Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.
Announcements
This community is designed for sharing of public information. Please do not share Intel or third-party confidential information here.
6588 Discussions

Question on Redistribution related to numpy/scipy

Lynn_O_
Beginner
258 Views

The only binary distribution for 64-bit numpy/scipy that I've been able to find is built using MKL.  Is there any problem with distributing programs that I write using these libraries?

I'll mention that I had no idea what MKL was, or that it was included in the distribution, until I stumbled across some of the dll names while debugging and took the time to search out the information.  Under Python you generally just install the needed libraries and the dependencies are taken care of by the install tools; in my case I needed matplotlib, which brought along numpy/scipy, which eventually led me here.  A request for information from Intel support led to instructions to ask for an official answer here.

My first preference would be to find a 64-bit binary distribution for numpy/scipy that does not use the MKL runtime, or workable instructions for building from source.  Second choice, if redistribution is not an issue, would be figuring out how to make my build tools work with the MKL runtime.  So far I haven't found anyone who knows anyone who has succeeded in building for 64-bit without using the MKL runtime.

Thanks for any info...

0 Kudos
1 Reply
TimP
Black Belt
258 Views
It seems the support people should have referred you to an article such as http://software.intel.com/en-us/articles/redistribution-of-application-binaries-built-for-microsoft-... (there is a linux version of that article, but the answer is much the same). If you have a license including MKL, you can redistribute library shared objects as the article indicates. If you don't want to support MKL threading, you can link statically with mkl_sequential, as described in the link advisor, which we have been urging should be restored to this web site. This would eliminate dependence on MKL runtime library. I don't know why people should have trouble building these applications with open source or free alternatives to MKL. It's good to hear that MKL is more convenient than those alternatives (which is why I'm guessing you're on Windows).
Reply