21#ifndef UTILS_MPI_DETAIL_SIZE_AND_OFFSET_HPP
22#define UTILS_MPI_DETAIL_SIZE_AND_OFFSET_HPP
28#include <boost/mpi/collectives.hpp>
29#include <boost/mpi/communicator.hpp>
36int size_and_offset(std::vector<int> &
sizes, std::vector<int> &
displ,
37 int n_elem,
const boost::mpi::communicator &comm,
39 auto const world_size =
static_cast<unsigned int>(comm.size());
40 sizes.resize(world_size);
41 displ.resize(world_size);
49 for (
unsigned int i = 0; i <
sizes.size(); i++) {
57inline void size_and_offset(
int n_elem,
const boost::mpi::communicator &comm,
cudaStream_t stream[1]
CUDA streams for parallel computing on CPU and GPU.