data_evaluator_impl.h

Go to the documentation of this file.

/*
 * Copyright (c) 2011-2017:  G-CSC, Goethe University Frankfurt
 * Authors: Andreas Vogel, Arne Naegel
 *
 * This file is part of UG4.
 *
 * UG4 is free software: you can redistribute it and/or modify it under the
 * terms of the GNU Lesser General Public License version 3 (as published by the
 * Free Software Foundation) with the following additional attribution
 * requirements (according to LGPL/GPL v3 §7):
 *
 * (1) The following notice must be displayed in the Appropriate Legal Notices
 * of covered and combined works: "Based on UG4 (www.ug4.org/license)".
 *
 * (2) The following notice must be displayed at a prominent place in the
 * terminal output of covered works: "Based on UG4 (www.ug4.org/license)".
 *
 * (3) The following bibliography is recommended for citation and must be
 * preserved in all covered files:
 * "Reiter, S., Vogel, A., Heppner, I., Rupp, M., and Wittum, G. A massively
 *   parallel geometric multigrid solver on hierarchically distributed grids.
 *   Computing and visualization in science 16, 4 (2013), 151-164"
 * "Vogel, A., Reiter, S., Rupp, M., Nägel, A., and Wittum, G. UG4 -- a novel
 *   flexible software system for simulating pde based models on high performance
 *   computers. Computing and visualization in science 16, 4 (2013), 165-179"
 *
 * This program 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.
 */
 
#ifndef __H__UG__LIB_DISC__SPATIAL_DISC__DATA_EVALUATOR_IMPL__
#define __H__UG__LIB_DISC__SPATIAL_DISC__DATA_EVALUATOR_IMPL__
 
namespace ug{
 
template <typename TDomain, typename TElemDisc>
DataEvaluatorBase<TDomain, TElemDisc>::
DataEvaluatorBase(int discPart,
              const std::vector<TElemDisc*>& vElemDisc,
              ConstSmartPtr<FunctionPattern> fctPat,
              const bool bNonRegularGrid,
              LocalVectorTimeSeries* pLocTimeSeries,
              const std::vector<number>* pvScaleMass,
              const std::vector<number>* pvScaleStiff)
   : m_spFctPattern(fctPat)
{
//  remember infos
  m_discPart = discPart;
  m_pLocTimeSeries = pLocTimeSeries;
  m_bNeedLocTimeSeries = false; // initially
  m_bUseHanging = false; // initially
 
  m_vElemDisc[PT_ALL] = vElemDisc;
 
//  create FunctionIndexMapping for each Disc
  for(size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
  {
  //  get elem disc
    TElemDisc* disc = m_vElemDisc[PT_ALL][i];
 
  //  handle time dependency
    if(pLocTimeSeries != NULL && pvScaleMass != NULL && pvScaleStiff != NULL){
      disc->set_time_dependent(*pLocTimeSeries, *pvScaleMass, *pvScaleStiff);
    }
    else if(pLocTimeSeries != NULL){
      disc->set_time_dependent(*pLocTimeSeries, std::vector<number>(), std::vector<number>());
    }
    else{
      disc->set_time_independent();
    }
 
  //  checks
    disc->check_setup(bNonRegularGrid);
 
  //  cache time dependency
    m_bNeedLocTimeSeries |= disc->local_time_series_needed();
 
  //  cache grid type required
    if(bNonRegularGrid)
      m_bUseHanging |= disc->use_hanging();
 
  //  sort by process type
    ProcessType process;
    if(!disc->is_time_dependent()) process = PT_STATIONARY;
    else process = PT_INSTATIONARY;
    m_vElemDisc[process].push_back(disc);
  }
}
 
 
// DataEvaluatorBase Setup
 
 
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::
add_data_to_eval_data(std::vector<SmartPtr<ICplUserData<dim> > >& vEvalData,
                      std::vector<SmartPtr<ICplUserData<dim> > >& vTryingToAdd)
{
//  if empty, we're done
  if(vTryingToAdd.empty()) return;
 
