Not really an issue, but I was not sure where to put this.. so feel free to close/move this if you want.

As I was struggling to get the official CUDA-based viewer running on my windows machine, I have reimplemented the renderer inside of unity based on compute shaders.
The ply-loader implementation is not very robust, but should work for radfoam models which match the format of the released models in this repo.

I have also experimented with reordering the voronoi-cell-points in morton order, and did see a small, but noticeable, improvement in performance. Maybe reordering the points each x-iterations could help improve training times?

The source code is available here: https://github.com/ChristianSchott/radfoam-unity

Note that something still seems a bit off with my color calculations, in case someone wants to take a look at that. Also, I would be interested in how performance compares to the CUDA implementation.

The repo also contains a proof-of-concept implementation for a WebGL-compatible renderer, but it is not really practical..