How to define a local array in mpi?

Question

When I define an array in an mpi program, does the all processors share the same address, that is the same array(pointer) or they point to different arrays(pointers)? For example, int the following code

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

int main(int argc, char** argv) {
    int num_procs, rank;

    // Initialize MPI
    int argcc=8;
    char **argvc;
    //print argc and argv
    std::cout << "argc: " << argc << std::endl;
    for (int i = 0; i < argc; ++i) {
        std::cout << "argv[" << i << "]: " << argv[i] << std::endl;
    }
    MPI_Init(&argcc, &argvc);

    // Get the number of processors
    MPI_Comm_size(MPI_COMM_WORLD, &num_procs);

    // Get the rank of the current process
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);

    // Print the number of processors and the rank of the current process
    std::cout << "Number of processors: " << num_procs << std::endl;
    std::cout << "Rank: " << rank << std::endl;
    int count;
        
    if (rank == 0) {
        std::vector<int> numbers;
        std::ifstream inputFile("input.txt");
        int number;
        while (inputFile >> number) {
            numbers.push_back(number);
        }
        inputFile.close();
        // Print the numbers
        for (int i = 0; i < numbers.size(); ++i) {
            std::cout << "Number " << i << ": " << numbers[i] << std::endl;
        }
        // Send the one part of numbers to the other processes
         count = numbers.size() / (num_procs - 1);
        int offset = 0;
        for (int dest = 1; dest < num_procs; ++dest) {
            MPI_Send(&numbers[offset], count, MPI_INT, dest, 0, MPI_COMM_WORLD);
            offset += count;
        }
    }
    //create a vector to store the numbers
    std::vector<int> numbersx;

        // Receive the numbers from the root process
        MPI_Status status;
        MPI_Probe(0, 0, MPI_COMM_WORLD, &status);
        MPI_Get_count(&status, MPI_INT, &count);
        std::vector<int> numbers(count);
        MPI_Recv(&numbersx[0], count, MPI_INT, 0, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
        // Print the numbers
        for (int i = 0; i < numbers.size(); ++i) {
            std::cout << "Number " << i << ": " << numbers[i] << std::endl;
        }

   
    // Finalize MPI
    MPI_Finalize();

    return 0;
}

is the numbersx array shared or does each processor has its own separate array? Or what should I do for each processor to have its own local array?

Answer 1

MPI is process-based. So if your code says vector<int> x then every process creates its own vector, at an address that is totally dependent on the process. You can also look at it as: everything you allocate is local; there are no global data structures.