GAMESS-US
Useful Links
Access
GAMESS is licensed software. See GAMESS License Agreement
All HECToR users have access to the GAMESS binaries. If you would like access to the HECToR-specific GAMESS source code then please (contact the HECToR Helpdesk).
Running
To run GAMESS you need to add the correct module. The current build of GAMESS on HECToR is dated 1/10/2010:
module add gamess-us
will give you access to the GAMESS binary and the rungms script that is needed to run the job.
Note: The rungms script has been modified from the standard one to enable efficient running on the Cray architecture. The script depends on the $GAMESS_HOME (location of the bin/ and common/ directories) and the $SCRATCH environment variable being set before the script is called. The $GAMESS_HOME variable is set by the module but you must specify the $SCRATCH variable in your job submission script to point to a directory on the /work filesystem that you have write access to (an example of how to do this is shown below in the example job submission script).
Note: You must use less than 32 cores per node for GAMESS to function correctly. We recommend using 16 cores per node on HECToR phase 3 (see example job submission script below).
An example GAMESS job submission script is shown below.
#!/bin/bash --login #PBS -N gamess_job #PBS -V #PBS -l mppwidth=1024 #PBS -l mppnppn=32 #PBS -l walltime=03:00:00 # Change this to your account code #PBS -A z01 # Move to directory that script was submitted from export PBS_O_WORKDIR=$(readlink -f $PBS_O_WORKDIR) cd $PBS_O_WORKDIR module add gamess-us # Set the number of threads to 1 # This prevents any system libraries from automatically # using threading. export OMP_NUM_THREADS=1 # This should be done for performance export ONESIDED_USE_BLOCKING_WAITS=1 # You must set the SCRATCH environment variable for the job to run export SCRATCH=$PBS_O_WORKDIR/scratch # "rungms" takes five options # 1) The name of your input file (without ".inp") # 2) The type of binary (use "asis" on HECToR) # 3) The number of cores in total # 4) The number of cores per node (must be less than 32, we recommend using 16 cores per node) # 5) The number of cores per NUMA region (if using 16 cores per node, this would be 4) rungms my_job asis 512 16 4
If you wish to keep the files in $SCRATCH when your job completes (for debugging or to use to continue runs) you should substitute the rungms script for rungms-keepscr.
