This work is based off of #468 and should ideally be rebased after that is merged.

Please rebase later today or tomorrow.

I think we decided we weren't bothering with this one just yet, right?

Sure. We can put it lower on the list.

I went ahead and rebased onto #433 and there were no conflicts.

@xylar could you rebase this again? There were some conflicts on some python import lines that I was unsure of.

I rebased and there was no conflict.

@xylar, can you describe the case, computer, and python package you tested with, and what the directory init/e3sm_coupling/prescribed_ismf contained when you were done?
I tried on grizzly on a compute node using QU240wISC and package compass_0.1.2. I first run the init step. Then I get this:

cd e3sm_coupling/
./run.py
****** prescribed_ismf ******
running: mpirun -np 36 /usr/projects/climate/SHARED_CLIMATE/anaconda_envs/base/envs/compass_0.1.2/bin/ESMF_RegridWeightGen --source /tmp/tmpar9c01qz/src_mesh.nc --destination /tmp/tmpar9c01qz/dst_mesh.nc --weight ./map_5600.0x5600.0km_1.0km_Antarctic_stereo_to_oQU240wISC.nc --method conserve --netcdf4 --no_log --src_regional --dst_regional --ignore_unmapped
Operating system error: Cannot allocate memory
Allocation would exceed memory limit

But this is a brand new functionality. So if it worked for you, I'm happy to merge it and work out the kinks for this grizzly set-up later.

@mark-petersen, sometimes a user takes up all the memory on the compute node. I've had this issue before, sure is frustrating.

@pwolfram Are you thinking of a login node? I should be the only one on that compute node. There must be some problem with the size of the files it is loading.

Sorry @mark-petersen , I was thinking head node -- I agree that the file might be too big.

@mark-petersen, I believe you are running with the serial version of compass. If that's the case, you will do horrible things if you try to run ESMF_RegridWeightGen with mpirun. It might work but it will just be running 36 independent processes trying to write the same file. You need to set nprocs to 1 until we can get the parallel environment working again (#480). This problem should be with any mapping files you're creating, not just this case, by the way.