[Numpy-discussion] numpy arrays, data allocation and SIMD alignement

David Cournapeau cournape at gmail.com
Thu Aug 9 11:03:31 EDT 2007

On 8/9/07, Charles R Harris <charlesr.harris at gmail.com> wrote:
> On 8/9/07, David Cournapeau <david at ar.media.kyoto-u.ac.jp> wrote:
> > Charles R Harris wrote:
> > >
> > > Ah, you want it in C.
> > What would be the use to get SIMD aligned arrays in python ?
> If I wanted a fairly specialized routine and didn't want to touch the guts
> of numpy, I would pass the aligned array to a C function and use the data
> pointer. The python code would be just a high level wrapper. You might even
> be able to use ctypes to pass the pointer into a library function. It's not
> necessary to code everything in C using the python C API.
I certainly do not argue on this point. But if it was specialized,
there would be no point putting in in numpy in the first place. What I
hope is that at some point, the aligned allocators can be used inside
core numpy to optimize things internally (ufunc, etc...). Those
facilities would be really useful for many optimized libraries, which
are all C: as such, doing it in C makes sense, no ?


More information about the NumPy-Discussion mailing list