LAM supports communication between processes using shared memory. sysv rpi
(where synchronization is achieved using system V semaphores) and usysv
rpi (where synchronization is achieved using spin locks) are provided for
this purpose.
Using '-ssi rpi sysv' switch to the mpirun, you would be
able to take advantage of SMPs. More information on the ssi parameters is
available in LAM FAQ and User Guide.
Just a quick note: In a boot schema you need to say cpu=2 (Not np=2).
More information is available in LAM FAQ (Q:How do I lamboot multi-processor
machines?)
FYI: In LAM, some collectives (Bcast, reduce, allreduce, barrier) are also
implemented according to the MagPIe's algorithms. These algorithms also
take advantage of the SMP machines. Note that, with the 'sysv' rpi, you
can get more speed while working with above collectives.
Hope this helps.
Amey S. Dharurkar
----------------------------------------------------------
LAM/MPI Developer
Graduate Student, Indiana University
Ph. O:(812)855-3609, H:(812)331-8203
On Tue, 14 Oct 2003, Carlos Lopez Nataren wrote:
> Hello, I haven't been able to find the answer to the next question, I greatly
> appreciate anyone giving me some advice.
>
> Does anyone know if LAM can take advantage of SMP machines??? For example,
> suppose you have a Beowulf cluster with SMP nodes, you boot with a hipotetical
> lamhosts file like this:
>
> master np=2
> node1 np=2
> node2 np=2
> node3 np=3
>
> and the application makes a lot of communication in pairs, I mean, if I use
> this boot scheme in LAM and let's say that in node1 there are a couple of
> processes and they have a lot of communication between them, LAM makes use of
> the shared memory to perform this communication of it uses the network interface?
> or does it have to be instructed to do this???
>
> Thanks a lot in advance.
> --
> Carlos Lopez Nataren <natorro_at_[hidden]>
> Instituto de Fisica, UNAM
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
