Just a learning question if you wouldn't mind: Why do directional light shadow views need to cull based on the camera's world position rather than the light's position?

Just a learning question if you wouldn't mind: Why do directional light shadow views need to cull based on the camera's world position rather than the light's position?

I don’t mind at all!

As far as I understand, visibility ranges are only meant to define distances at which meshes start to fade out from view / fade into view when viewed from a user defined camera, as opposed to a shadow camera. This is specifically about visibility range culling as opposed to the more general frustum culling, even though the logic is in the same block in the wgsl file.

In the specific example for visibility_rangeon main, you can observe the results of the existing behavior (where it doesn’t use the user defined camera). When the user moves their camera at certain angles at distances, the shadow starts to cull even though the object hasn’t. I don’t know much about shadow cameras (I’ve really only dove into this part of the code within the past few weeks!), but I assume it’s because the distance from the shadow camera is not within the visibility range of the shadow, which shows up as premature fading on the user camera. The mesh and its shadow also end up using two different cameras to do visibility range culling instead of the same user camera, which means the mesh and shadow could end up disappearing/reappearing at different times.

@pcwalton just review requesting you since i think you might be most familiar with the things this code is touching / have strong opinions. please look whenever most convenient for you!