Thread: Basic MPI

I wrote a program which sends a starting and ending range to other processes and the processes calculate the prime numbers in that range and return the count of prime numbers to the head process, process 0. But this is not working properly at the moment. I realize I still have to split up the range based on how many processes I have...I still have not figured out how I want to set that up. If you see anything wrong with what I've got at the moment, I'd appreciate your opinion. Thanks!

Code:

#include <stdio.h>
#include <stdlib.h>
#include <mpi.h>

int isPrime(int num);

int main(int argc, char **argv){

  int myID;
  int numProcs;
  int primeCount = 0;
  
 // printf("The starting and ending values are: %d %d\n", starting, ending);
  
  MPI_Init(&argc, &argv);
  MPI_Comm_size(MPI_COMM_WORLD, &myID);
  MPI_Comm_rank(MPI_COMM_WORLD, &numProcs);
  
  if(myID == 0) { // Process 0 (head process)
    int starting = 10;
    int ending = 100;
    MPI_Status status;
    for(int i = 1; i < numProcs; i++){
        MPI_Send(&starting, 1, MPI_INT, i, 1, MPI_COMM_WORLD);
        MPI_Send(&ending, 1, MPI_INT, i, 2, MPI_COMM_WORLD);
    }
    int total = 0;
    for (int i = 1; i < numProcs; i++) {
       int primeCount;
       MPI_Recv(&primeCount, 1, MPI_INT, i, 1, MPI_COMM_WORLD, &status); // blocking
       total += primeCount;
    }
    
    } else { // All other processes
      MPI_Status status;
      int starting, ending;
      for(int i = 1; i < numProcs; i++){
        MPI_Recv(&starting, 1, MPI_INT, i, 1, MPI_COMM_WORLD, &status); // blocking
        MPI_Recv(&ending, 1, MPI_INT, i, 2, MPI_COMM_WORLD, &status); // blocking
      }
      for (int i = starting; i < ending; i++) {
        if(isPrime(i)){
            primeCount++;
        }
       MPI_Send(&primeCount, 1, MPI_INT, 0, 1, MPI_COMM_WORLD);
    }

    printf("Prime count: %d\n", primeCount);
  }
    MPI_Finalize();
}



    int isPrime(int num){
        for (int i = 2; i < num; i++){
            if (num % i == 0 && i != num) return 0;
        }
        return 1;
    }

My output right now is:
Prime count: 1
Prime count: 1

and then it freezes

Code:

    int isPrime(int num){
        for (int i = 2; i < num; i++){
            if (num % i == 0 && i != num) return 0;
        }
        return 1;
    }

Seems wasteful to divide each number by 4,6,8 ...

That makes sense Salem, thank you, but now it's going through and giving me zero as the result. Here is my revised code:

Code:

int main(int argc, char **argv){
  int myID;
  int numProcs;
  
  MPI_Init(&argc, &argv);
  MPI_Comm_size(MPI_COMM_WORLD, &numProcs);
  MPI_Comm_rank(MPI_COMM_WORLD, &myID);
 
  if(myID == 0) { // Process 0 (head process)
    int starting = 10;
 int ending = 100;
 printf("The starting and ending values are: %d %d\n", starting, ending);
 MPI_Status status;
 for(int i = 1; i < numProcs; i++){
  MPI_Send(&starting, 1, MPI_INT, i, 1, MPI_COMM_WORLD);
  MPI_Send(&ending, 1, MPI_INT, i, 2, MPI_COMM_WORLD);
 }
 int totalPrime = 0;
 int primeCount = 0;
    for (int i = 1; i < numProcs; i++) {
       MPI_Recv(&primeCount, 1, MPI_INT, i, 1, MPI_COMM_WORLD, &status);
       totalPrime += primeCount;
    }
 printf("Prime count: %d\n", totalPrime);
 } else { // All other processes
   MPI_Status status;
   int starting, ending;
   int localPrimeCount = 0;
   MPI_Recv(&starting, 1, MPI_INT, 0, 1, MPI_COMM_WORLD, &status);
   MPI_Recv(&ending, 1, MPI_INT, 0, 2, MPI_COMM_WORLD, &status); 
   for (int i = starting; i < ending; i++) {
  if(isPrime(i)){
   localPrimeCount++;
  }
    MPI_Send(&localPrimeCount, 1, MPI_INT, 0, 1, MPI_COMM_WORLD);
 }
  }
    MPI_Finalize();
}

For a start you could check the return values of MPI_Send and MPI_Recv:

Code:

if (MPI_Send(...) != MPI_SUCCESS)
    // we have a problem

You could also check the status structure and the actual values you send/recieve.

In general, you should learn to debug parallel processes.

Bye, Andreas

Originally Posted by AndiPersti

For a start you could check the return values of MPI_Send and MPI_Recv:

Code:

if (MPI_Send(...) != MPI_SUCCESS)
    // we have a problem

You could also check the status structure and the actual values you send/recieve.

In general, you should learn to debug parallel processes.

Bye, Andreas

When I get a moment I will do that, thank you for the recommendation

Never mind, solved it