Test MCNP with Parallel universe

[makeclean]

Test simple hello world program in the Parallel universe, then extend to full MCNP5 calculation.

[makeclean]

#include <iostream>
#include "mpi.h"

void print_id(int id_number, int num_proc);

int main(int argc, char* argv[])
  {

  int id; // cpu id
  int p; // number of procs

  MPI::Init (argc,argv);

  id = MPI::COMM_WORLD.Get_rank( );
  p = MPI::COMM_WORLD.Get_size( );

  if ( id == 0 )
    {
      std::cout << " I am the master " << id << std::endl;
    }

  MPI::COMM_WORLD.Barrier( );

  print_id(id,p);

  MPI::COMM_WORLD.Barrier( );
  MPI::Finalize( );

  return 0;
}


void print_id(int id,int p)  
{
  std::cout << "I am processor " << id << " of " << p << std::endl;
  return;
}

[makeclean]

Simple hello world mpi program, compile as

mpicxx mpitest.cpp -o mpitest

[gonuke]

You should just add these as files in the repo, with a makefile. Rather than as comments here.

[makeclean]

This is done, 36d54ed

[gonuke]

A hello world test is done, right. Do you want to make another issue (or a few more) towards the goal of testing MCNP?

[makeclean]

Currently parallel universe is non-responsive awaiting assistance from CHTC, the condor parallel examples in the manual work as expected, however MPI parallel programs do not.

[makeclean]

Parallel universe still unresponsive, however can now run MPI programs on the same node as long we do not exceed 32 core, see commit 60fbbf7 or #22