You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In addition to the quest to run ufs-weather-model on my Mac (to date unsuccessful for me with any spack-stack version), I also have used spack-stack (1.6 on linux and 1.7 on mac) to run the simple land-only system ufs-land-driver. This system only requires Fortran, MPI and NetCDF.
As an aside, when I try to help other users of ufs-land-driver test their systems, I put together a simplified version of the driver that only tests MPI. Since my ufs-land-driver is failing with 1.8.0, I went back to test this MPI-only code and am getting this:
mpif90 -o ufsLand.exe ../driver/test_vector_sum.o ../mpi/mpi_land.o
ld: warning: -commons use_dylibs is no longer supported, using error treatment instead
ld: warning: -commons use_dylibs is no longer supported, using error treatment instead
ld: common symbol '_mpi_fortran_argv_null_' from '/Users/michael.barlage/src/testing/netcdf_mpi/mpi_no_netcdf/driver/test_vector_sum.o' conflicts with definition from dylib '_mpi_fortran_argv_null_' from '/Users/michael.barlage/packages/spack-stack-1.8.0/envs/ufs/install/apple-clang/15.0.0/openmpi-5.0.3-hmarleo/lib/libmpi_usempi_ignore_tkr.40.dylib'
collect2: error: ld returned 1 exit status
with minimal loaded modules
$ module list
Currently Loaded Modules:
1) stack-apple-clang/15.0.0 3) zlib/1.2.13 5) stack-openmpi/5.0.3
2) pmix/5.0.1 4) openmpi/5.0.3
This seems similar to this. 1.8.0 uses openmpi/5.0.3 and it seems maybe this issue has been dealt with in openmpi/5.0.4, but I'm out of my range of understanding on these things. I see develop is now using openmpi/5.0.5. So, after that long intro, I guess my questions are:
should I expect 1.8.0 to work for this simple test? and if so, is there something I need to add (flags, etc.) to address this linker conflict?
would you recommend that I use some hash past 1.8.0 that uses an openmpi >= 5.0.4 where this may have been addressed?
My system: MacOS/M3/Sonoma 14.7.2/Xcode 15.4/CLT 15.3
GNU Fortran (Homebrew GCC 13.3.0) 13.3.0
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
As part of the discussion to get ufs-weather-model to work on MacOS, I followed instructions from @natalie-perlin for building 1.8.0 ufs-weather-model-env. This seems to have been successful.
In addition to the quest to run ufs-weather-model on my Mac (to date unsuccessful for me with any spack-stack version), I also have used spack-stack (1.6 on linux and 1.7 on mac) to run the simple land-only system ufs-land-driver. This system only requires Fortran, MPI and NetCDF.
As an aside, when I try to help other users of ufs-land-driver test their systems, I put together a simplified version of the driver that only tests MPI. Since my ufs-land-driver is failing with 1.8.0, I went back to test this MPI-only code and am getting this:
with minimal loaded modules
This seems similar to this. 1.8.0 uses openmpi/5.0.3 and it seems maybe this issue has been dealt with in openmpi/5.0.4, but I'm out of my range of understanding on these things. I see develop is now using openmpi/5.0.5. So, after that long intro, I guess my questions are:
My system: MacOS/M3/Sonoma 14.7.2/Xcode 15.4/CLT 15.3
GNU Fortran (Homebrew GCC 13.3.0) 13.3.0
Beta Was this translation helpful? Give feedback.
All reactions