Hi,
thanks for advice. But I think processes are executed in round-robin, instead of one after one. And I believe that the time slot for each process is very small compared to the process lifetime (hours).It is true that computation cost will be increased by N times if there are N processes on it. But why the communication cost is also increased?
----- Original Message -----
From: dburbano_at_[hidden]
Date: Monday, March 29, 2004 4:12 pm
Subject: Re: LAM: is that possible to reduce the communication cost by assigning processes on the same node
> Hi,
>
> The idea is to parallelize the program in multiples processors no
> processes, when you assing multiples processes to one machine, your
> processes are executing one after other. The computation time
> should be N
> times where N is the number of processes. Now, if the information
> thatyou communicate depend of all processes, the communication
> finish until
> the last process finish.
>
> The idea is match the number of processes with the number of
> processors.
>
> > Hi,
> >
> > I assigned multiple processes on one machine instead of several
> > machines. In this way, I expect the communication cost will be
> reduced> compared with assigning them to several machines.
> However, the result is
> > weird. both computation cost and communication cost are increased
> > sharply, especially MPI_Reduce, MPI_Sendrecv. It seems that the
> > underlying implementation of lam_mpi doesn't favor multiple
> processes on
> > the same host.
> >
> > Your help will be greatly appreciated.
> >
> > thanks
> >
> > _______________________________________________
> > This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
>
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
