Hello again,
I am having problems running in parallel using mpi. The program works fine with one processor, but does not do anything when I use mpi (it compiles using mpi). The error is below, and the script file I used to run in the queue is
#!/bin/bash
#PBS -l nodes=6,walltime=28:00:00
#PBS -j oe
module load netcdf-4.0.1
module load mvapich/1.1-intel
cd /home/rpacheco/ROMS/MyTest/cdw/CDW1
mpiexec oceanG ocean_cdw.in
Here is the error message that I got (also attached to this message). I will appreciate your help.
Model Input Parameters: ROMS/TOMS version 3.5
Tuesday - July 12, 2011 - 5:18:11 AM
-----------------------------------------------------------------------------
Model Input Parameters: ROMS/TOMS version 3.5
Tuesday - July 12, 2011 - 5:18:11 AM
-----------------------------------------------------------------------------
Model Input Parameters: ROMS/TOMS version 3.5
Tuesday - July 12, 2011 - 5:18:11 AM
-----------------------------------------------------------------------------
Model Input Parameters: ROMS/TOMS version 3.5
Tuesday - July 12, 2011 - 5:18:11 AM
-----------------------------------------------------------------------------
^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
Operating system : Linux
CPU/hardware : x86_64
Compiler system : ifort
Compiler command : /packages/intel-ct/Compiler/11.1/072/bin/intel64/ifort
Compiler flags : -heap-arrays -fp-model precise -g -check uninit -ftrapuv -traceback -free -free -free
problems running in parallel using mpiexec
problems running in parallel using mpiexec
- Attachments
-
- runjobMPI.o3821071.gz
- (1.77 KiB) Downloaded 182 times
Re: problems running in parallel using mpiexec
In ROMS, just about every print goes something like:
What you are getting is all of the processes thinking they are the Master. You probably want to try running something much simpler than ROMS to see if the MPI "hello, world" sorts of tests work. Have you tried any such code? Do you have system people to work with to sort this out?
Code: Select all
IF (Master) WRITE (stdout,10)