[Beowulf] Cluster OpenMP
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Lombard, David N david.n.lombard at intel.comTue May 16 15:21:13 PDT 2006
- Previous message: [Beowulf] coprocessor to do "physics calculations"
- Next message: [Beowulf] substantial RX packet drops during Pallas over e1000 (Rocks 4.1)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
From: Douglas Eadline on Tuesday, May 16, 2006 1:51 PM > > <http://softwareforums.intel.com/ids/board/message?board.id=11&message.i d=3793> > > > > Maybe it will be interesting for many people here. > > > > In short, recompile your multithreaded OpenMP program and get a cluster > > version. > > Were it that simple and it all depends. With parallel computing > and clusters in particular, deciding what *should* be parallel > for a given hardware environment is difficult. Just because > something is concurrent does not mean it should be executed in parallel. Lest anyone think the whitepaper suggests performance is that simple: "Obviously, not all applications are suitable for Cluster OpenMP. An application with fine-grained synchronization is forced to spend a large percentage of its time fetching data from remote nodes and paying thousands of times more for memory accesses than it would on an SMP. An important consideration for a potential user of Cluster OpenMP is whether the prospective application is suitable for Cluster OpenMP." The UserGuide.pdf linked on the web page above goes into the details. -- David N. Lombard My statements represent my opinions, not those of Intel Corporation
- Previous message: [Beowulf] coprocessor to do "physics calculations"
- Next message: [Beowulf] substantial RX packet drops during Pallas over e1000 (Rocks 4.1)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
