2
votes

I am using Julia Version 0.6.0-pre.alpha.47 under MacOS 10.13.

When attempting using PyPlot, any basic "plot" command causes a fatal error with the message:

Intel MKL FATAL ERROR: Cannot load libmkl_intel_thread.dylib.

I have tried rebuilding PyCall with a number of different python installations. I have no issue using matplotlib from python.

I see the same issue in Julia versions 0.4.5 and 0.5.0.

Any suggestions?

2
I was able to resolve this by switching to nomkl version in python and pointing the Julia PYTHON environment variable to this python installation. - Shawn

2 Answers

1
votes

This can be avoided by adding your Conda library directory (in 0.6 for example this is $HOME/.julia/v0.6/Conda/deps/usr/lib) to your LD_LIBRARY_PATH before launching Julia.

See https://github.com/JuliaPy/PyPlot.jl/issues/315#issuecomment-334179119

The root cause may be fixed soon but this has worked to mitigate the issue for now.

0
votes

I was having the same issue with julia 0.6 on macOS 10.12.5.

To solve the problem, in a fresh install of julia, I did:

Pkg.add("Conda")
using Conda
Conda.add("nomkl")
Conda.add("numpy")
Conda.add("scipy")
Pkg.add("PyCall")
Pkg.add("PyPlot")