//  search for element in already scheduled data
  typename std::vector<SmartPtr<ICplUserData<dim> > >::iterator it, itEnd;
  it = find(vEvalData.begin(), vEvalData.end(), vTryingToAdd.back());
 
//  if found, skip this data
  if(it != vEvalData.end())
  {
    vTryingToAdd.pop_back();
    return;
  }
 
//  search if element already contained in list. Then, the element
//  did start the adding procedure before and a circle dependency
//  is found
  itEnd = vTryingToAdd.end(); itEnd--;
  it = find(vTryingToAdd.begin(), itEnd, *itEnd);
 
//  if found, return error of circle dependency
  if(it != itEnd)
    UG_THROW("DataEvaluatorBase::add_data_to_eval_data:"
            " Circle dependency of data detected for UserData.");
 
//  add all dependent datas
  SmartPtr<ICplUserData<dim> > data = vTryingToAdd.back();
  for(size_t i = 0; i < data->num_needed_data(); ++i)
  {
  //  add each data separately
    vTryingToAdd.push_back(data->needed_data(i));
    add_data_to_eval_data(vEvalData, vTryingToAdd);
  }
 
//  add this data to the evaluation list
  vEvalData.push_back(data);
 
//  pop last one, since now added to eval list
  if(!vTryingToAdd.empty())
    vTryingToAdd.pop_back();
}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::extract_imports_and_userdata(int subsetIndex, int discPart)
{
  clear_extracted_data_and_mappings();
 
//  queue for all user data needed
  std::vector<SmartPtr<ICplUserData<dim> > > vEvalData;
  std::vector<SmartPtr<ICplUserData<dim> > > vTryingToAdd;
 
//  In the next loop we extract all needed UserData:
//  We only process the DataImport if there has been set data to the import
//  since otherwise no evaluation is needed.
//  If there is data given, we get the connected UserData and add it to the vector
//  of EvaluationData. This simply adds the UserData to the queue for UserData, if
//  the data does not depend on other Data. But if the UserData itself has
//  dependencies to other UserData, this data is added first (in a recursive
//  process). Of course, no circle dependency between UserData is allowed.
 
//  In the same loop over the data imports, we schedule the DataImports for
//  evaluation and compute the correct FunctionMapping for the linearization
//  of the defect and the Data, the Import is connected to:
//  If the UserData does not depend on the primary unknowns, we're done. Else
//  we have to setup the Function mappings between the common function group
//  and the DataImport-FunctionGroup. This is simply the same function map as
//  for the element discretization, since the DataImport depends by definition
//  from and only from the primary variables of its associated IElemDisc.
 
//  loop elem discs
  for(size_t d = 0; d < m_vElemDisc[PT_ALL].size(); ++d)
  {
    TElemDisc* disc = m_vElemDisc[PT_ALL][d];
 
  //  loop imports
    for(size_t i = 0; i < disc->num_imports(); ++i)
    {
    //  get import
      IDataImport<dim>* iimp = &(disc->get_import(i));
 
    //  skip non-given data (no need for evaluation)
      if(!iimp->data_given()) continue;
 
    //  check part
      if( !(iimp->part() & discPart) ) continue;
 
    //  check correct process type
      if(iimp->part() == MASS)
        if(!disc->is_time_dependent()) continue;
 
    //  push export on stack of needed data
      vTryingToAdd.push_back(iimp->data());
 
    //  add data and all dependency to evaluation list
      try{
        add_data_to_eval_data(vEvalData, vTryingToAdd);
      }
      UG_CATCH_THROW("DataEvaluatorBase:"
            " Circle dependency of data detected for UserData.");
 
    //  check that queue is empty now, else some internal error occured
      if(!vTryingToAdd.empty())
        UG_THROW("DataEvaluatorBase:"
            " Internal Error, UserData queue not empty after adding.");
 
