Sabre

Testing Sabre

Sabre manual

A number of Sabre examples are available for downloading. They may be used to test your implementation of Sabre.

First edit the example .sab file to change the log file specified in the OUT command (so that the example log file is not overwritten)
Start SABRE interactively, and wait for the screen to clear and the <S> prompt to appear.
Read in the example .sab file by issuing the INPUT command: INPUT filename
e.g.:
INPUT wage.sab
Compare the results in the log file produced with those given in the example log file. (In the case of the example wage.sab the example log file is wage.log.)

Running Parallel Jobs

The way in which parallel jobs are run is implementation specific. To run a parallel Sabre job on the HPC at Lancaster a command file has to be written and submitted to run in batch mode. A command file to run parallel Sabre on the example wage.sab in the directory $HOME/sabre/examples would look like the following:

   #!/bin/bash
   #$ -o $HOME/sabre/examples/wage.stdout
   #$ -e $HOME/sabre/examples/wage.stderr
   #$ -S /bin/bash
   #$ -v LD_LIBRARY_PATH=/usr/lib:/usr/local/lib
   #$ -v LD_LIBRARY_PATH_64=/usr/lib/sparcv9:/usr/local/lib/sparcv9
   echo "Got $NSLOTS slots."
   cd $HOME/sabre/examples
   mpirun -np $NSLOTS -machinefile $TMPDIR/machines sabre_parallel < wage.sab

If this command file is called wage_parallel.com it could then be submitted for running with a request for, say, 16 processors as follows:

   qsub -pe mpich 16 wage_parallel.com

Other links: Centre for e-Science | Centre for Applied Statistics