We are happy to announce a new Anaconda add-on product called MKL Optimizations”, which allows packages in Anaconda to take advantage of the Math Kernel Library (MKL) by Intel.

MKL was released by Intel 10 years ago, and provides optimized math routines for science, engineering, and financial applications. Core math functions include BLAS, LAPACK, sparse solvers, FFTs, and vector math. NumPy, SciPy, scikit-learn and NumExpr are open source packages which may optionally be linked against MKL, and are included in Anaconda. The new MKL Optimizations add-on includes the MKL (version 11.0) runtime library, as well as versions of NumPy, SciPy, scikit-learn and NumExpr, which are compiled and linked against MKL, so that you have everything in place to run your same code faster. In addition, we have created a small interface to MKL service functions, which allows changing the number of threads MKL is using at runtime.

If you are using Accelerate on Linux, you already have this functionality. However, we now extend these MKL optimizations to MacOSX and Windows, such that all Anaconda supported platforms and architectures benefit. With the launch of the MKL Optimizations, we are also releasing a new version of Accelerate, which includes MKL Optimizations on all platforms.

Try out MKL Optimizations

As with all our commercial products, we offer a 30 day trail period. To install MKL Optimizations:

$ conda update conda
$ conda install mkl

Existing Accelerate customers should update Accelerate via conda:

$ conda update conda
$ conda update accelerate