This web mail archive is frozen.
This page is part of a frozen web archive of this mailing list.
You can still navigate around this archive, but know that no new mails
have been added to it since July of 2016.
Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.
On Jun 29, 2009, at 9:48 PM, Ashika Umanga Umagiliya wrote:
> I am parallelizing tomography algorithm which takes about 5 hours
> using single processor.
> I want to gain full performance and should reduce the
> computational time as short as possible.
> I was wondering the SSh/RSH launcher could be a performance
> issue?(I am just guessing)
Unlikely. The time to launch an ssh/rsh process is negligible
compared to a 5 hour run.
> What kind of software/middleware I should use to solve the network/
> communication overhead.
I'm not sure what you mean by this question...?
In your environment, Open MPI is using TCP sockets to communicate
between MPI processes on different hosts and likely using shared
memory to communicate between MPI processes on the same host. But in
your MPI application, you just use MPI_SEND / MPI_RECV (etc.) -- Open
MPI automatically chooses the "best" transport underneath and gets
your message to the other side as fast as possible.
You might want to take a short MPI tutorial to get familiar with MPI
concepts. There's an excellent free introduction to MPI tutorial at
Register for a (free) account and then you'll see the course listing.
This tutorial will introduce you to MPI concepts such as sending,
receiving, MPI collective operations, etc.