I have been using the daily build (2015-09-24), and keep getting the message light sampler can only handle 255 lights efficiently.
for this particular project there is no way around it i need a lot more lights, is there a way to get rid of the message, it pop ups every time i render, clicking ignore or dismiss all, makes no difference.
by the way is it planed to support more then 255 lights? it seems like a limitation to me.
enabling develop mode shows options for solver and sampling mode, can we get some information about the options there? there are no tooltips for them.
That is why we are working on scalable solution. We have a reasonably working prototype already done, and it will be probably release in 1.4 stable release (together with adaptivity)
in this project i have thousands of light . but i made an script to grouping lights together by distance , and then my render is faster and also cleaner : https://vimeo.com/141227043
with this technique you will have less light count and calculation will be faster , see the attachment to get the idea ,
This is interesting, vrMan. So are they “grouped” as in instances?
Ondra - very good to hear there’s efforts being made in this area. We frequently work on scenes with many hundreds of lights (even if there’s perhaps only 30 or so unique ones and the rest are instances), and it’s currently manageable but speed could be better.
I think there is some miscommunication going on. There’s only improvements in error / warning system in recent daily builds. Improvements on many light performance is planned in V1.4. Am i right?
I have a project with thousands of lights and got some help from Ludvik, here’s my layman’s conclusion, devs please feel free to correct me:
the important thing here is ‘light groups’ displayed in brackets next to the lights count in the stats
for light objects and geometry with a LightMtl assigned, a single face counts as a ‘light’, whereas a mesh consisting of - let’s say - 32 faces, counts as 32 ‘lights’ but one ‘light group’
the limit of 255 lights applies to ‘light groups’
example: if you have 1000 spheres with a LightMtl assigned, they’ll count as 1000 ‘light groups’ and Corona will use a slower algorithm. Now if you collapse these spheres and you end up with 10 objects, each consisting of 100 spheres, they’ll count as 10 ‘light groups’ and will use the faster algorithm.
the rule is to limit the number ‘light groups’ to 255
Instances are still separate geometries so they’ll still count as separate ‘light groups’, these would still need to be collapsed to larger groups of objects.
This does not apply to CoronaMtls with self-illumination, these are not handled as lights and are sampled separately.
There may be side effects to this but I’ll let this one to the devs to explain in detail. I wish I’d known this earlier and there would be a sticky thread explaining this somewhere (or maybe there is…?)
this is correct with one important note, collapsing light objects may trigger using the better algorithm, but if the collapsed objects are too big or are created from light objects too far away, the results will be even worse than before. So there it is not “collapse all lights into single object and everything will be perfect”
Any sign of progress on this issue? We are working on a project where this would be a serious problem (5000 lights might be enough). Really want to use corona for it!!
