Line data    Source code

       1            0 : // Distributed under the MIT License.
       2              : // See LICENSE.txt for details.
       3              : 
       4              : #pragma once
       5              : 
       6              : #include <cstddef>
       7              : 
       8              : namespace LinearSolver::Schwarz {
       9              : 
      10              : /*!
      11              :  * \brief Abstract base class for the subdomain operator, i.e. the linear
      12              :  * operator restricted to an element-centered Schwarz subdomain
      13              :  *
      14              :  * A subdomain operator must implement these member function templates
      15              :  * - `operator()`: Applies the operator to the element-centered subdomain data.
      16              :  *   It must take these arguments, in this order:
      17              :  *   - `gsl::not_null<LinearSolver::Schwarz::ElementCenteredSubdomainData*>`:
      18              :  *     The data where the result of the operator applied to the operand should
      19              :  *     be written to
      20              :  *   - `LinearSolver::Schwarz::ElementCenteredSubdomainData`: The operand data
      21              :  *     to which the operator should be applied
      22              :  *   - `db::DataBox`: The element's DataBox. Can be used to retrieve any data on
      23              :  *     the subdomain geometry and to access and mutate persistent memory
      24              :  *     buffers.
      25              :  *
      26              :  * Since the subdomain operator has access to the element's DataBox, it can
      27              :  * retrieve any background data, i.e. data that does not depend on the variables
      28              :  * it operates on. Background data on overlap regions with other elements can be
      29              :  * either initialized in advance if possible or communicated with the
      30              :  * `LinearSolver::Schwarz::Actions::SendOverlapFields` and
      31              :  * `LinearSolver::Schwarz::Actions::ReceiveOverlapFields` actions. Note that
      32              :  * such communication should not be necessary between iterations of the Schwarz
      33              :  * solve, but only between successive solves, because background data should not
      34              :  * change during the solve. The Schwarz algorithm takes care of communicating
      35              :  * all variable data that the subdomain operator operates on, whenever
      36              :  * necessary. This variable data is passed to the operator as the `operand`
      37              :  * argument (see above). It includes the data on the central element of the
      38              :  * subdomain, as well as the data on overlap regions with neighbors. Since data
      39              :  * on the entire subdomain is available, applying the subdomain operator
      40              :  * requires _no_ communication between elements. This is the strength of the
      41              :  * Schwarz algorithm: all subdomain solves are independent of each other (see
      42              :  * `LinearSolver::Schwarz::Schwarz` for details).
      43              :  *
      44              :  * Here's an example of a subdomain operator that is the restriction of an
      45              :  * explicit global matrix:
      46              :  *
      47              :  * \snippet Test_SchwarzAlgorithm.cpp subdomain_operator
      48              :  */
      49              : template <size_t Dim>
      50            1 : class SubdomainOperator {
      51              :  public:
      52            0 :   static constexpr size_t volume_dim = Dim;
      53              : 
      54            0 :   SubdomainOperator() = default;
      55              : };
      56              : 
      57              : }  // namespace LinearSolver::Schwarz