    //  done if and only if zero-derivative
      if(iimp->zero_derivative()) continue;
 
    //  remember Import
      ProcessType process;
      if(!disc->is_time_dependent()) process = PT_STATIONARY;
      else process = PT_INSTATIONARY;
 
      m_vImport[PT_ALL][iimp->part()].push_back(iimp);
      m_vImport[process][iimp->part()].push_back(iimp);
    }
  }
 
//  Now, we have processed all imports, that must be evaluated and have a long
//  vector of UserData that is connected to those imports. The UserData is already
//  sorted in this way: Data that depends on other data appears after the data
//  it depends on. This is important since we will schedule now the data for
//  evaluation and the data, that is needed by other data, will be computed
//  first. In addition, the data linker have to update their FunctionGroup and
//  must be sure that the data they depend on has already a correct FunctionGroup
//  set. This all is ensured by the (already produced) correct ordering.
//
//  In the next loop we process all UserData, that will be evaluated during
//  assembling (i.e. is connected to an Import). First, we check if the data
//  is constant. If so simply add it the the Constant Data vector; nothing more
//  has to be done here. Else we check if the data depends on the primary
//  unknowns. If this is not the case, the UserData must be a Position-dependent
//  data, but not constant. Thus, schedule it at the Position Data vector.
//  If the data depends on the primary unknowns we must proceed as follows:
//  First, we update the FunctionGroup of the Data, since it could be a linker
//  and having an incorrect FunctionGroup (iff the FunctionGroup of the data
//  the linker depends on has been changed). Then we create the function
//  mapping between the functions the linker depends on and the common Function
//  Group.
 
//  loop all needed user data and group it
  for(size_t i = 0; i < vEvalData.size(); ++i)
  {
  //  get the user data
    SmartPtr<ICplUserData<dim> > ipData = vEvalData[i];
 
  //  update function pattern (this will update functionGroups and Map of Data)
    try{
      ipData->set_function_pattern(m_spFctPattern);
    }
    UG_CATCH_THROW("DataEvaluatorBase: Cannot set FunctionPattern to UserData.");
 
  //  sort data into const and non-solution dependent
    if(ipData->constant()) {m_vConstData.push_back(ipData); continue;}
    if(ipData->zero_derivative()){m_vPosData.push_back(ipData); continue;}
 
  //  save as dependent data
    m_vDependentData.push_back(ipData);
  }
 
//  Handle time dependency
//  NOTE: constant data is not processed.
  if(m_pLocTimeSeries != NULL){
    for(size_t i = 0; i < m_vPosData.size(); ++i)
      m_vPosData[i]->set_times(m_pLocTimeSeries->times());
    for(size_t i = 0; i < m_vDependentData.size(); ++i)
      m_vDependentData[i]->set_times(m_pLocTimeSeries->times());
  }
 
//  NOTE: constant data is not processed, since constant == independent of si
  for(size_t i = 0; i < m_vPosData.size(); ++i)
    m_vPosData[i]->set_subset(subsetIndex);
  for(size_t i = 0; i < m_vDependentData.size(); ++i)
    m_vDependentData[i]->set_subset(subsetIndex);
}
void DataEvaluatorBase<TDomain, TElemDisc>::extract_imports_and_userdata(int subsetIndex, int discPart) {…}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::clear_extracted_data_and_mappings()
{
  for(int type = 0; type < MAX_PROCESS; ++type){
    m_vImport[type][MASS].clear();
    m_vImport[type][STIFF].clear();
    m_vImport[type][RHS].clear();
  }
 
  m_vConstData.clear();
  m_vPosData.clear();
  m_vDependentData.clear();
}
void DataEvaluatorBase<TDomain, TElemDisc>::clear_extracted_data_and_mappings() {…}
 
