useMPI vs compiler_suite vs compiler_mpi #1297
Description
In different machine files, we use these three switches:
- useMPI
- compiler_suite
- compiler_mpi
They do similar, sometimes identical things. This results in strange choose blocks like:
"2.5":
git-repository: https://gitlab.dkrz.de/ec-earth/xios-2.5.git
archfile: ESMTOOLS_generic_oasis_intel
choose_computer.useMPI: # JS: This entire block is a missnomer. It's not the MPI thats imporant but the compiler cray ftn vs intel.
cray_mpich:
archfile: ESMTOOLS_generic_oasis_cray
intelmpi:
archfile: ESMTOOLS_generic_oasis_intel
parastationmpi:
archfile: ESMTOOLS_generic_oasis_intel
"*":
foo: bar
# MA: yet another block, this time for albedo. We cannot fix this now, but in the
# future we should adopt the compiler suite logic for the other computers. Already
# included in ESM-Tools clickup
choose_computer.compiler_suite:
gcc:
archfile: ESMTOOLS_generic_oasis_gcc
intel-oneapi:
archfile: ESMTOOLS_levante_oasis_intel
"*":
foo: bar
# JS: On levante this switch has yet another name. Again only the compiler part matters.
choose_computer.name:
levante:
choose_computer.compiler_mpi:
gnu11_ompi4:
archfile: ESMTOOLS_generic_oasis_gcc
"*":
archfile: ESMTOOLS_levante_oasis_intel
juwels:
choose_computer.compiler_mpi:
gnu2023_ompi2023:
archfile: ESMTOOLS_generic_oasis_gcc
"*":
archfile: ESMTOOLS_levante_oasis_intel
I have a dream...
I dont want these to be called different names... Some day.