We are running into an issue with RuntimePartitioned PCG actors and GPU-based Static mesh spawner. Sometimes, a portion or even every single mesh stops rendering on the actor.
The only way to fix this issue when it appears is to manually refresh the graph or move away from the camera far enough so that the PCGRuntimePartitionActor gets destroyed, and then move back closer so it gets initialized again. On its own, the actor will never recover without manual refresh/reinitialization.
We are using this setting for the PCGWorldActor:
[Image Removed]
Unfortunately, I don’t have a concrete repro for this, but I can mention a couple of scenarios when this usually occurs:
After loading into a world-partitioned level
Changing a material property on a mesh that is rendered by the RuntimePCG actor - everything disappears, not only the meshes with the material that was changed
It’s possible these are all caused by the same root cause, we have a known issue where flushing of the render state was not properly detected and that can cause instances to go missing until regenerated. Another way this could be triggered is changing the quality settings at runtime.
Both loading into a world and changing/adding materials can sometimes cause a flush, too.
We’ve recently submitted a changelist that addresses some of these issues.
Could you try out CL 50449310 in Main and see if that improves the situation for you?
I spoke to my colleagues who handle merging Unreal Main into our code base, and they were concerned about cherry picking this change since there are probably other dependencies that we would have to merge. So I’ll probably have to wait till this change gets propagated into our codebase.
Once I get the chance to test it out, I’ll let you know if the issue still prevails.