[SciPy-dev] [SciPy-user] Benchmark data

Travis Oliphant oliphant.travis at ieee.org
Fri Dec 9 13:31:34 CST 2005


Gerard Vermeulen wrote:

>On Fri, 09 Dec 2005 03:14:49 -0700
>Travis Oliphant <oliphant.travis at ieee.org> wrote:
>
>  
>
>>I'd like people to try out scipy core in SVN.  I made improvements to the
>>buffered ufunc section of code that I think will make a big difference
>>in the recently published benchmarks. 
>>
>>    
>>
>
>Hi Travis,
>
>indeed, it made a big difference (for big arrays scipy is now fastest on some
>statements).
>
>Below are my benchmark results on my DIY python, see
>http://www.scipy.org/mailinglists/mailman?fn=scipy-user/2005-December/006057.html
>
>  
>
Thanks for these tests.  I'd like to gather more data, but it seems that 
the real difference here is the optimizations that are set for your 
tests.  On my tests no optimizations were done by the compiler (I'm 
using a debug build and scipy is clearly doing better).  On your system 
with optimizations set numarray starts to edge out for very large 
arrays.    Apparently, the compiler is able to optimize the inner loops 
for numarray better (at least on your machine).

If anybody has opinions on that, it would be great.   One idea is that 
the inner loops for scipy and Numeric use strided memory access and 
pointer dereferencing on the inner loop.   Perhaps the optimizer could 
do a better job if the dereferencing were done outside of the inner loop?  

Thanks,

-Travis







More information about the Scipy-dev mailing list