topic Todd, in Intel® Distribution for Python*

Anaconda Python

GAnthony_R_Intel — Sun, 08 Apr 2018 20:10:08 GMT

Is the Intel version of Python installed when I grab the latest release of Anaconda? Or is the Intel version still a separate download? Will Anaconda ever use the Intel version by default?

Thanks.

-Tony

Hello Tony,

Todd_T_Intel — Tue, 24 Apr 2018 21:08:16 GMT

Hello Tony,

No, Anaconda does not use the Intel build of python as the default in their Anaconda product distribution. However, we work very closely with them, and they have adopted into their default distribution several of the performance enhancements we initially provided.

For example, their default numpy package now comes with MKL linked for BLAS operations. They have also included several other patches to numpy that optimize memory operations, allow vectorization, and leverage more of MKL along with our mkl_fft and mkl_random packages for acceleration of those workloads. Intel's OpenMP is also included on their main package channel (named intel-openmp).

All of our packages are available on the Intel channel on the Anaconda Cloud*: anaconda.org/intel.

We are continuing to work with Anaconda to provide the broadest set of users the best performance available.

Best Wishes, Todd

Todd,

gaston-hillar — Sat, 26 May 2018 04:46:27 GMT

Todd,

It would be great if the information you are providing in this thread is included in the documentation. I explained what you mention many times to Python developers that work with Anaconda and the lack of documentation about this performance improvements brought by Intel is a problem. I haven't read all the documentation. However, I didn't find any mention of Intel's improvements in most of the documentation I've read. Hence, I think it should be a good idea to include these improvements in the documentation.

I've been working with Intel Distribution for Python since its beta versions and it would be great for many Anaconda users to know that many performance improvements are brought by the hard work with Intel. ;)