ComputeUserObjectsThread Class Reference

Class for threaded computation of UserObjects. More...

#include <ComputeUserObjectsThread.h>

Inheritance diagram for ComputeUserObjectsThread:

Public Member Functions
	ComputeUserObjectsThread (FEProblemBase &problem, const TheWarehouse::Query &query)
	ComputeUserObjectsThread (ComputeUserObjectsThread &x, Threads::split)
virtual	~ComputeUserObjectsThread ()
virtual void	onElement (const Elem *elem) override
	Assembly of the element (not including surface assembly) More...
virtual void	onBoundary (const Elem *elem, unsigned int side, BoundaryID bnd_id, const Elem *lower_d_elem=nullptr) override
	Called when doing boundary assembling. More...
virtual void	onInternalSide (const Elem *elem, unsigned int side) override
	Called when doing internal edge assembling. More...
virtual void	onExternalSide (const Elem *elem, unsigned int side) override
	Called when iterating over external sides (no side neighbor) More...
virtual void	onInterface (const Elem *elem, unsigned int side, BoundaryID bnd_id) override
	Called when doing interface assembling. More...
virtual void	post () override
	Called after the element range loop. More...
virtual void	subdomainChanged () override
	Called every time the current subdomain changes (i.e. More...
void	join (const ComputeUserObjectsThread &)
virtual void	caughtMooseException (MooseException &e) override
	Called if a MooseException is caught anywhere during the computation. More...
virtual bool	keepGoing () override
	Whether or not the loop should continue. More...
virtual void	preElement (const Elem *elem) override
	Called before the element assembly. More...
virtual void	preInternalSide (const Elem *elem, unsigned int side) override
	Called before evaluations on an element internal side. More...
virtual void	preBoundary (const Elem *elem, unsigned int side, BoundaryID bnd_id, const Elem *lower_d_elem=nullptr) override
	Called before the boundary assembly. More...
virtual void	neighborSubdomainChanged () override
	Called every time the neighbor subdomain changes (i.e. More...
virtual void	operator() (const ConstElemRange &range, bool bypass_threading=false)
virtual void	pre ()
	Called before the element range loop. More...
virtual void	postElement (const Elem *elem)
	Called after the element assembly is done (including surface assembling) More...
virtual void	postInternalSide (const Elem *elem, unsigned int side)
	Called after evaluations on an element internal side. More...

Protected Member Functions
void	printGeneralExecutionInformation () const override
	Print general information about the loop, like the ordering of class of objects. More...
void	printBlockExecutionInformation () const override
	Print information about the loop, mostly order of execution of particular objects. More...
template<typename T >
void	printVectorOrdering (std::vector< T *> uos, const std::string &name) const
	Format output of vector of UOs. More...
void	prepareElement (const Elem *elem)
void	clearVarsAndMaterials ()
void	printExecutionOrdering (const std::vector< T * > &objs, const bool print_header=true, const std::string &line_prefix="[DBG]") const
	Routine to output the ordering of objects within a vector of pointers to these objects. More...
void	printExecutionOrdering (const std::vector< std::shared_ptr< T >> &objs_ptrs, const bool print_header=true, const std::string &line_prefix="[DBG]") const
virtual void	printBoundaryExecutionInformation (const unsigned int) const
	Print information about the particular ordering of objects on each boundary. More...
void	resetExecPrintedSets () const
	Resets the set of blocks and boundaries visited. More...
virtual bool	shouldComputeInternalSide (const Elem &elem, const Elem &neighbor) const
	Whether to compute the internal side for the provided element-neighbor pair. More...

Protected Attributes
FEProblemBase &	_fe_problem
MooseMesh &	_mesh
THREAD_ID	_tid
SubdomainID	_subdomain
	The subdomain for the current element. More...
SubdomainID	_old_subdomain
	The subdomain for the last element. More...
SubdomainID	_neighbor_subdomain
	The subdomain for the current neighbor. More...
SubdomainID	_old_neighbor_subdomain
	The subdomain for the last neighbor. More...
std::set< SubdomainID >	_blocks_exec_printed
	Keep track of which blocks were visited. More...
std::set< BoundaryID >	_boundaries_exec_printed
	Keep track of which boundaries were visited. More...

Private Member Functions
template<typename T >
void	querySubdomain (Interfaces iface, std::vector< T > &results)
template<typename T >
void	queryBoundary (Interfaces iface, BoundaryID bnd, std::vector< T > &results)

Private Attributes
const TheWarehouse::Query	_query
TheWarehouse::QueryCache< AttribThread, AttribSubdomains, AttribInterfaces >	_query_subdomain
TheWarehouse::QueryCache< AttribThread, AttribBoundaries, AttribInterfaces >	_query_boundary
std::vector< InternalSideUserObject * >	_internal_side_objs
std::vector< InterfaceUserObject * >	_interface_user_objects
std::vector< ElementUserObject * >	_element_objs
std::vector< ShapeElementUserObject * >	_shape_element_objs
std::vector< DomainUserObject * >	_domain_objs
std::vector< DomainUserObject * >	_all_domain_objs
AuxiliarySystem &	_aux_sys

Detailed Description

Class for threaded computation of UserObjects.

Definition at line 37 of file ComputeUserObjectsThread.h.

Constructor & Destructor Documentation

ComputeUserObjectsThread() [1/2]

ComputeUserObjectsThread::ComputeUserObjectsThread	(	FEProblemBase &	problem,
		const TheWarehouse::Query &	query
	)

Definition at line 29 of file ComputeUserObjectsThread.C.

