BoB Documentation

Helpful hints on submitting jobs to the BoB queues:

For a deeper understanding of the queueing system SGE, please read the SGE user guide.
For MPI jobs (8 queues of 16 processors named mpi8*): use mpi.sh as a simple script example. (A copy of this script may also be found in /home/fs/mpi.sh on bob)
MPI documentation available here
For NAMD jobs, use namd.sh as a simple script example. (A copy of this script may also be found in /home/fs/namd.sh on bob)
Namd documentation available here.
For CHARMM jobs, use charmm.sh as a simple script example. (A copy of this script may also be found in /home/fs/charmm.sh on bob)
Charmm documentation available here.
For Gamess jobs, use gamess.sh
Gamess documentation available here.
Interface with Charmm: to come.
For AMBER jobs, use amber.sh as a simple script example. (A copy of this script may also be found in /home/fs/amber.sh on bob)
Amber documentation available here.
Amber tutorials available here.
For Gaussian jobs, use gaussian.sh as a simple script example. (A copy of this script may also be found in /home/fs/gaussian.sh on bob)
Gaussian documentation here.
For Gaussian parallel jobs, use gaussianp.sh as a simple script example. (A copy of this script may also be found in /home/fs/gaussianp.sh on bob).

A word of warning: when considering using parallel gaussian you should remember that spawning slaves and interprocess communications are orders of magnitude slower than pure computation. It's quite common that for small problems parallel gaussian may take longer than the sequential gaussian. In the words of the technical support people at Gaussian: "The principle methods which benefits from parallelization are HF or DFT energies, gradients and frequencies and MP2 energies and gradients.
For Jaguar jobs, use jaguar.sh as a simple script example. (A copy of this script may also be found in /home/fs/jaguar.sh on bob).
Jaguar documentation available here.

Rule of thumb for number of processors: when you are choosing the number of processors to use for a parallel Jaguar job, take the number of basis functions for the job and divide by 100 (for HF or DFT jobs) or 80 (for LMP2 jobs), then discard any portion of this number after the decimal place. This number is the maximum number of processors advised for an efficient parallel run. For instance, if your molecule had 486 basis functions, the maximum number of processors advised for an HF or DFT calculation would be 4, and the maximum number of processors for an LMP2 job would be 6.
For Macromodel jobs: to come.
Macromodel documentation available here.
For Maestro interface with SunGrid Engine scheduler: to come.
Maestro documentation available here.
For Towhee jobs, use towhee.sh as a simple script example. (A copy of this script may also be found in /home/fs/towhee.sh on bob).
Towhee user manual available here, and towhee example available here.
For LAMMPS jobs, use lammps.sh as a simple script example. (A copy of this script may also be found in /home/fs/lammps.sh on bob).
Lammps user manual available here, and Lammps example available here.
For serial jobs (protomol serial, etc...), use serial.sh.
For further help / suggestions, please contact me at jducom@nd.edu or at x10381.

Updated on July 8, 2002