MSS crahes opening Top View if MKL libraries are not available on Anaconda with osx-64 (OSX 10.10)

Issue #304 new
Erik Kretschmer created an issue

Here is a dependency check possibly missing and/or an issue with Anaconda2

Running MSS 1.6.2 on Anaconda with osx-64 (OSX 10.10). All clean and new installation. MSS can start fine and the table view can be opened. As soon as one opens the top view, the process crashes with the following error in the console (no error in the MSS log):

Intel MKL FATAL ERROR: Cannot load libmkl_intel_thread.dylib.

The MKL libraries are not supported on osx-64 10.10 and require OSX >= 10.12.

Replacing the numpy and scipy in Anaconda with the nomkl versions fixed the issue:

(mssenv) ekmacpro:/ kretsch$ conda install nomkl
Fetching package metadata .............
Solving package specifications: .

Package plan for installation in environment /Users/kretsch/anaconda2/envs/mssenv:

The following NEW packages will be INSTALLED:

    nomkl: 1.0-0                

The following packages will be DOWNGRADED:

    numpy: 1.13.1-py27hd567e90_2 --> 1.13.1-py27_nomkl_0      [nomkl]
    scipy: 0.19.1-py27hf01dd8f_3 --> 0.19.1-np113py27_nomkl_0 [nomkl]

Proceed ([y]/n)? y

nomkl-1.0-0.ta 100% |##########################################################################################################| Time: 0:00:00 401.35 kB/s
numpy-1.13.1-p 100% |##########################################################################################################| Time: 0:00:02   1.68 MB/s
scipy-0.19.1-n 100% |##########################################################################################################| Time: 0:00:02   5.67 MB/s

With the nomkl downgrade, the top view works as expected.

Comments (4)

  1. Reimar Bauer

    This is currently an issue of anachonda 4.

    The actual issue is that Anaconda linked with mkl, but not with libmkl_core.so, thus it has a missing symbol, and can be seen by ...

    some more info in that discussion https://github.com/scikit-learn/scikit-learn/issues/5046

    and explaination of options if mkl makes trouble

    https://www.anaconda.com/blog/developer-blog/anaconda-25-release-now-mkl-optimizations/

    We could add nokml with a preprocessing selector for OSX.

    https://conda.io/docs/user-guide/tasks/build-packages/define-metadata.html#preprocessing-selectors

    If you compare with some of the colleagues none OSX systems can you say something about that you think it is slowed down for some of the actions you do?

  2. Log in to comment