Getting ready for Cuda
- explicit random state instead of local-thread storage
- multiple fixes to the renderer to get it closer to Mitsuba
In addition I made a lot of modifications to simplify the code and make it easier for me to use Cuda:
- I removed Unity, C#, Mac, and multi-threading related code
- I simplified the Windows main to be a simple console application that runs a renders a predefined number of frames into a .png image (thanks to stb_image_write).
On my laptop I got 8M rays/s using the simplified single-threaded implementation.
Here is a link to the full source code