<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 5.5.2448.0">
<TITLE>RE: Parallelization check list</TITLE>
</HEAD>
<BODY>

<P><FONT SIZE=2>Sure. One of the things to look for, definitely, are nested loops.  You can increase "big oh" efficienty by factors of ten when you parallelize these types of loops.  Not sure about any actual benchmarks on the approach, however.</FONT></P>

<P><FONT SIZE=2>Richard Schilling</FONT>
<BR><FONT SIZE=2>Web Integration Programmer</FONT>
<BR><FONT SIZE=2>Affiliated Health Services</FONT>
</P>

<P><FONT SIZE=2>-----Original Message-----</FONT>
<BR><FONT SIZE=2>From: Gallip William C PHDN [<A HREF="mailto:gallipwc@nswcphdn.navy.mil">mailto:gallipwc@nswcphdn.navy.mil</A>]</FONT>
<BR><FONT SIZE=2>Sent: Thursday, June 29, 2000 10:45 AM</FONT>
<BR><FONT SIZE=2>To: beowulf@beowulf.org</FONT>
<BR><FONT SIZE=2>Subject: Parallelization check list</FONT>
</P>
<BR>

<P><FONT SIZE=2>We are starting to put together a check-list of things to consider when</FONT>
<BR><FONT SIZE=2>taking a serial code and determining its parallelizability, either for a</FONT>
<BR><FONT SIZE=2>distributed memory and/or SMP environment.  Has anyone already gone through</FONT>
<BR><FONT SIZE=2>a similar exercise and if so could you share the results?  Any help would be</FONT>
<BR><FONT SIZE=2>greatly appreciated.</FONT>
</P>

<P><FONT SIZE=2>Bill Gallip</FONT>
<BR><FONT SIZE=2>Fleet Advanced Supercomputing Technology Center</FONT>
<BR><FONT SIZE=2>NAVSEA Dam Neck</FONT>
<BR><FONT SIZE=2>Virginia Beach, VA</FONT>
</P>

<P><FONT SIZE=2>_______________________________________________</FONT>
<BR><FONT SIZE=2>Beowulf mailing list</FONT>
<BR><FONT SIZE=2>Beowulf@beowulf.org</FONT>
<BR><FONT SIZE=2><A HREF="http://www.beowulf.org/mailman/listinfo/beowulf" TARGET="_blank">http://www.beowulf.org/mailman/listinfo/beowulf</A></FONT>
</P>

</BODY>
</HTML>