3. Why a new FFT?
The timings I just posted probably justify writing a new FFT. However
that is not what concretely motivated me to write one.
I began thinking about writing a new FFT about the end of September
this year. At the time I skimmed through a huge number of papers on
the FFT for someth
I've done some more timings. This time I turned off MPIR's sqrt2 trick
which I don't have yet.
For some multiplications this causes MPIR to take way too long due to
tuning issues. For example, there are points where larger numbers take
less time to multiply.
I made a comparison showing how much f