Skip to content

Conversation

@mildsunrise
Copy link

  • Bunch of obvious fixes so that things work again
  • Add more sizes
  • Make benchmarks a bit more exact by having them only calculate FFTs, nothing else
  • Add an FFT I wrote which is about ~50 lines of code and performs more or less like PulseFFT
  • Some other small improvements

- don't compute total as part of the benchmarks pls
- add 8192 size
- report FFT points, not total
- pre-generate a few buffers
- use random data
- remove dead code
- extract two things to variables
- use f32 (for fairness with pulsefft which also uses f32)
implement full Cooley-Tuckey (not possible with WebGL,
which is what I wrote this code for)
@mildsunrise
Copy link
Author

mildsunrise commented Aug 28, 2025

in case someone is interested, I made a more proper benchmark that adds more implementations (some are ~twice as fast as PulseFFT), performs the execution in a dedicated worker, and plots the results in a nice graph. However it only compares complex FFTs.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant