PAW on the Cobalt-Cluster

Other Utilities

Queueing

To queue a job like paw_run or paw_opt you just have to create a little script like that:
(assuming your inputfile chewbacca.1.inp is in your directory ~yoda/forces)

#$ -l PAW
#
cd ~yoda/forces
/usr/people/program/paw/scripts/paw_opt chewbacca.1 > chewbacca.1.echo

To submit this with qsub should do the trick.
It is much better to specify memory requirements for your jobs (all in the #$ lines; please refer here to the
qsub manpages etc., and the info sent out by Serguei).

Remember to cd to your directory first, because the queueing system starts all jobs in your homedirectory!!!

You might also want to have a look at the scripts qpaw (submits paw_opt to PAW-queue) or qpawrun (submits
paw_run to the PAW-queue). They don not specify memeory reqs., either and are given only for convenience.
You might want to copy those e.g. to your bin-directory and modify for your own needs.

qpaw <jobname>

qpawrun <jobname>

Keyword Manipulation

It is quite easy to implement own strategies. It is strongly recommended to have a look at the code of
paw_opt. It uses standard csh-commads (which is quite easy to understand), the paw_run script, and two other
little scripts usefull to manipulate keywords between stages. These are set_directive and set_keyword.

set_directive <jobname> <directive>

replaces the directive in <jobanem>.inp with <directive>.

set_keyword <jobname> <keyword> <value>

replaces the value for <keyword> in <jobanem>.inp with <value> (limitation: this works currently only for "single
parameter" keywords like steps but not for something with more parameters like lattice).

prot2xyz

if you like movies (who doesn't :-) you can use prot2xyz to grab the xyz coords written to the prot-file every time
the restart file is written (usually every 50 steps) into a file that can be read by xmol or molden.

prot2xyz chewbacca.3.prot > chewbacca.3.xyz

paw_opt does that automatically for the last step (of course :-)

Rochus Schmid (30. 11. 98)