Observe.hpp

// Distributed under the MIT License.
// See LICENSE.txt for details.
 
#pragma once
 
#include <cstddef>
#include <string>
#include <utility>
#include <vector>
 
#include "IO/Observer/ObservationId.hpp"
#include "IO/Observer/ObserverComponent.hpp"
#include "IO/Observer/ReductionActions.hpp"
#include "IO/Observer/TypeOfObservation.hpp"
#include "Parallel/GlobalCache.hpp"
#include "Parallel/Invoke.hpp"
#include "Parallel/Reduction.hpp"
#include "ParallelAlgorithms/LinearSolver/Tags.hpp"
#include "Utilities/Functional.hpp"
#include "Utilities/PrettyType.hpp"
#include "Utilities/System/ParallelInfo.hpp"
 
namespace LinearSolver {
namespace observe_detail {
 
using reduction_data = Parallel::ReductionData<
    // Iteration
    Parallel::ReductionDatum<size_t, funcl::AssertEqual<>>,
    // Residual
    Parallel::ReductionDatum<double, funcl::AssertEqual<>>>;
 
template <typename OptionsGroup>
struct Registration {
  template <typename ParallelComponent, typename DbTagsList,
            typename ArrayIndex>
  static std::pair<observers::TypeOfObservation, observers::ObservationKey>
  register_info(const db::DataBox<DbTagsList>& /*box*/,
                const ArrayIndex& /*array_index*/) noexcept {
    return {observers::TypeOfObservation::Reduction,
            observers::ObservationKey{pretty_type::get_name<OptionsGroup>()}};
  }
};
 
/*!
 * \brief Contributes data from the residual monitor to the reduction observer
 */
template <typename OptionsGroup, typename Metavariables>
void contribute_to_reduction_observer(
    const size_t iteration_id, const double residual_magnitude,
    Parallel::GlobalCache<Metavariables>& cache) noexcept {
  const auto observation_id = observers::ObservationId(
      iteration_id, pretty_type::get_name<OptionsGroup>());
  auto& reduction_writer = Parallel::get_parallel_component<
      observers::ObserverWriter<Metavariables>>(cache);
  Parallel::threaded_action<observers::ThreadedActions::WriteReductionData>(
      // Node 0 is always the writer, so directly call the component on that
      // node
      reduction_writer[0], observation_id, static_cast<size_t>(sys::my_node()),
      // When multiple linear solves are performed, e.g. for the nonlinear
      // solver, we'll need to write into separate subgroups, e.g.:
      // `/linear_residuals/<nonlinear_iteration_id>`
      std::string{"/" + Options::name<OptionsGroup>() + "Residuals"},
      std::vector<std::string>{"Iteration", "Residual"},
      reduction_data{iteration_id, residual_magnitude});
}
 
}  // namespace observe_detail
}  // namespace LinearSolver