over the last 6 weeks i've been developing a scalar FFT implementation in rust, and today it finally exceeded the performance of rustfft!
i wrote an article going step-by-step from DFT to a highly optimized FFT: yummers.dev/a-fast-cpu-f...
caveats: AVX/SSE off, input sizes 256, 512, 1024, 4096
yummers.dev
a fast CPU FFT
Deriving a fast scalar FFT implementation.