 
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::clear_positions_in_user_data()
{
//  remove ip series for all used UserData
  for(size_t i = 0; i < m_vConstData.size(); ++i) m_vConstData[i]->clear();
  for(size_t i = 0; i < m_vPosData.size(); ++i)   m_vPosData[i]->clear();
  for(size_t i = 0; i < m_vDependentData.size(); ++i) m_vDependentData[i]->clear();
}
void DataEvaluatorBase<TDomain, TElemDisc>::clear_positions_in_user_data() {…}
    return; {…}
 
 
// prepare / finish
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::set_time_point(const size_t timePoint)
{
  for(size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
    m_vElemDisc[PT_ALL][i]->set_time_point(timePoint);
 
  // NOTE: constant data is not processed.
  for(size_t i = 0; i < m_vPosData.size(); ++i)
    m_vPosData[i]->set_time_point(timePoint);
  for(size_t i = 0; i < m_vDependentData.size(); ++i)
    m_vDependentData[i]->set_time_point(timePoint);
}
void DataEvaluatorBase<TDomain, TElemDisc>::set_time_point(const size_t timePoint) {…}
 
 
 
// Error estimator's routines
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::
prepare_err_est_elem_loop(const ReferenceObjectID id, int si)
{
//  prepare loop (elem disc set local ip series here)
  try{
    for(size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_prep_err_est_elem_loop(id, si);
  }
  UG_CATCH_THROW("DataEvaluatorBase::prepare_err_est_elem_loop: "
            "Cannot prepare element loop.");
 
//  extract data imports and user data
  try{
    extract_imports_and_userdata(si, m_discPart);
  }
  UG_CATCH_THROW("DataEvaluatorBase::prepare_err_est_elem_loop: "
          "Cannot extract imports and userdata.");
 
//  check setup of imports
  try{
    for(size_t i = 0; i < m_vImport[PT_ALL][MASS].size(); ++i)
      m_vImport[PT_ALL][MASS][i]->check_setup();
    for(size_t i = 0; i < m_vImport[PT_ALL][STIFF].size(); ++i)
      m_vImport[PT_ALL][STIFF][i]->check_setup();
    for(size_t i = 0; i < m_vImport[PT_ALL][RHS].size(); ++i)
      m_vImport[PT_ALL][RHS][i]->check_setup();
  }
  UG_CATCH_THROW("DataEvaluatorBase::prepare_err_est_elem_loop: Import not correctly implemented.");
 
//  prepare and check dependent data
  try{
    for(size_t i = 0; i < m_vDependentData.size(); ++i){
      m_vDependentData[i]->check_setup();
    }
  }
  UG_CATCH_THROW("DataEvaluatorBase::prepare_err_est_elem_loop: Dependent UserData "
           " (e.g. Linker or Export) is not ready for evaluation.");
 
//  evaluate constant data
  for(size_t i = 0; i < m_vConstData.size(); ++i)
    m_vConstData[i]->compute((LocalVector*)NULL, NULL, NULL, false);
}
void DataEvaluatorBase<TDomain, TElemDisc>:: {…}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::finish_err_est_elem_loop()
{
//  finish each elem error estimator disc
  try{
    for(size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_fsh_err_est_elem_loop();
  }
  UG_CATCH_THROW("DataEvaluatorBase::finish_err_est_elem_loop: Cannot finish error estimator element loop");
 
//  clear positions at user data
  clear_positions_in_user_data();
}
void DataEvaluatorBase<TDomain, TElemDisc>::finish_err_est_elem_loop() {…}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::prepare_err_est_elem
(LocalVector& u, GridObject* elem, const MathVector<dim> vCornerCoords[],
const LocalIndices& ind, bool bDeriv)
{
  try
  {
    for (size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_prep_err_est_elem(u, elem, vCornerCoords);
  }
  UG_CATCH_THROW("DataEvaluatorBase::compute_elem_err_est: Cannot prepare element.");
 
//  evaluate position data
  for (size_t i = 0; i < m_vPosData.size(); ++i)
    m_vPosData[i]->compute(&u, elem, vCornerCoords, false);
 
//  process dependent data:
//  We can not simply compute exports first, then Linker, because an export
//  itself could depend on other data if implemented somehow in the IElemDisc
//  (e.g. using data from some DataImport). Thus, we have to loop the sorted
//  vector of all dependent data (that is correctly sorted the way that always
//  needed data has previously computed).
 
