IPARS Parallel Runs

• Prepare an input file and copy it to the directory where you want to run ipars
  gxl@bevo2:~/ipars/IPARSv2/work> mkdir Run
  gxl@bevo2:~/ipars/IPARSv2/work>cd Run
  gxl@bevo2:~/ipars/IPARSv2/work>cp ../data/hydro1.dat ./
  NOTE: Input file can be any that runs in serial mode. Special (additional keywords) may include METHOD, NPROCS.
• Bevo2 uses a queuing system to submit, manage, and monitor the jobs being performed. You must write a submit script in order to submit a job to the queue.
• Create a job submit script file (gxljob)
  #!/bin/bash
  #$ -pe mpich 16
  #$ -cwd
  #$ -N gxljob
  #$ -S /bin/bash
  Source ~/.bashrc
  module load intel
  module load mvapich
  module load mkl
  /share/apps/mvapich/intel/bin/mpirun \
  -np $NSLOTS -machinefile $TMP/machines \
  ipars
• Create batch file IPARS.IN
  gxl@bevo2:~/ipars/IPARSv2/work/Run> cat IPARS.IN
  Hydro1.dat
  Hydro1.out
• Submit bevojob to the queue
  gxl@bevo2:~/ipars/IPARSv2/work/Run> qsub gxljob
• Check the status of your job
  gxl@bevo2:~/ipars/IPARSv2/work/Run> qstat
  job-ID prior name user state submit/start at queue master ja-task-ID
  --------------------------------------------------------------------
  774 0 gxljob gxl qw 08/07/2005 23:39:47
  gxl@bevo2:~/ipars/IPARSv2/work/Run > qstat
  job-ID prior name user state submit/start at queue
  --------------------------------------------------
  774 0 gxljob gxl t 08/07/2005 23:39:55 compute-0- MASTER
      0 gxljob gxl t 08/07/2005 23:39:55 compute-0- SLAVE
      0 gxljob gxl t 08/07/2005 23:39:55 compute-0- SLAVE
  774 0 gxljob gxl t 08/07/2005 23:39:55 compute-1- SLAVE
      0 gxljob gxl t 08/07/2005 23:39:55 compute-1- SLAVE
• Cancel your job
  gxl@bevo2:~/ipars/IPARSv2/work/Run> qdel <jobID>
• View hydro1.out for standard output
• View <jobname>.o<jobID> for screen output
• View <jobname>.e<jobID> for error message