使用 MPI 的数组总和

先决条件： MPI——分布式计算变得简单

消息传递接口 (MPI) 是一个例程库，可用于在 C 或 Fortran77 中创建并行程序。它允许用户通过创建并行进程并在这些进程之间交换信息来构建并行应用程序。
MPI 使用两个基本的通信例程：

• MPI_Send ，向另一个进程发送消息。
• MPI_Recv ，从另一个进程接收消息。

MPI_Send 和 MPI_Recv 的语法是：

int MPI_Send(void *data_to_send, 
             int send_count, 
             MPI_Datatype send_type, 
             int destination_ID, 
             int tag, 
             MPI_Comm comm); 

int MPI_Recv(void *received_data, 
             int receive_count, 
             MPI_Datatype receive_type,
             int sender_ID, 
             int tag, 
             MPI_Comm comm, 
             MPI_Status *status);

为了降低程序的时间复杂度，子数组的并行执行是通过并行进程运行来计算它们的部分和，然后最后，主进程（根进程）计算这些部分和的总和以返回总和数组。

例子：

Input : {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
Output : Sum of array is 55

Input : {1, 3, 5, 10, 12, 20, 4, 50, 100, 1000}
Output : Sum of array is 1205

注 –您必须在基于 Linux 的系统上安装 MPI，才能执行以下程序。有关详细信息，请参阅 MPI – 分布式计算变得容易

使用以下代码编译并运行程序：

mpicc program_name.c -o object_file
mpirun -np [number of processes] ./object_file

下面是上述主题的实现：

#include 
#include 
#include 
#include 
  
// size of array
#define n 10
  
int a[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 };
  
// Temporary array for slave process
int a2[1000];
  
int main(int argc, char* argv[])
{
  
    int pid, np,
        elements_per_process,
        n_elements_recieved;
    // np -> no. of processes
    // pid -> process id
  
    MPI_Status status;
  
    // Creation of parallel processes
    MPI_Init(&argc, &argv);
  
    // find out process ID,
    // and how many processes were started
    MPI_Comm_rank(MPI_COMM_WORLD, &pid);
    MPI_Comm_size(MPI_COMM_WORLD, &np);
  
    // master process
    if (pid == 0) {
        int index, i;
        elements_per_process = n / np;
  
        // check if more than 1 processes are run
        if (np > 1) {
            // distributes the portion of array
            // to child processes to calculate
            // their partial sums
            for (i = 1; i < np - 1; i++) {
                index = i * elements_per_process;
  
                MPI_Send(&elements_per_process,
                         1, MPI_INT, i, 0,
                         MPI_COMM_WORLD);
                MPI_Send(&a[index],
                         elements_per_process,
                         MPI_INT, i, 0,
                         MPI_COMM_WORLD);
            }
  
            // last process adds remaining elements
            index = i * elements_per_process;
            int elements_left = n - index;
  
            MPI_Send(&elements_left,
                     1, MPI_INT,
                     i, 0,
                     MPI_COMM_WORLD);
            MPI_Send(&a[index],
                     elements_left,
                     MPI_INT, i, 0,
                     MPI_COMM_WORLD);
        }
  
        // master process add its own sub array
        int sum = 0;
        for (i = 0; i < elements_per_process; i++)
            sum += a[i];
  
        // collects partial sums from other processes
        int tmp;
        for (i = 1; i < np; i++) {
            MPI_Recv(&tmp, 1, MPI_INT,
                     MPI_ANY_SOURCE, 0,
                     MPI_COMM_WORLD,
                     &status);
            int sender = status.MPI_SOURCE;
  
            sum += tmp;
        }
  
        // prints the final sum of array
        printf("Sum of array is : %d\n", sum);
    }
    // slave processes
    else {
        MPI_Recv(&n_elements_recieved,
                 1, MPI_INT, 0, 0,
                 MPI_COMM_WORLD,
                 &status);
  
        // stores the received array segment
        // in local array a2
        MPI_Recv(&a2, n_elements_recieved,
                 MPI_INT, 0, 0,
                 MPI_COMM_WORLD,
                 &status);
  
        // calculates its partial sum
        int partial_sum = 0;
        for (int i = 0; i < n_elements_recieved; i++)
            partial_sum += a2[i];
  
        // sends the partial sum to the root process
        MPI_Send(&partial_sum, 1, MPI_INT,
                 0, 0, MPI_COMM_WORLD);
    }
  
    // cleans up all MPI state before exit of process
    MPI_Finalize();
  
    return 0;
}

输出：

Sum of array is 55

以下是计算其部分和的过程的快照：