//  compute the data
  try
  {
    for (size_t i = 0; i < m_vDependentData.size(); ++i)
    {
      u.access_by_map(m_vDependentData[i]->map());
      m_vDependentData[i]->compute(&u, elem, vCornerCoords, false);
    }
  }
  UG_CATCH_THROW("DataEvaluatorBase::compute_elem_err_est: Cannot compute data for Export or Linker.");
}
void DataEvaluatorBase<TDomain, TElemDisc>::prepare_err_est_elem {…}
 
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::
compute_err_est_A_elem
( LocalVector& u,
  GridObject* elem,
  const MathVector<dim> vCornerCoords[],
  const LocalIndices& ind,
  const number scaleMass,
  const number scaleStiff)
{
  UG_ASSERT(m_discPart & STIFF, "Using compute_err_est_A_elem, but not STIFF requested.");
  try
  {
    for (std::size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_compute_err_est_A_elem(u, elem, vCornerCoords, scaleStiff);
  }
  UG_CATCH_THROW("DataEvaluatorBase::compute_err_est_A_elem: Cannot assemble stiffness part of error estimator");
}
void DataEvaluatorBase<TDomain, TElemDisc>:: {…}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::
compute_err_est_M_elem
( LocalVector& u,
  GridObject* elem,
  const MathVector<dim> vCornerCoords[],
  const LocalIndices& ind,
  const number scaleMass,
  const number scaleStiff)
{
  UG_ASSERT(m_discPart & MASS, "Using compute_err_est_M_elem, but not MASS requested.");
  try
  {
    for (std::size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_compute_err_est_M_elem(u, elem, vCornerCoords, scaleMass);
  }
  UG_CATCH_THROW("DataEvaluatorBase::compute_err_est_A_elem: Cannot assemble stiffness part of error estimator");
}
void DataEvaluatorBase<TDomain, TElemDisc>:: {…}
 
template <typename TDomain, typename TElemDisc>
void DataEvaluatorBase<TDomain, TElemDisc>::
compute_err_est_rhs_elem
( GridObject* elem,
  const MathVector<dim> vCornerCoords[],
  const LocalIndices& ind,
  const number scaleMass,
  const number scaleStiff)
{
  UG_ASSERT(m_discPart & RHS, "Using compute_err_est_rhs_elem, but not RHS requested.");
  try
  {
    for (std::size_t i = 0; i < m_vElemDisc[PT_ALL].size(); ++i)
      m_vElemDisc[PT_ALL][i]->do_compute_err_est_rhs_elem(elem, vCornerCoords, scaleStiff);
  }
  UG_CATCH_THROW("DataEvaluatorBase::compute_err_est_rhs_elem: Cannot assemble rhs part of error estimator");
}
void DataEvaluatorBase<TDomain, TElemDisc>:: {…}
 
 
} // namespace ug
 
 
#endif
  if(it != vEvalData.end()) {…}
  typename std::vector<SmartPtr<ICplUserData<dim> > >::iterator it, itEnd; {…}
  if(vTryingToAdd.empty()) return; {…}
add_data_to_eval_data(std::vector<SmartPtr<ICplUserData<dim> > >& vEvalData, {…}
void DataEvaluatorBase<TDomain, TElemDisc>:: {…}
  {…}
  } {…}
    ProcessType process; {…}
      m_bUseHanging |= disc->use_hanging(); {…}
      disc->set_time_independent(); {…}
  {…}
DataEvaluatorBase<TDomain, TElemDisc>:: {…}