Henrik Enquist

I played a little with iai just now, looks very promising! https://github.com/HEnquist/RustFFT/blob/iai/benches/bench_iai.rs The only problem is that it can't exclude the setup. So each test must run twice, once with...

Those 32 tests (or really 16 since I need to run each one twice) ran in 7 seconds on my ryzen laptop. Leaving it overnight should produce a nice big...

I made some simple macros to generate the tests, and ran a series for radix 4. Looks pretty nice!  Length on x, estimated cycles on y. Next I'll look...

Actually, that seems to have changed after this: https://github.com/ejmahler/RustFFT/commit/6ab56f96fa0c0779a8886d2755c979bd836e8f41#diff-07bb71e908a04b41bccc8c3665eae0e78f9d562af2e4425df3c442c2337f5bdc That "slightly faster" seems to grow with length, so I don't really see mixed radix winning any more (at length 4194304...

I had forgotten about this and just rediscovered it. I can't help wondering what this tests that isn't already covered by the existing tests. All algorithms already have tests that...

How about just adding a simple example? ``` Output of a 6-point FFT: FFT(x) = [(X0r, X0i), (X1r, X1i), (X2r, X2i), (X3r, X3i), (X4r, X4i), (X5r, X5i)] with frequencies: [0,...

It also needs a short text to go with it. `fs` is the sampling frequency. I think it makes sense to keep an example like this very simple and just...

Something like this? # The FFT output format. In this example `x` is a vector of 6 complex values. The values are samples of a complex signal that changes as...

Related: https://github.com/ejmahler/RustFFT/issues/73

Your solution is not far off, you "just" need to transpose the data between the FFT steps so that each FFT runs along the correct axis. But have you looked...