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

Arnd Baecker arnd.baecker at web.de
Mon Dec 5 00:55:12 CST 2005


On Sun, 4 Dec 2005, Travis Oliphant wrote:

[...]

> The other issue of vector-vector and vector-scalar operations, I'm less
> convinced about.   Do we really need a whole other class of functions in
> the ufunc machinery.   If so, I'm inclined to included them in the math
> operations for array-scalars, rather than the ufunc machinery.

It seems that on some machines there are special math libraries
available which specifically deal with such operations
(havn't tried any of those yet, so take this comment with a bit of care).
If there is a way to make use of them from scipy, that
*could* give some further speed improvement ...

And a (related) question:
 - installing scipy_core works without ATLAS
 - full scipy needs ATLAS
 - if a user installs ATLAS only in the second step,
   will the dot operation from scipy_core
   use dotblas, or not?
   (Also wondering about how eggs will handle such cases ...)

Best,

Arnd




More information about the SciPy-user mailing list