Hey, has anyone here done any experiments with volume lights? I happened upon a rather promising article proposing an alternative to traditional ray-marching. Essentially a dense modelled polygonal grid is placed in front of the light. The depthmap of the light is used to deform the vertices with vertex texture fetch or algorithmically. Seems pretty reasonable, and promises a huge speedup compared to ray marching. Now, would it be possible to implement it on default light types in Unity, are the depthmaps they use accessible? Would be great if someone from Unity would say a few words on the feasibility of the technique, I don't want to go rummaging around in undocumented features just yet. The site and paper are here: http://www.cse.chalmers.se/~billeter/pub/volumetric/ I tested the VTF with a static depthmap,works pretty well,and fast.
the depth buffer is per camera, not per light and only present if you tell the cam to render it at all. But yes its accessable, see the shader lab documentation on depth rendering etc in the reference area of the documentation to find the variable you can find it in.
Depth buffers are also used for shadows, and there is at least one per shadow-casting light. Unfortunately, I don't think it's accessible from arbitrary shaders. It obviously needs to be accessed in fragment shaders for lighting, but you might find it difficult to bend that to your needs.