OpenMPI

OpenMPI is a open-source implementation of the Message Passing Interface (MPI) standard, which designed for high-performance parallel computing. Supporting distributed-memory systems, clusters, and supercomputers and enabling efficient communication between processes in parallel applications.

Key Features

• Efficiently handles small to large-scale clusters
• Optimized for low-latency, high-throughput communication
• Supports multiple programming languages: C, C++, Fortran

Use Cases

• Scientific Computing
• Simulations in physics, chemistry, and biology
• Large-scale numerical computations
• Big Data and Machine Learning
• Distributed training of models
• Parallel processing of large datasets

MPI Program Source Code (Example)

Example source code path → /home/$USER/job_template/C/mpi_pi.c

// This program is to calculate PI using MPI
// The algorithm is based on integral representation of PI. If f(x)=4/(1+x^2), then PI is the integral of f(x) from 0 to 1

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

#define N 1E9
#define d 1E-9
#define d2 1E-18

int main (int argc, char* argv[])
{
  int rank, size, error, i;

  double pi=0.0, result=0.0, sum=0.0, begin=0.0, end=0.0, x2;

  error=MPI_Init (&argc, &argv);

  //Get process ID
  MPI_Comm_rank (MPI_COMM_WORLD, &rank);

  //Get processes Number
  MPI_Comm_size (MPI_COMM_WORLD, &size);

  //Synchronize all processes and get the begin time
  MPI_Barrier(MPI_COMM_WORLD);
  begin = MPI_Wtime();

  //Each process will calculate a part of the sum
  for (i=rank; i<N; i+=size)
  {
    x2=d2*i*i;
    result+=1.0/(1.0+x2);
  }

  //Sum up all results
  MPI_Reduce(&result, &sum, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);

//Synchronize all processes and get the end time
  MPI_Barrier(MPI_COMM_WORLD);
  end = MPI_Wtime();

  //calculate and print PI
  if (rank==0)
  {
    pi=4*d*sum;
    printf("np=%2d;    Time=%fs;    PI=%lf\n", size, end-begin, pi);
  }

  error=MPI_Finalize();
  return 0;
}

For more information about OpenMPI, please refer to OpenMPI Official Site