Closed brahma-dev closed 2 years ago
Nevermind. It works.
fft_analysis->input = fft_input0
@brahma-dev Yes, it should be possible! You can allocate your own memory for the input and output buffer. Then, you can just update the content of the array and re-run the FFT.
fft_init
requires some CPU cycles to pre-compute the "twiddle factors" of the FFT, but if you only deal with fixed size, you could pre-compute them at compile time, or create the look-up table in advance.
Then you could get rid of fft_init
altogether.
I am trying to run fft on two input arrays back and forth.
Both arrays are of same size. Is it possible to use it such that I don't have to run fft_init every time and instead just change the input and run fft_execute ? I want to do so because fft_init takes more time than fft_execute. And I am short on CPU time.