Homec4science
Diffusion tamaas (features/netcdf-mpi) Browse

tamaas/src/core/loops/loop_omp.hh
f76320c57280features/netcdf-mpi

Last Change
Raw File

loop_omp.hh
View Options

/**
* @file
*
* @author Lucas Frérot <lucas.frerot@epfl.ch>
*
* @section LICENSE
*
* Copyright (©) 2017 EPFL (Ecole Polytechnique Fédérale de
* Lausanne) Laboratory (LSMS - Laboratoire de Simulation en Mécanique des
* Solides)
*
* Tamaas is free software: you can redistribute it and/or modify it under the
* terms of the GNU Lesser General Public License as published by the Free
* Software Foundation, either version 3 of the License, or (at your option) any
* later version.
*
* Tamaas is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
* A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
* details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with Tamaas. If not, see <http://www.gnu.org/licenses/>.
*
*/
/* -------------------------------------------------------------------------- */
#ifndef __LOOP_OMP_HH__
#define __LOOP_OMP_HH__
/* -------------------------------------------------------------------------- */
#include "loop.hh"
#include "loop_utils.hh"
/* -------------------------------------------------------------------------- */
__BEGIN_TAMAAS__
namespace detail {
template<bool strided, typename Functor, typename... Grids>
void exec_loop_omp(Functor&& func, Grids&&... grids) {
const UInt loop_size = loopSize<strided>(grids...);
#pragma omp parallel for
for (UInt index = 0 ; index < loop_size ; ++index) {
if (!strided) func(grids(index)...);
else func(grids(index * grids.getNbComponents())...);
}
}
template <operation op, bool strided, typename Functor, typename... Grids>
auto reduce_loop_omp(Functor&& func, Grids&&... grids)
-> decltype(func(grids(0)...)) {
const UInt loop_size = loopSize<strided>(grids...);
using ret_type = decltype(func(grids(0)...));
ret_type res = reduction_helper<op>::template init<ret_type>();
#pragma omp parallel
{
ret_type res_local = res;
#pragma omp for
for (UInt index = 0 ; index < loop_size ; ++index) {
if (!strided) reduction_helper<op>::reduce(res_local, func(grids(index)...));
else reduction_helper<op>::reduce(res_local, func(grids(index * grids.getNbComponents())...));
}
#pragma omp critical
reduction_helper<op>::reduce(res, res_local);
}
return res;
}
}
__END_TAMAAS__
#endif // __LOOP_PROXY_HH__
c4science · Help