[gmx-users] 1replica/1cpu problem

[Mark Abraham]

On 19/07/2012 12:32 AM, francesco oteri wrote:
> Dear gromacs users,
> I am trying to run a replica exchange simulation using the files you find
> in http://dl.dropbox.com /u/40545409/gmx_mailinglist/inputs.tgz
>
> The 4 replicas have been generated, as following:
> grompp -p rest2.top -c 03md.gro -n index.ndx -o rest2_0  -f rest2_0.mdp
> grompp -p rest2.top -c 03md.gro -n index.ndx -o rest2_1  -f rest2_1.mdp
> grompp -p rest2.top -c 03md.gro -n index.ndx -o rest2_2  -f rest2_2.mdp
> grompp -p rest2.top -c 03md.gro -n index.ndx -o rest2_3  -f rest2_3.mdp
>
> The simulation was started with command, using gromacs 4.5.5  with the
> latest bug fix:

Which bug fix? How did you apply it?

Mark

> mpirun -np 4  mdrun_mpi -s rest2_.tpr -multi 4 -replex 1000 >& out1
>
> giving the following error:
>
> [etna:10799] *** An error occurred in MPI_comm_size
> [etna:10799] *** on communicator MPI_COMM_WORLD
> [etna:10799] *** MPI_ERR_COMM: invalid communicator
> [etna:10799] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
> --------------------------------------------------------------------------
> mpirun has exited due to process rank 0 with PID 10796 on
> node etna exiting without calling "finalize". This may
> have caused other processes in the application to be
> terminated by signals sent by mpirun (as reported here).
> --------------------------------------------------------------------------
> [etna:10795] 3 more processes have sent help message
> help-mpi-errors.txt / mpi_errors_are_fatal
> [etna:10795] Set MCA parameter "orte_base_help_aggregate" to 0 to see
> all help / error messages
>
>
> The nice thing is that the same error doesn't appear either if I use
> the 4.5.5 without applying tha patches!!!
>
> mpirun -np 4  mdrun_mpi -s rest2_.tpr -multi 4 -replex 1000 >& out2
>
> or the bug fixed with multiple processors per replica:
>
> mpirun -np 8  mdrun_mpi -s rest2_.tpr -multi 4 -replex 1000 >& out3
>
> Since I have to use more then 4 replicas, I need to run 1cpu/replica.
>
> Has someone any idea of the probem?
>
> Francesco