[SciPy-user] NumPy vs. SciPy and other speed comparisons

David Cournapeau david@ar.media.kyoto-u.ac...
Wed Jun 11 06:45:09 CDT 2008


Ivo Maljevic wrote:
> 2008/6/11 Robert Kern <robert.kern@gmail.com 
> <mailto:robert.kern@gmail.com>>:
>
> Specifically, it is "from scipy import *" that imports the
> subpackages. "import scipy" does not.
>
> Sorry, but I am not sure I understand everything here.
>
> 1. "import scipy as sp" vs "import numpy as np", and then I added 
> prefixes in front of all the functions. The times are the same, 44 and 
> 23 seconds, respectively. There goes that theory.
>
> 2. I tried explicitly importing only the required functions, e.g,  
> from numpy import sqrt, arange, ones, zeros, random, where, ceil, sign
> Agian, the same results.
>
> Would the loading time of the two packages account for over 20 seconds 
> difference in execution time? 

Ok, the bad guy is..... sqrt. scipy.sqrt is much slower than numpy.sqrt 
(note that in your script, you could avoid computing the scale of the 
normal in the loop).

cheers,

David


More information about the SciPy-user mailing list