MPQC on BigBen

To run jobs, first download the template script mpqcScript. You can transfer files to the server using the scp command. Files can be edited on the server using emacs, pico, vi, etc.

The qchemScript file contains the following…

#!/bin/csh

#PBS -N mpqc sets the job name to mpqc (this doesn’t really affect anything… this is just what will show up when you type qstat)

#PBS -l walltime=0:05:00 sets the walltime to 5 minutes – you’ll want longer

#PBS -j oe don’t know and don’t care… it’s just supposed to be there

#PBS -l size=32 tells script to use 32 processors. You can go as large as 128 without losing scaling.

#PBS -m ae tells server to e-mail when job completes or if it aborted due to an error

#PBS -M jjwilke@ccc.uga.edu the e-mail address notice is to be sent to

module load mpqc
cd $PBS_O_WORKDIR
setenv XT_SYMMETRIC_HEAP_SIZE 512M
echo `pwd`
set scratch = ~/scratch/mpqc.$PBS_JOBID
mkdir $scratch
 
cp input.dat $scratch
cd $scratch
 
pbsyod -SN $MPQC_DIR/bin/mpqc input.dat > $PBS_O_WORKDIR/output.dat.$PBS_O_SIZE
 
rm -rf $scratch

Warnings!!!

1. For this submit script, my input file must be named input.dat because of the lines
cp input.dat $scratch
pbsyod –SN $MPQC_DIR/bin/mpqc input.dat > …

2. Never specify anything but a multiple of 2 processors. BigBen is a bunch of 2 processor nodes – so make it an even number.

3. You are allowed a maximum of 12 hours wall time. This machine is for massive jobs (integral direct preferably) that you want done fast.
Any long optimizations should go to Rachel or one of our machines. Keep in mind, however, that MPQC parallelizes pretty well up to as many as 128 processors. Be greedy with MPQC. The priority for your job will not be affected that much.

4. Make sure you submit the PBS script from the same folder your input file is in.

To actually submit the job, you must type

qsub [scriptname]