At the moment we are wasting a large number of computational resources for minimal benefits and have an easy-to-break behaviour regarding shadows and spotlights. We can look at UsdMaya's calculation of shadow matrices (which directly grabs shadow projection matrices from the vp2 context). It's not as trivial to implement as there since we both need to support setting up simple lights and other hydra light sprims.
Added a quick take at the problem in feature/shadow-matrices. It's still buggy, but I think it's a small difference between the shadow matrix Maya provides and the one HdSt expects.
At the moment we are wasting a large number of computational resources for minimal benefits and have an easy-to-break behaviour regarding shadows and spotlights. We can look at UsdMaya's calculation of shadow matrices (which directly grabs shadow projection matrices from the vp2 context). It's not as trivial to implement as there since we both need to support setting up simple lights and other hydra light sprims.