Hi John, > I got the MPI error about init being disallowed, etc. I think this problem has been observed and discussed earlier.
Can you elaborate on this? What was the exact error message you saw? If the conda package is having MPI problems, it is likely because it's trying to use a different MPI on the system rather than the one that is bundled in the conda environment. I would double check that `which mpirun` points to the mpirun in your conda environment, and make sure you don't have any variables (PATH, LD_LIBRARY_PATH, etc.) pointing to specific MPI installations in your .bashrc or similar. For the second error, it looks like it's loading an older version of Meep instead of the one in the conda environment. I would check your .bashrc for anything pointing to an old meep and make sure PYTHONPATH is unset. You can also run `ldd ~/.conda/envs/pmp_test/lib/python3.6/site-packages/meep/_meep.so` to make sure it's linking with the correct libmeep.so. I verified that both the 1.5 release, and the development package work on a clean Ubuntu 18.04. Chris On Sat, Aug 4, 2018 at 9:01 AM, John Weiner <johwei...@gmail.com> wrote: > Hello to all, > > Having gotten the impression that future development of meep will be > essentially in python meep, I have taken the first tentative baby steps to > get python meep working on my pc Ubuntu box. Scheme meep, both stand-alone > and parallel, is now working on this box running Ubuntu 18.04 LTS. > > Using the CONDA package, I was able to install one-core python meep > without incident. It works. > > Then I tried to install the parallel version using the instructions at > https://meep.readthedocs.io/en/latest/Installation/#conda-packages, and > in particular, > > conda create -n pmp -c chogan -c conda-forge pymeep-parallelsource activate > pmp > > But when I tried to run a .py script using > > mpirun -np 2 python foo.py, > > I got the MPI error about init being disallowed, etc. I think this > problem has been observed and discussed earlier. > > I then tried the nightly builds and was able to create the environment > pmp_test. When I entered the same mpirun command line as above, I got the > following message, > > File "<string>", line 1, in <module> > File > "/home/john/.conda/envs/pmp_test/lib/python3.6/site-packages/meep/__init__.py", > line 8, in <module> > from . import _meep > ImportError: > /home/john/.conda/envs/pmp_test/lib/python3.6/site-packages/meep/_meep.so: > undefined symbol: _ZN4meep6fields17add_volume_sourceENS_9componentERKNS_ > 8src_timeERKNS_6volumeEPKcS9_St7complexIdE > > I thought that maybe the first line in the foo.py file, > > import meep as mp > > was not correct. So I deleted it and ran the script again. Then I got > the error message, > > Traceback (most recent call last): > File "straight_waveguide.py", line 2, in <module> > cell = mp.Vector3(16, 8, 0) > NameError: name 'mp' is not defined > > which I take to mean that maybe the meep library is not "imported" without > the import command or there is some more serious problem in the meep > library. > > I don't know what to do to move forward with python parallel meep and > would appreciate any help. > > Best regards, > > John > > > > > _______________________________________________ > meep-discuss mailing list > meep-discuss@ab-initio.mit.edu > http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss >
_______________________________________________ meep-discuss mailing list meep-discuss@ab-initio.mit.edu http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
