is it worth to go native?

[Georg Heiler]

Is it worth to go native without MKL? I have a matrix of about 450 elements in size (400-500) and want to perform some multiplications and some decompositions to calculate: 

- a gaussian process

- correlation matrix

My problem is that I will have several million of these matrices and I wonder what is good to move forward.

Best,

Georg

[Georg Heiler]

In particular, I would be interested in opinions on:

- https://nd4j.org

- http://ejml.org/wiki/index.php?title=Main_Page

- https://github.com/scalanlp/breeze