User Tools

Site Tools


MPI Scatter

Using MPI_Scatter to distribute parts of an array from the master process to all processes in an MPI communicator.
MPI_Scatter is used to divide an array into equal segments and distribute each segment to a different process.

#include <mpi.h>
#include <stdio.h>
int main(int argc, char** argv) {
    // Initialize the MPI environment
    MPI_Init(&argc, &argv);
    // Get the rank of the process
    int world_rank;
    MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
    // Get the total number of processes
    int world_size;
    MPI_Comm_size(MPI_COMM_WORLD, &world_size);
    // The master process (rank 0) prepares the data
    const int elements_per_proc = 3;
    int *send_data = NULL;
    if (world_rank == 0) {
        const int total_elements = elements_per_proc * world_size;
        send_data = new int[total_elements];
        // Initialize the array with values
        for (int i = 0; i < total_elements; i++) {
            send_data[i] = i + 1;
    // Each process will receive a portion of the array from the master process
    int recv_data[elements_per_proc];
    // Scatter the data from the master process to all processes
    MPI_Scatter(send_data, elements_per_proc, MPI_INT, recv_data, elements_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
    // Each process prints its received portion
    printf("Process %d received elements: ", world_rank);
    for (int i = 0; i < elements_per_proc; i++) {
        printf("%d ", recv_data[i]);
    // Clean up
    if (world_rank == 0) {
        delete[] send_data;
    // Finalize the MPI environment

Initialization: The MPI environment is initialized with MPI_Init.

Data Preparation: The master process (rank 0) prepares an array (send_data) with data to be distributed.

Scatter Operation: MPI_Scatter is called by all processes. It divides the send_data array into segments of elements_per_proc and distributes them to each process. Each process receives its segment in the recv_data array. The parameters of MPI_Scatter include the send buffer (only significant at the master process), the number of elements sent to each process, the datatype of the elements, the receive buffer, the number of elements received, the datatype of the receive buffer, the rank of the sending process, and the communicator.

Output: Each process prints the elements it received.

Cleanup: The master process frees the allocated memory.

Finalization: The MPI environment is finalized with MPI_Finalize.

The number of total elements in the send_data array should be a multiple of the number of processes to ensure even distribution.
c/mpi/scatter.txt · Last modified: 2024/01/17 10:36 by odefta