02_domain_decomposition.cpp

Once the World object has been defined, the next step is to determine how the world will be divided among different MPI processes. In OpenPFC, the calculation area is partitioned into smaller parts using the Decomposition class. This class utilizes the excellent functionality provided by HeFFTe. The partitioning is done in such a way that minimizes the surface area between the regions, which in turn minimizes the MPI communication required during FFT calculations. In the example, we can see how the domain decomposition can be done "manually". However, in practice, it is most effective to divide the calculation area into the same number of parts as the number of computing nodes allocated from the HPC cluster.

This example demonstrates how to use the World class to create a simulation world and after that, decompose world to smaller domains using Decomposition class.

// SPDX-FileCopyrightText: 2025 VTT Technical Research Centre of Finland Ltd
// SPDX-License-Identifier: AGPL-3.0-or-later
 
#include <iostream>
#include <mpi.h>
#include <openpfc/core/decomposition.hpp>
#include <openpfc/core/strong_types.hpp>
#include <openpfc/core/world.hpp>
#include <openpfc/factory/decomposition_factory.hpp>
 
using namespace std;
using namespace pfc;
 
int main(int argc, char *argv[]) {
  MPI_Init(&argc, &argv);
  // Domain decomposition can be done in a manual manner, just by giving the
  // size of the calculation domain, id number and total number of subdomains:
  int comm_rank = 0, comm_size = 2;
  auto world1 = world::create({32, 4, 4});
  auto decomp1 = decomposition::create(world1, comm_size);
  cout << decomp1 << endl;
 
  MPI_Finalize();
  return 0;
}