#C++

Copy this into the file hello_processes.cpp

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

int main(int argc, char **argv)
{
    // Initialise MPI
    MPI_Init(&argc, &argv);

    // Get the number of cores in the MPI cluster
    int nprocs = MPI::COMM_WORLD.Get_size();

    // Get the ID number of this core in the MPI cluster
    int rank = MPI::COMM_WORLD.Get_rank();

    std::cout << "I am process " << rank
              << ". The number of processes is " << nprocs << ".\n";

    // Shut down MPI
    MPI::Finalize();

    return 0;
}

This example calls two MPI functions;

• MPI::COMM_WORLD.Get_size() : Returns the number of processes in the MPI processteam
• MPI::COMM_WORLD.Get_rank() : Returns the rank of the calling process.

Note that using these functions requires you to include the mpi.h header file. For some MPI implementations, you may also need to include an extra C++ MPI header file, called something like mpic++.h. See the documentation of your MPI library for more details. Also, the C++ wrappers were only standardised in MPI 2. Most MPI libraries now support MPI 2, but older MPI libraries may be missing the C++ wrappers.

Note that all MPI functions are in the MPI namespace. Also note that the functions are called from the object MPI::COMM_WORLD. This object represents the global communicator that is used to coordinate communication between all processes in the MPI process team. It is possible to create and use different MPI communicators in a program, and calling the same MPI functions using the object representing the new communicator. Creating and managing different communicators is beyond the scope of this simple introduction, so all examples in this course will always use MPI::COMM_WORLD.

You can compile this program as you did for the hello_mpi.cpp in the last section, e.g.

mpicxx hello_processes.cpp -o hello_processes

This will produce the executable, hello_processes.

Return to the previous page.