New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
FFT - DSP(Double)SplitComplex and conjugate symmetric #96
Comments
Not entirely sure how it came to be. Would you be willing to create a PR with your proposed changes? |
We're currently validating an implementation using the packed FFT algorithm with several test-vectors against Matlab's FFT implementation. If this works out, I could certainly create a PR for this. I'll keep you updated. Still, it would be worth knowing what you design decision was. I'm not a DSP expert and in fact only collected some thoughts from stackoverflow to get some kind of understanding of FFTs in general. |
Maybe @mattt can shed some light. |
I'm currently evaluating several FFT implementations and I tried out your implementation (of the vDSP algorithm collection). After some reading there're two things I don't fully understand and I hope you might help us out here:
vDSP_fft_zrip
/vDSP_fft_zripD
? I understand that your implementation works with real input vectors. So what's the benefit of not (!) using the "packed" implementation with the 'r' in the middle of the function name. The 'packed' implementation needs some additional packing / unpacking - but that's it.I'm specifically referring to the statement in https://github.com/jseales/numpy-style-fft-in-obj-c
Certainly your implementation works, but it seems using the different method it should run faster or at least more efficient? On that note it's worth exploring the "in-place" variants of the methods as well.
conjugate symmetric fft
"half of the result" could be discarded anyhow (for real input values).The text was updated successfully, but these errors were encountered: