[NumPy-Tickets] [NumPy] #1260: Use FFT in np.correlate/convolve?

NumPy Trac numpy-tickets@scipy....
Fri Feb 18 10:48:56 CST 2011


#1260: Use FFT in np.correlate/convolve?
-------------------------------------+--------------------------------------
 Reporter:  roger                    |       Owner:  somebody      
     Type:  enhancement              |      Status:  needs_decision
 Priority:  normal                   |   Milestone:  Unscheduled   
Component:  numpy.core               |     Version:  1.3.0         
 Keywords:  convolve correlate slow  |  
-------------------------------------+--------------------------------------

Comment(by endolith):

 Replying to [comment:3 pv]:
 > One thing is that `fftconvolve` uses 2*M-1 sized FFT. It's never a power
 of two, which seems to make it slow. I fixed that in
 http://projects.scipy.org/scipy/changeset/5968

 Shouldn't that be optimized in the fft function itself, instead of
 modifying all the functions that use it?  It could be optimized to use the
 real FFT when appropriate, too.

 Replying to [comment:8 cdavid]:
 > FFT has several disadvantages: it does not work for object arrays

 > I think we should keep the current implementation based on naive
 implementation, and add a FFT-based one. The current function could take
 an argument to force naive, fft or heuristic (to try to pick the fastest
 one).

 This is what I was asking for.  Can object arrays be automatically handled
 by the naive implementation?

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1260#comment:10>
NumPy <http://projects.scipy.org/numpy>
My example project


More information about the NumPy-Tickets mailing list