<html><body>
<DIV>-------------- Original message -------------- <BR>From: Geoff Jacobs <gdjacobs@gmail.com> <BR></DIV>
<DIV>> <BR>> But isn't CAF (and UPC, and Titanium) implicitly message passing for a <BR>> Beowulf anyway? It's attractive because it simplifies the process and <BR>> might be able to optimize communication, but underneath everything it's <BR>> still message passing. <BR>> <BR></DIV>
<DIV>Most of what you say here is true ...</DIV>
<DIV> </DIV>
<DIV>It is low-level message passing between nodes, and can be either within ... depending</DIV>
<DIV>what optimizations the compiler does. Still, the code generated is one layer closer to</DIV>
<DIV>the network adapter hardware and has a small potential performance advantage because</DIV>
<DIV>of this (although MPI can be used as a conduit). </DIV>
<DIV> </DIV>
<DIV>PGAS languages push the problem of managing latency off onto the compiler </DIV>
<DIV>while offering a more implicit, language integrated approach to dealing with remote</DIV>
<DIV>references. The []s are light-weight symbols that remind the programmer of the</DIV>
<DIV>overhead implicit in make remote references, but the work of actual making them</DIV>
<DIV>effecient is left up to the compiler.</DIV>
<DIV> </DIV>
<DIV>rbw<BR>-- <BR><BR>"Making predictions is hard, especially about the future." <BR><BR>Niels Bohr <BR><BR>-- <BR><BR>Richard Walsh <BR>Thrashing River Consulting-- <BR>5605 Alameda St. <BR>Shoreview, MN 55126 <BR><BR>Phone #: 612-382-4620</DIV></body></html>