[Pw_forum] Running Example 01 (Parallel Running on a single PC with 3 cores, AMD, LAM-MPI)

[O. Baris Malcioglu]

Dear Masoud

First of all, let me kindly mention that opening several threads for
the same problem is not very nice, these messages are sent to many
people that follow the topics thread by thread, and not everyone is
interested in this particular problem. It is also not very nice for
people searching the forum.

Now, regarding your problem: I would recommend you to read some
documentation regarding how to set up and use parallel execution
environments before setting up Q-E. It seems to me that the problem
you are experiencing is due to lack of proper communication between
the different processes. The observation you made that mpi wrappers
for compilers are working does not imply that the parallel environment
is working. Did you try a simple test program like :

#include <stdio.h>
#include "mpi.h"
int main(int argc, char **argv) {
   int rank;
   int namelen;
   MPI_Init(&argc, &argv);
   MPI_Comm_rank(MPI_COMM_WORLD, &rank);
   printf("Hello world (Rank: %d)\n", rank);
   fflush(stdout);
   MPI_Finalize();
   return 0;
}

My personal preference on the matter is to use openmpi, not lam, but
its your choice. Be careful about lamboot, and how you set the spawned
processes to communicate.

Best.
Baris



2010/3/1 Masoud Nahali <masoudnahali at live.com>:
> Dear QE's Users
>
> Sometimes the running of exapmle 01 stops !! and sometimes aborts because of
> MPI_INIT !!
>
> koa at linux-xd0f:~/QSPRESSO4.1.2/espresso-4.1.1-Example/examples/example01>
> ./run_example
> /home/koa/QSPRESSO4.1.2/espresso-4.1.1-Example/examples/example01 : starting
> This example shows how to use pw.x to calculate the total energy and
> the band structure of four simple systems: Si, Al, Cu, Ni.
> executables directory: /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin
> pseudo directory: /home/koa/QSPRESSO4.1.2/espresso-4.1.2/pseudo
> temporary directory: /home/koa/tmp
> checking that needed directories and files exist... done
> running pw.x as: mpirun -np 3
> /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin/pw.x -npool 1 -inp
> running bands.x as: mpirun -np 3
> /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin/bands.x -npool 1 -inp
> cleaning /home/koa/tmp... done
> running the scf calculation for
> Si...-----------------------------------------------------------------------------
> The selected RPI failed to initialize during MPI_INIT. This is a
> fatal error; I must abort.
> This occurred on host linux-xd0f (n0).
> The PID of failed process was 9075 (MPI_COMM_WORLD rank: 0)
>
> or:
>
>
> oa at linux-xd0f:~/QSPRESSO4.1.2/espresso-4.1.1-Example/examples/example01>
> ./run_example
> /home/koa/QSPRESSO4.1.2/espresso-4.1.1-Example/examples/example01 : starting
> This example shows how to use pw.x to calculate the total energy and
> the band structure of four simple systems: Si, Al, Cu, Ni.
> executables directory: /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin
> pseudo directory: /home/koa/QSPRESSO4.1.2/espresso-4.1.2/pseudo
> temporary directory: /home/koa/tmp
> checking that needed directories and files exist... done
> running pw.x as: mpirun -np 3
> /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin/pw.x -npool 1 -inp
> running bands.x as: mpirun -np 3
> /home/koa/QSPRESSO4.1.2/espresso-4.1.2/bin/bands.x -npool 1 -inp
> cleaning /home/koa/tmp... done
> running the scf calculation for Si...    ---> stops at this line
>
>
>    using the -inp that is suggested by Paolo Giannozzi did not solve my
> problem.
> I have tested my mpif77 and mpicc and I think that they works well , but I
> feel that the problem is related to the MPI !
> but I can not find it. My QE has been compiled well by mpif77 and all
> binaries created well. I am not sure that LAM-MPI
> recognizes the 3 cores of a single PC or not ? I am comfused ! another
> amazing paradox is that once I ran this example.
>
> ________________________________
> Hotmail: Powerful Free email with security by Microsoft. Get it now.
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
>