[SciPy-user] FFT speed ?

Andrew Jaffe a.h.jaffe@gmail....
Wed Feb 7 14:48:06 CST 2007


Scott Ransom wrote:
> On Wed, Feb 07, 2007 at 12:13:12PM -0500, Anne Archibald wrote:
>> On 07/02/07, Samuel GARCIA <sgarcia@olfac.univ-lyon1.fr> wrote:
>>>  Yes I know this.
>>>  But 186888 186889 and 186890 are not power of 2 and the computation time is
>>> very very different just for a difference of size of only one point.
>>>  What is the reason ?
>>>  And how to deal with that ? (I realy need to compute fft with a random
>>> nomber of point)


It's probably worth pointing out that 186889 is, in fact, prime, which 
is certainly the worst case for any algorithm.

Andrew



More information about the SciPy-user mailing list