I spent today mostly cleaning up the GPU warp alignment/padding code I wrote yesterday, fixing some bugs, making it a big more effective, and writing some tests. I spent a bunch of time verifying weird edge cases to convince myself that it's doing what it's supposed to, and it seems to be. The best outcome from today is a fuzz test that generates random presets and then serializes them into GPU format and runs a block of simulations. This is really helpful to convince me that at least the new warp alignment code does not crash, even against tens of thousands of huge random presets. This is also nice because it opens the door to more tests of the simulator/GPU code, which are sorely needed.
While cleaning this up and fixing bugs, I also noted a few possibilities for even better optimization. They're all a bit trickier than what I've done, which is already substantially tricky, so I am going to backburner that work while I finish getting modulators working properly. As much as I want to dig into more fun optimizations, I really want to get modulators into a state where I can demo them in a longer-format video.
