If you use MATLAB you don't need to worry as it does it for you. If you want to do it 'by hand' so to speak, then you need to study FFT algorithms. I like the book by O. Brigham.

If you are using MATLAB, it does it for you, and quite efficiently.

If you are doing it 'by hand' then you need to study how FFTs work. Get a good book and study.

FFTs come in 2 flavours: Decimation in time (DIT) and decimation in frequency (DIF). Both result in the same speed gains and the choice is personal. Then there is the issue of the radix. In radix 2 one gains in processing time by knowing that the twiddle factors that are pi (half a turn) apart differ in sign only so the multiplications by 1 and -1 are not needed and are simply incorporated into the code and for radix 4 the twiddle factors differ by j, so, again the multiplications by 1, j, -1 and -j are not needed and are incorporated into the code.

Before you use mixed radices FFTs you need to understand well how the twiddle factors of single a radix FFT behave.

I like the book "The Fast Fourier Transform" by E. Oran Brigham and the chapters on implementation of FFTs on the Texas Instruments DSPs by Panos Papamichalis in "Digital Signal Processing Applications with the TMS320 Family". Yes, they are both oldies, but they are excellent!

I include 2 publications of mine where I explain an implementation of a mixed radix FFT (radix 4 and radix 2) to obtain the FFT of 256 real values treated as 128 complex values:

. “Real-time spectral analysis of Doppler signals using a digital signal processor and a microcomputer”. Schlindwein, F.S. and Evans, D.H, In: Physics in Medical Ultrasound II, chapter 15, pp.112-118. The Institute of Physical Sciences in Medicine, report N.57, 1988.

. “Spectral analysis of Doppler signals and computation of the normalized first moment in real time using a digital signal processor”, Schlindwein, F.S.; Smith, M.J. and Evans, D.H. - Medical & Biological Engineering & Computing, vol.26, pp. 228-232, 1988

Best of luck.