MPAS standalone: Add gnu >=10 support on Cori

[xylar]

This merge adds the -fallow-argument-mismatch flag to the gnu-nersc target for MPAS standalone builds. This got missed in #4821.

[BFB]

[xylar]

Testing

I compiled the code on Cori-Haswell with this branch and the updated Gnu modules (#4839), and ran the compass pr test suite.

The ocean/global_ocean/QU240/PHC/files_for_e3sm test failed on one of the calls to ESMF_RegridWeightGen, so that deservest some follow-up in compass (MPAS-Dev/compass#335). But other tests were successful.

[xylar]

@mark-petersen, @matthewhoffman, @trhille and @akturner, this is ready for you to review. (@trhille, I can't add you as a reviewer so please just comment below).

[trhille]

This enabled me to build MALI in COMPASS on Cori using @xylar's Spack instance from /global/cfs/cdirs/e3sm/software/compass/cori-haswell/test_spack

I cherry-picked commit ef2a1d9 onto MALI-Dev/develop and built using:
make clean gnu-nersc ALBANY=true DEBUG=true

I still get warnings about rank mismatch, but they are no longer errors. I get a dozen or so warnings like this one from mpas_global_sum_mod.F:

Warning: Rank mismatch between actual argument at (1) and actual argument at (2) (rank-1 and scalar)
mpas_global_sum_mod.F:1470:25:

 1470 |       call MPI_ALLREDUCE(localSum, globalSum, 1, MPI_INTEGER8, &
      |                         1
......
 6407 |       call MPI_ALLREDUCE(localSum, globalSum, nFields, &
      |                         2

As well as another dozen from mpas_dmpar.F:

Warning: Rank mismatch between actual argument at (1) and actual argument at (2) (scalar and rank-2)
mpas_dmpar.F:941:28:

  941 |          call MPI_Allreduce(sendbuf, recvbuf, 1, MPI_2INTEGERKIND, MPI_MINLOC, dminfo % comm, mpi_ierr)
      |                            1
......
 1733 |         call MPI_Allreduce(nUniqueNeededList, totalSize, 1, MPI_INTEGERKIND, MPI_MAX, dminfo % comm, mpi_ierr)
      |                           2 

I ran the full_integration landice test suite, which failed during test execution for landice_hydro_radial_restart_test
and landice_hydro_radial_decomposition_test and during validation for landice_circular_shelf_decomposition_test and landice_thwaites_restart_test.

[xylar]

Thanks @trhille!

I ran the full_integration landice test suite, which failed during test execution for landice_hydro_radial_restart_test
and landice_hydro_radial_decomposition_test and during validation for landice_circular_shelf_decomposition_test and landice_thwaites_restart_test.

I just want to note that, at least as far as I understand it, we don't have any reason to think these errors have anything to do with the changes in this PR.

[xylar]

I still get warnings about rank mismatch, but they are no longer errors.

I believe that's the nature of the flag I added. I also always see the warnings. Presumably, it's something we'll want to address later on if we can.

[mark-petersen]

I compiled ocean, sea ice, and mali with gnu 10 on cori and this works. Specifically, with today's compass/master 127b6d08:

./conda/configure_compass_env.py --conda ~/miconda3 --compiler gnu
source load_dev_compass_1.0.0_cori-haswell_gnu_mpt.sh
module list
  17) gcc/10.3.0
cd /global/homes/m/mpeterse/repos/E3SM/pr/components/mpas-ocean
make gnu-nersc USE_PIO2=true OPENMP=false DEBUG=true GEN_F90=true

Also for mpas-seaice and mpas-albany-landice.

[mark-petersen]

@matthewhoffman and @akturner if a successful compile of your cores with gnu10 is sufficient testing, you can approve based on that.

[xylar]

Thanks @mark-petersen, @matthewhoffman and @akturner!

[jonbob]

passes sanity testing -- merged to next

[xylar]

Thank you, @jonbob!

[jonbob]

merged to master

[xylar]

@jonbob, thanks again!