[meta-freescale] Numpy using optimized librairies

[Jean-Michel Hautbois]

Hi all,

I am using numpy in a program which mainly does matrix multiplications.
I use the nitrogen6x board.
I have profiled it, and as expected, the function numpy.core.multiarray.dot
is the one which consumes the most cpu.
I saw several librairies which may help (BLAS, LAPACK, ATLAS, etc.) but I
can't find any of these in the Yocto recipes. Are those only usable on
Intel processors ?

Another way I wish to explore is using pyGL but maybe do you have better
suggestions ?
The matrix are ~(100*20000) in size...

Thanks in advance,
JM
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/meta-freescale/attachments/20140402/ccee52c1/attachment.html>