The OASIS Coupler Forum

  HOME

Coupled ICON+NEMOv3.6+HD via OASIS3-MCT v4.0 hangs at MPI_COMM_Split in oasis_enddef

Up to Specific issues in real coupled models

Posted by Anonymous at August 11 2022

Hello everyone,

I'm working on a coupled model of ICON + NEMO v3.6 + HD via OASIS3-MCT v4.0 (without XIOS) on the computing system Levante (DKRZ, Hamburg, Germany). The coupled model hangs at line:

CALL MPI_COMM_Split(mpi_comm_global,icpl,1,newcomm,ierr)

of subroutine oasis_enddef.
I printed out the value of mpi_comm_local just before the call MPI_COMM_Split and saw the value of 9 for ICON model, meanwhile in debug.root.01 (for ICON):

(oasis_init_comp) mpi_comm_local =            3

The values of mpi_comm_local printed for NEMO and HD models are all 3, like in debug.root.02 and debug.root.03, respectively.

Within another coupled model COSMO-CLM + NEMO v3.6 + HD via the same OASIS3-MCT version, on the same computer, the values of mpi_comm_local are all 3, either printed by myself or in the debug.root files.

Do you have any idea what would cause that hanging problem?

Cheers,
Ha

Posted by Anonymous at August 25 2022

Dear Ha,

I am not sure what is happening. You should have a look at how the initialisation calls oasis_init_comp, oasis_get_localcomm are implemented in both codes and compare. I suppose that the components of the coupled model are launched the same way in both case?
Maybe you can reply by sending a mail to oasishelp@cerfacs.fr so that we can interact offline on this?
  With best regards,
 Sophie
Reply to this