Is there a way of 'switching off' or downgrading the use of Intel the Maths Kernal Library (mkl) and using 'raw' (non mkl) Numpy functions?

Asked Oct 16 '21 at 09:54

Active Oct 16 '21 at 09:54

Viewed 65 times

I am suffering with a bug in my code using numpy FFT and one suggestion has been that it could be due to the use of the Intel Maths Kernal Library (mkl) in Anaconda.

To test this, I would like to re-run without using mkl.

So is there a way of using a non mkl version of numpy with Python in Anaconda?

My discussion in the numpy issues log on github relates: https://github.com/numpy/numpy/issues/20113

asked Oct 16 '21 at 09:54

If you are on linux there's a nomkl package https://stackoverflow.com/questions/46656367/how-to-create-an-environment-in-anaconda-with-numpy-nomkl – Rani Sharim Oct 16 '21 at 10:03
Rani - Thank you. A useful reference. Sadly, I am using Windows. But I will follow that lead and see if it goes anywhere. – Oct 16 '21 at 10:52

Is there a way of 'switching off' or downgrading the use of Intel the Maths Kernal Library (mkl) and using 'raw' (non mkl) Numpy functions?

0 Answers0