[SciPy-user] some benchmark data for numarray, Numeric and scipy-newcore

Travis Oliphant oliphant.travis at ieee.org
Sat Dec 3 14:25:01 CST 2005


Gerard Vermeulen wrote:

>I have benchmarked some array manipulations using the default
>Numeric and numarray on my Mandrake-10.2 and a recent snapshot
>of the new scipy core.
>  
>
Thanks for the benchmarks.   There definitely may be additional 
code-optimizations we can do.  I would like to figure out why those 
specific lines you mention are taking longer --- it may be some simply 
thing.

The other thing to keep in mind is the buffer size.  SciPy core has a 
user-settable buffer size that determines when "mis-behaved" arrays (or 
type-cast-needed arrays) are copied and when they aren't.  Playing with 
that size can make a difference on a per-platform basis.   I just picked 
a default size that seemed reasonable.

There are definitely some other optimizations that could be applied.   
Benchmarks like these can help us find them.   If there is really 
something inherently slow in the ufunc algorithm then that needs to be 
fixed.

-Travis





More information about the SciPy-user mailing list