[Beowulf] Re: MPI Implementations for SMP use
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.
Isaac Dooley idooley2 at uiuc.eduSun Dec 19 15:37:20 PST 2004
- Previous message: [Beowulf] Re: MPI Implementations for SMP use
- Next message: [Beowulf] use for a cluster -- wall of monitors
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
>>You would think that, but the nice thing about pure MPI is that >>locality is perfect. So in most cases, a pure MPI code beats a hybrid >>code. That's good, because hybrid programming is more complicated than >>straight MPI or straight OpenMP. >> >> Not necessarily. Charm++ uses an abstraction that does not concern the programmer with the location/node of a given object. Hence it is not more complicated than straight MPI. Its AMPI(an adaptive MPI implementation) works the same way, and both are free. (http://charm.cs.uiuc.edu). There are some papers at the site on this topic. Also locality may be an issue for some, but as we all know, applications have different bottlenecks and issues. Locality may be usefull for small programs which fit in cache. However there are many real applications which have network bottlenecks, IO bottlenecks or other problems which make the CPU and/or memory locality not the bottleneck. Maybe I'm just misunderstanding your reference to locality though. cheers, Isaac
- Previous message: [Beowulf] Re: MPI Implementations for SMP use
- Next message: [Beowulf] use for a cluster -- wall of monitors
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
