[Numpy-discussion] vectorizing loops

Francesc Altet faltet@carabos....
Fri Nov 2 08:55:21 CDT 2007


A Thursday 01 November 2007, Timothy Hochberg escrigué:
> On Nov 1, 2007 7:14 AM, David M. Cooke <cookedm@physics.mcmaster.ca> > 
> Another issue is that numexpr is still in the scipy sandbox, so
> > only those who enable it will use it (or use it through PyTables).
> > One problem with moving it out is that Tim reports the compile
> > times on Windows are ridiculous (20 mins!).
>
> While this is true at the default optimization (O2), it compiles
> reasonably quickly at O1 and I've never been able to detect a speed
> difference between versions compiled with O1 versus O2. It would
> probably be sufficient to crank back the optimization on Windows.

Yes. This has been my experience too on Windows/MSVC boxes.

> > Maybe numexpr should become a
> > scikit? It certainly doesn't need the rest of scipy.

Call me intrepid, but I've always felt that numexpr belongs more to 
numpy itself than scipy.  However, I agree that perhaps it should be a 
bit more polished (but not much; perhaps just adding some functions 
like, exp, log, log10... would be enough) before being integrated.  At 
any rate, Numexpr would be a extremely useful complement to NumPy.

My two cents,

-- 
>0,0<   Francesc Altet     http://www.carabos.com/
V   V   Cárabos Coop. V.   Enjoy Data
 "-"


More information about the Numpy-discussion mailing list