another attempt to speed up with cuda

[radiotommy]

I didn't know biergaizi has done some great job already.

I tried accelerating it myself using CUDA. My implementation is still in the proof-of-concept (PoC) stage, but I’ve successfully gotten the Engine, excitation, and UPML components working—enough to run basic simulations.

I haven’t had the chance to do detailed profiling yet, but on my RTX 3050, I’m seeing around 350 Mcells/s. That’s roughly 2.5x faster than the multithreaded SSE engine, which is promising! You can check out my work here:

https://github.com/radiotommy/openEMS-Project
Ideally, I’d love to load everything onto the GPU and run the entire simulation there to minimize the overhead of data transfers between GPU and CPU. However, I’ve found this to be quite challenging without significant refactoring of the codebase. Still, it’s a work in progress—any thoughts or suggestions are welcome!