MPI is not being initialized for SLUMP

[cobayasampler]

MPI is not being initialized for SLUMP

[JesusTorrado]

Hi @dwhan89

Glad you like it, and thanks for the comprehensive bug report!

Indeed, the MPI environment check that we perform is by no means ideal and needs some generalisation. I'll try to fix it in the next few days (I need to check the encapsulation of some MPI calls before applying your fix, because IIRC, this would cause some errors for systems without a working MPI installation).

I'll update here when it's done.

Cheers
Jesús

[JesusTorrado]

Hi @dwhan89

You should be able to use the last commit as is. Please, let me know if it works and whether you observe any unexpected side effect.

At some point, I intended to create a doc page similar to the one for the Amazon Cloud (https://cobaya.readthedocs.io/en/latest/cluster_amazon.html), summarizing the non-trivial necessary configuration steps for Cori, but never got around to do it. If you have the time to write such a summary, I would be very grateful and happy to include it in the docs with you as the author, of course.

Cheers
Jesús

[dwhan89]

Hi, @JesusTorrado

Thanks for the fast fix! I will close the issue now.

And, sure, I am happy to help you with the summary page. So far, everything ran as I expected, except this minor bug. I will let you know if I find any other Cori related quirks.

Cheers!
DW

[JesusTorrado]

Hi @dwhan89

If everything works as expected, then there is no need to create a special page for Cori. Did you have to follow the instructions at http://www.nersc.gov/users/data-analytics/data-analytics-2/python/anaconda-python/#buildingmpi4py ? I remember needing to do it if I wanted to use GNU compilers, but maybe I was not doing things right.