The oSDFT seems to be performing better than the current mSDFT:
- It's definitely faster (not much, but still)
- It's less noisy (assuming there are no bugs)
- It's easier to implement (at least in Python using Numba)
I'm considering to replace the mSDFT with oSDFT and dropping the C implementation. The current mSDFT implementation will be moved to an archive branch.