Hi, Mohammed,
The answer to your first question depends on the platform, but for most
platforms the default is "pgas_sockets". You can launch the resulting
executables either with the manager/launcher combination, or with
mpirun/mpiexec from some versions of MPI (MPICH2 should work, but I
think there were some issues with the more recent versions -- 1.3 -- of
OpenMPI).
The "standalone" transport is designed to work within a single node only.
You simply launch the (single) executable (e.g., "./a.out"), and it forks
off one copy for each place. The number of places is controlled by the
X10RT_STANDALONE_NUMPLACES environment variable, and defaults to 1.
Places communicate using shared memory. There is no need to use an
external launcher for these executables, but you also cannot run them
across a cluster.
Finally, if you build with the "mpi" transport, what you get is a normal
MPI executable, which you launch with the appropriate mpirun/mpiexec
(just make sure it's from the same version of MPI that you built with).
Using "-x10rt mpi" will pick up mpicxx from the PATH, so make sure the
one you want to use is first in the PATH (verify using "which mpicxx").
Hope this helps,
Igor
mohammed elsaeedy <mohammed.elsae...@kaust.edu.sa> wrote on 07/21/2010
10:10:07 AM:
> Dear list,
>
>
> The issue for the 3rd question, was solved by following tuto:
Augmenting a
> pre-built X10 Release with the MPI version of
> X10RT<http://x10.codehaus.org/X10RT+Implementations>
> but then my first two question are still unresolved for me.
> another thing is that after I compiler with MPI, and I want to run it
over a
> cluster, how can I do that please?
>
> Thank you all
> Mohammed @ KAUST
>
> On Wed, Jul 21, 2010 at 9:08 AM, mohammed elsaeedy <
> mohammed.elsae...@kaust.edu.sa> wrote:
>
> > Dear list,
> >
> >
> > I have developed several applications that work very well locally
(one
> > place or several places), and I want to deploy them over a cluster, so
I
> > have several questions
> > regarding that:
> >
> > 1) When I compile with "x10c++ -o Hello Hello.x10" which x10rt
> > implementation is used? is it the sockets or the standalone? and whats
the
> > difference between both?
> >
> > 2) When I compile with "x10c++ -x10rt standalone -o Hello Hello.x10"
> > and when I get to run with (launcher -t n) it executes, but always
as
> > only one place as it gives me the following message:
> > "X10RT_STANDALONE_NUMPLACES not set. Assuming 1 place" . How can i
set it?
> >
> > 3) When I compile with "x10c++ -x10rt MPI -o Hello Hello.x10", at
first
> > it asked for the file named "x10rt_MPI.properties" which was not there
at
> > first, so I created one and
> > filled it with the following:
> >
> > CXX=mpicxx
> > CXXFLAGS=
> > LDFLAGS=
> > LDLIBS=-lx10rt_mpi
> >
> > when I compile now, it tells me that it cant find "-lx10rt_mpi"
> >
> > I have OpenMPI installed over my Ubuntu, and I want to try my
> program over MPI, I'm using X10 2.0.4
> >
> > Is there anywhere where we can understand these x10rt property file
> compiler flags or commands?
--
Igor Peshansky (note the spelling change!)
IBM T.J. Watson Research Center
X10: Parallel Productivity and Performance (http://x10-lang.org/)
XJ: No More Pain for XML's Gain (http://www.research.ibm.com/xj/)
"I hear and I forget. I see and I remember. I do and I understand" --
Confucius
------------------------------------------------------------------------------
This SF.net email is sponsored by Sprint
What will you do first with EVO, the first 4G phone?
Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first
_______________________________________________
X10-users mailing list
X10-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/x10-users
