source: CIVL/examples/mpi/collective/scatter.c

/**
 * This example demonstrates the usage of MPI collective operations,
 * which should be called in the same orders for all MPI processes.
 * This example has an error when there are more than two MPI processes.
 */
#include<mpi.h>
#include<stdlib.h>

int main(int argc, char * argv[])
{
    int rank;
    int procs;
    int* values;
    
    MPI_Init(&argc,&argv);
    MPI_Comm_size(MPI_COMM_WORLD, &procs);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    
    if (rank == 0 || rank == 2) {
        values = (int*)malloc(sizeof(int)*procs);
        for(int i = 0; i < procs; i++){
            values[i] = i;
        }
    }else{
        values = (int*)malloc(sizeof(int));
    }
    
#if defined TYPE
    if (rank != 2)
        MPI_Scatter(values, 1, MPI_INT, values, 1, MPI_INT, 0, MPI_COMM_WORLD);
    else
        MPI_Scatter(values, 1, MPI_FLOAT, values, 1, MPI_FLOAT, 0, MPI_COMM_WORLD);
#elif defined ROOT
    if (rank != 2)
        MPI_Scatter(values, 1, MPI_INT, values, 1, MPI_INT, 0, MPI_COMM_WORLD);
    else
        MPI_Scatter(values, 1, MPI_INT, values, 1, MPI_INT, 2, MPI_COMM_WORLD);
#else
    if (rank != 2)
        MPI_Scatter(values, 1, MPI_INT, values, 1, MPI_INT, 0, MPI_COMM_WORLD);
#endif
    
    free(values);
    MPI_Finalize();
    return 0; 
}