For those of you interested by using NumPy with the MKL, here is a rough benchmark :
I will continue improving the post with new tests and try to make it more robust.
Comments or feedback are welcome ;-)
Didrik, thanks for the information. I have a couple of notes, make sense to explicitly add the size of the task you are benchamarking and to do the same things on the latest architectures like Nechalem...
thanks a lot. Gennady
I have updated the scripts. It is attached for testing. Please report your test results ;-)
Regarding the Nechalem architecture, I have no access to those one at the moment.