A hybrid job uses both OpenMP and MPI for parallelization. This requires the number of tasks and CPUs per task to be set up correctly. Here is an example of such a job. We will use Intel MPI in this example, but this also works using OpenMPI.

Code

#include <mpi.h>
#include <stdio.h>
 
int main(int argc, char** argv) {
    // Initialize the MPI environment
    MPI_Init(NULL, NULL);
 
    // Get the number of processes
    int world_size;
    MPI_Comm_size(MPI_COMM_WORLD, &world_size);
 
    // Get the rank of the process
    int world_rank;
    MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
 
    // Get the name of the processor
    char processor_name[MPI_MAX_PROCESSOR_NAME];
    int name_len;
    MPI_Get_processor_name(processor_name, &name_len);
 
    int nthreads, tid;
 
    /* Fork a team of threads giving them their own copies of variables */
    #pragma omp parallel private(nthreads, tid)
    {
   
    /* Obtain thread number */
    tid = omp_get_thread_num();
    printf("Hello World from thread = %d, processor %s, rank %d out of %d processors\n", tid, processor_name, world_rank, world_size);
   
    /* Only master thread does this */
    if (tid == 0)
      {
      nthreads = omp_get_num_threads();
      printf("Number of threads = %d\n", nthreads);
      }
   
    }  /* All threads join master thread and disband */
 
 
    // Finalize the MPI environment.
    MPI_Finalize();
}

Compilation

This can be compiles with both the Intel or the GNU compiler.

module load intel
module load impi
mpiicc -qopenmp  -Wl,-rpath,$LD_RUN_PATH -o hybrid_hello_world.bin hybrid_hello_world.c

module load gcc
module load impi
mpigcc -fopenmp -Wl,-rpath,$LD_RUN_PATH -o hybrid_hello_world.bin hybrid_hello_world.c

Batch Job

#!/bin/bash
#SBATCH --time=00:10:00
#SBATCH --nodes=2
#SBATCH --tasks-per-node=4
#SBATCH --cpus-per-task=10
#SBATCH --partition=standard96:test
 
module load impi
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
srun ./hybrid_hello_world.bin

Related articles

• Seite:
  Slurm usage
• Seite:
  Workflow CPU CLX
• Seite:
  Advanced options
• Seite:
  Hybrid Jobs
• Seite:
  INTEL-MPI version 19 slower than INTEL-MPI version 18