After a long and agonizing period of struggle with individual package management, I found Intel Python and I was really hopeful. It could fix all my problems as I already had MKL and MPI libraries installed with Parallel Studio. But no luck.
I use Visual Studio 2015, have Intel Parallel Studio and Visual Studio Python Tools. I installed Intel Parallel Studio XE 2016 Cluster Edition after Visual Studio and python tools. And finally I installed Intel Python after an unsuccessful trial with Anaconda. My OS is Win 10 64 bit.
Python version should be 3.3 and higher for the code I am trying to run.
My first problem:
I do not know why but numpy MKL libraries are completely wrong. It is
= C:\Program Files (x86)\IntelSWTools\compilers_and_libraries_2016.0.110\windows\mkl\lib\intel64;C:\Program Files (x86)\IntelSWTools\compilers_and_libraries_2016.0.110\windows\compiler\lib\intel64
But it should be = D:\IntelSWTools\...2016.2.180\....
I checked system variables and the only thing that points to C:\Program Files (x86) is INTEL_DEV_REDIST with path C:\Program Files (x86)\Common Files\Intel\Shared Libraries\
Could it be the reason?
But when I consider that even the version is wrong I think that the issue is with installer configurations.
I do not know why it is not asked during installation. I could enter the correct library for MKL.
Is it possible to point libraries during installation? I extracted the exe file. If there is a configuration file in that folder, I can change it.
The second issue:
I could somehow deal with numpy and others but the main reason I installed Intel Python was to have mpi4y installed as written on the website:"Intel® Distribution for Python* 2017 Beta program is now available. The Beta product adds new Python packages like scikit-learn, mpi4py, numba, conda, tbb (Python interfaces to Intel® Threading Building Blocks) and pyDAAL (Python interfaces to Intel® Data Analytics Acceleration Library). "
But I get the error that there is no mpi4py module. I won't even try with conda or pip because I am sure that it will install with wrong directories. Or it won't work. (This was the cause of all my suffering for the past couple of days). There is always an error with mpi4py.
And my last question is about the path variable. I chose to append path variable during custom installation. But when I check system variables there is no such a thing as pythonpath in the path variables or an independent variable as pythonpath. Was it wrong to choose append option?
I wish you all a nice weekend,
First off, thank you very much for downloading our Beta release!
To address your first question, I would like to help you troubleshoot and will be contacting you directly via email.
As for your second question, Linux is the only platform where we support mpi4py for Beta. We are exploring including mpi4py on Windows for a future release. I'll look into how we can make this more clear in our documentation.
For your last question, you did everything correctly. Appending simply adds Intel Python to PATH and does not use the PYTHONPATH environment variable.
Please let me know if you have any other questions or concerns and I'll do my best to address them. Thanks!
Maybe others could also be interested.
Is it possible to change numpy library paths after installation? If it is possible how can we do it?
How should we fill these paths ('library_dirs', 'libraries', 'include_dirs') in the config file? I could not find an mkl include folder in the IntelPyhton35 folder.