Technically there's only a limit if you use an OpenCL engine, the CPU engines have no limit.
But in the Blender addon I'm always using the limit of 8 to make my life easier.
We could use a higher number if requested.
@Dade, I guess this would not raise memory consumption when the max. number of light groups is not actually used, right?
B.Y.O.B. wrote: Sat May 09, 2020 10:54 am
@Dade, I guess this would not raise memory consumption when the max. number of light groups is not actually used, right?
Yup, it would increase the size of the SampleResult structure, there is a SampleResult structure for each thread and there are hundred of thousands of GPU threads.
Support LuxCoreRender project with salts and bounties
B.Y.O.B. wrote: Sat May 09, 2020 11:34 am
Oh, so my statement was wrong, it would have an impact even if no lightgroups are used at all.
Yes, GPU structures must have a fixed size or it would require conditional compilation (i.e. kernel re-compilation). There are other options (i.e. store the values in a dedicated buffer, not inside the SampleResult structure) but are complex.
Support LuxCoreRender project with salts and bounties