Open iesten opened 8 years ago
Out-of-place transforms are generally faster than in-place transforms, and FFTS is all about speed. Speed comes from runtime code generation, precalculated twiddle factors etc.. So you should be looking implementation using CORDIC based FFT for low memory devices.
But there are other ways to reduce memory usage within FFTS
Indeed they are. I am suggesting that it would be interesting to explore an in place version of the FFTS algorithm - perhaps it could achieve best in class speed for in place processing as well as out of place processing.
For low memory devices, support for in place processing would help reduce memory usage.