How many tiles had the wall. I have a level here, consisting of 1800 segments for walls, floors, ceilings, etc.
Sometimes I get a dozen in a row lit correctly, but then huge seams at another place. Maybe you were just “lucky” with your wall.
I was also able to attenuate the effect wen directly lit. But in the indirectly lit areas, I never could get rid of the issue.
If one thinks about the whole thing it becomes a bit “funny”:
The lightmap baking is distributed between multiple threads to improve baking performance.
Due to the thread separation we get inconsistent lighting. So this is not systemic to the way the lighting is calculated, but more like an optimization artefact.
If no interpolations were made and the lighting were explicitly calculated for each shading point, the lighting would be a-ok.
Then we need to crank lightmap resolutions up to mask this effect, which in turn, increases build times, thus largely defeating the intended goal of optimization.
Dont get me wrong. Optimizations are good and neccessary. And for most optimizations a compromise must be made.
However it should also be taken into account what “stage” is optimized: Designtime, Build/Compile time or Runtime.
Optimizations that help runtime performance are almost a must and if there are compromises to make, they may be unavoidable.
But that is something where the player (paying customer) can say: “Hey, I benefit from that”.
On the other hand, optimizations that speed up things during Designtime (enabling/disabling geometry rebuild), or lighting build time (multi thread baking) should be optional.
When I am able to build the lighting in 10 minutes for a level but have these artefacts, the paying customer will never know that I only spend 10 minutes on the baking job and he will probaply not even care. He will only judge the result. Whether I enjoy a book depends on the content. I would never say, hmm “crappy book, but wow, it was printed in 35 seconds…”
As it had been suggested already, a single thread baking would be highly appreciated.
Maybe call it another lighting quality stage… Preview - Production - Unreal
It would be intended to the final lighting build. Really the last build before the thing goes into production at the DVD plant.
Even if this would result in a 50 hours lighting job, for the final build it would be worth it.