<!DOCTYPE html>
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
<p>So let's talk about queue etiquette. While it may be easier for
you to build a job like this...</p>
<font face="monospace">#PBS -N JACOBI_TEST <br>
#PBS -m abe<br>
#PBS -j oe<br>
#PBS -l nodes=1:hammer:ppn=42,walltime=10:00:00<br>
#<br>
cd $HOME/jacobirepo<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi<br>
jacobi</font>
<p>to get multiple timings, this could potentially block the queue
for, in this case, up to 10 hours. If you know that that your
job will run for a while, it is generally much more considerate
(and often required at supercomputing installations) to build each
job to handle one calculation with a maximum time limit that is
not egregiously long. For example I would rewrite the above as</p>
<p><font face="monospace">#PBS -N JACOBI_TEST <br>
#PBS -m abe<br>
#PBS -j oe<br>
#PBS -l nodes=1:hammer:ppn=42,walltime=30:00<br>
#<br>
cd $HOME/jacobirepo<br>
jacobi</font></p>
<p>And then submit ten of them. Each job should return a log file
with the output. If there are multiple users in the queue the
"fair share" scheduler will then allow the queue to be shared
equitably.</p>
<p>Once your tests are complete you can then "cat" the output files
and "grep" for the times.<br>
</p>
<p><br>
</p>
<div class="moz-signature">-- <br>
<b>Andrew J. Pounds, Ph.D.</b><br>
<i>Professor of Chemistry and Computer Science</i><br>
<i>Director of the Computational Science Program</i><br>
<i>Mercer University, Macon, GA, 31207 (478) 301-5627 </i></div>
</body>
</html>