FFT performance boost - Ne10 backend

[pfeatherstone]

In my tests, the Ne10 library has much faster FFTs than the FFTW library on ARM. So it would be good to add this as a backend.

[pfeatherstone]

Though it only works for powers of 2

[davisking]

Yeah. You could have a code path for power of two sizes that switches to that 🤷

[pfeatherstone]

Though I might try adding a sprinkle of SIMD on the kissFFT backend at some point to see if I can close the gap on ARM.

[davisking]

Though I might try adding a sprinkle of SIMD on the kissFFT backend at some point to see if I can close the gap on ARM.

That would be cool. Especially since then it's not another dependency 😁