<!DOCTYPE html>
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
<p>Guys -- I am not planning on us meeting tomorrow and there will
be no class next Tuesday due to BEAR DAY. I do expect, however,
to see you working on code and running tests in the batch queues.
Your proof of concept to demonstrate that you can achieve better
porformance when using OpenMP should not be too difficult to pull
together. I recommend picking a fairly big matrix dimension and
running tests over 1 - 20 compute nodes. Then do the pick some
number of thread, say 4, and run your OpenMP code over 1 - 20
compute nodes. Graph the megaflops vs number of nodes for both
tests and you should have a pretty clear winner. BUT -- don't
wait to do this. There will invariably be some clean-up and
debugging to do.</p>
<p>As Tyler found out, my research lab on the 4th floor of Godsey is
a great place to work on this stuff because you are using your
files just like you were in GSC 218. You should all have keycard
access.</p>
<p>E-mail me with questions. I hope to have the BabyBLAS graded
soon.</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>