 01, 2018 6:15 am Well your own test there actually shows that it is not sky or hdri but a combination of sun and sky or hdri or flat color.
For example I rarely use HDRI cause they only make sense in some cases. For example in most archviz I do use sun and sky system, so what do I do if I want to use sun&sky and many interior lights that could benefit from DLCS?
Had simply changed World's Sky for HDRI (HDRI 360° Saalfelden, Austria) --- Gain: 3000
Personally, i prefer HDRI/IBL simply because of artistic freedom.
Re: DLSC test
DLCS is not design for this kind of scene or i'm completly wrong
Re: DLSC test
It should still not cause artifacts. Seems like a bug to me.
Re: DLSC test
maybe it's a problem of mostly uniform lighing that sky creates or completely uniform from flat color? Confusing lightcache?
Re: DLSC test
Yes of courseIt should still not cause artifacts. Seems like a bug to me.
I mean people should use correct tool for better perf ! if it is also cool if such test show buggs
Re: DLSC test
I think that is a sampling problem: DLSC trys to establish if the light is visible with very few rays (in order to speed up the pre-processing phase). But trying to establish if the Sky light source is visible with only ~16 rays can lead to false positive/negative in indoor scenes. I will check but I'm pretty sure this is the explanation.
The solution shouldn't be too complex: for instance I could handle Sky light as special case requiring more rays or always include Sky in the cache entry, etc.
Re: DLSC test
Maybe the DLSC building phase could also use the visibility map, if it was computed?