Jack,
I'll suggest one thing. Jeff Squyres writes an MPI column for
ClusterWorld magazine. The column is very good - easy and fun
to read and very informative. If you're interested in clusters and
MPI, I would suggest subscribing to the magazine
(www.clusterworld.com). The magazine is working on a project
to put the back issues on a CD.
Jeff
>>Yes, you can use MPI_ANY_SOURCE as the peer argument. I'm not a huge
>>fan of MPI_ANY_SOURCE, but it seems suitable for this situation. You
>>can find out who sent to you by examining the MPI_Status argument -- it
>>will be in the field named MPI_SOURCE.
>>
>>
>the line I tried was:
> n = MPI_Recv(&f, 1, MPI_DOUBLE, MPI_ANY_SOURCE, 1, MPI_COMM_WORLD,
>MPI_SOURCE);
>However, it won't compile. Perhaps I don't understand the MPI_Status part? I
>did a man on MPI_Recv, and it sais MPI_Status went at the end...
>
>
>
>>If you're an MPI newbie, you might want to get a good book on MPI, or
>>take one of the available on-line tutorials available. There's some
>>excellent ones out there, such as the one at NCSA:
>>
>> http://foxtrot.ncsa.uiuc.edu:8900/public/MPI/
>>
>>
>Actually I'm working my way through it, I guess I haven't foung the right part
>yet.
>
>Thanks,
>
>-Jack C
>_______________________________________________
>This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
>
>
|
