search

NOAA-GFDL / FMS

GFDL's Flexible Modeling System
Other
92 stars 134 forks source link

MOM6 ocean_only model hanging with gfortran 11.3 and FMS 2023.01.01 #1254

Open kshedstrom opened 1 year ago

kshedstrom commented 1 year ago

I'm running on chinook at UAF and getting some of my configurations to hang. It runs on gaea with ifort22. It runs with FMS 2023.01. It runs in debug mode. It fails on chinook with FMS 2023.01.01 and gfortran 11, but only when using multiple cores. CCS2 (coupled MOM6-SIS2) also fails on chinook.

The files for the first case to hang in my test suite are on gaea in: /lustre/f2/dev/Katherine.Hedstrom/ESMG/ESMG-configs/ocean_only/Supercritical

uramirez8707 commented 1 year ago

@kshedstrom Have you tried running this on Gaea with gfortran?

kshedstrom commented 1 year ago

Yes, I get a seg fault.

uramirez8707 commented 1 year ago

can you provide us with instructions on how to reproduce the error?

kshedstrom commented 1 year ago

The Makefile is /lustre/f2/dev/Katherine.Hedstrom/ESMG/Makefile.test. It is currently pointing to the source code under /lustre/f2/dev/Katherine.Hedstrom/ESMG/ESMG-configs/src that is being used. You need to load these:

  1001  15:04   module purge
  1003  15:04   module load PrgEnv-gnu
  1004  15:04   module avail
  1005  15:05   module load cray-hdf5
  1006  15:05   module load cray-netcdf

then "make -f Makefile.test". It will create the executable used by the script: /lustre/f2/dev/Katherine.Hedstrom/ESMG/ESMG-configs/ocean_only/Supercritical/run.sub Note that this is on c5.

kshedstrom commented 1 year ago

I should add that the Makefile only behaves rationally when used under tcsh. Or just simply ask it to make the explicit executable under ocean_only.