  : ThreadedElementLoop<ConstElemRange>(problem),
    _query(query),
    _query_subdomain(_query),
    _query_boundary(_query),
    _aux_sys(problem.getAuxiliarySystem())
{
}

ComputeUserObjectsThread() [2/2]

ComputeUserObjectsThread::ComputeUserObjectsThread	(	ComputeUserObjectsThread &	x,
		Threads::split
	)

Definition at line 40 of file ComputeUserObjectsThread.C.

  : ThreadedElementLoop<ConstElemRange>(x._fe_problem),
    _query(x._query),
    _query_subdomain(x._query_subdomain),
    _query_boundary(x._query_boundary),
    _aux_sys(x._aux_sys)
{
}

~ComputeUserObjectsThread()

ComputeUserObjectsThread::~ComputeUserObjectsThread ( )

virtual

Definition at line 49 of file ComputeUserObjectsThread.C.

{}

Member Function Documentation

caughtMooseException()

void ThreadedElementLoop< ConstElemRange >::caughtMooseException ( MooseException &  e )

overridevirtualinherited

Called if a MooseException is caught anywhere during the computation.

The single input parameter taken is a MooseException object.

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 105 of file ThreadedElementLoop.h.

Referenced by ComputeJacobianForScalingThread::operator()().

{
  Threads::spin_mutex::scoped_lock lock(threaded_element_mutex);

  std::string what(e.what());
  _fe_problem.setException(what);
}

clearVarsAndMaterials()

void ThreadedElementLoop< ConstElemRange >::clearVarsAndMaterials ( )

protectedinherited

Definition at line 195 of file ThreadedElementLoop.h.

Referenced by NonlinearThread::post().

{
  _fe_problem.clearActiveElementalMooseVariables(this->_tid);
  _fe_problem.clearActiveMaterialProperties(this->_tid);
}

join()

void ComputeUserObjectsThread::join

(

const ComputeUserObjectsThread &

)

Definition at line 325 of file ComputeUserObjectsThread.C.

{
}

keepGoing()

virtual bool ThreadedElementLoop< ConstElemRange >::keepGoing ( )

inlineoverridevirtualinherited

Whether or not the loop should continue.

Returns

true to keep going, false to stop.

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 45 of file ThreadedElementLoop.h.

Referenced by ComputeJacobianForScalingThread::operator()().

{ return !_fe_problem.hasException(); }

neighborSubdomainChanged()

void ThreadedElementLoop< ConstElemRange >::neighborSubdomainChanged ( )

overridevirtualinherited

Called every time the neighbor subdomain changes (i.e.

the subdomain of this neighbor is not the same as the subdomain of the last neighbor). Beware of over-using this! You might think that you can do some expensive stuff in here and get away with it... but there are applications that have TONS of subdomains....

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 139 of file ThreadedElementLoop.h.

{
  _fe_problem.neighborSubdomainSetup(ThreadedElementLoopBase<RangeType>::_neighbor_subdomain,
                                     ThreadedElementLoopBase<RangeType>::_tid);
}

onBoundary()

void ComputeUserObjectsThread::onBoundary ( const Elem *  elem, unsigned int  side, BoundaryID  bnd_id, const Elem *  lower_d_elem = nullptr  )

overridevirtual

Called when doing boundary assembling.

Parameters

elem	- The element we are checking is on the boundary.
side	- The side of the element in question.
bnd_id	- ID of the boundary we are at
lower_d_elem	- Lower dimensional element (e.g. Mortar)

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 162 of file ComputeUserObjectsThread.C.

{
  std::vector<UserObject *> userobjs;
  queryBoundary(Interfaces::SideUserObject, bnd_id, userobjs);
  if (userobjs.size() == 0 && _domain_objs.size() == 0)
    return;

  _fe_problem.reinitElemFace(elem, side, _tid);

  // Reinitialize lower-dimensional variables for use in boundary Materials
  if (lower_d_elem)
    _fe_problem.reinitLowerDElem(lower_d_elem, _tid);

  // Set up Sentinel class so that, even if reinitMaterialsFace() throws, we
  // still remember to swap back during stack unwinding.
  SwapBackSentinel sentinel(_fe_problem, &FEProblem::swapBackMaterialsFace, _tid);
  _fe_problem.reinitMaterialsFaceOnBoundary(bnd_id, elem->subdomain_id(), _tid);
  _fe_problem.reinitMaterialsBoundary(bnd_id, _tid);

  for (const auto & uo : userobjs)
    uo->execute();

  for (auto & uo : _domain_objs)
  {
    uo->preExecuteOnBoundary();
    uo->executeOnBoundary();
  }

  // UserObject Jacobians
  std::vector<ShapeSideUserObject *> shapers;
  queryBoundary(Interfaces::ShapeSideUserObject, bnd_id, shapers);
  if (_fe_problem.currentlyComputingJacobian() && shapers.size() > 0)
  {
    // Prepare shape functions for ShapeSideUserObjects
    const auto & jacobian_moose_vars = _fe_problem.getUserObjectJacobianVariables(_tid);
    for (const auto & jvar : jacobian_moose_vars)
    {
      unsigned int jvar_id = jvar->number();
      auto && dof_indices = jvar->dofIndices();

      _fe_problem.prepareFaceShapes(jvar_id, _tid);

      for (const auto & uo : shapers)
        uo->executeJacobianWrapper(jvar_id, dof_indices);
    }
  }
}

onElement()

void ComputeUserObjectsThread::onElement ( const Elem *  elem )

overridevirtual

Assembly of the element (not including surface assembly)

Parameters

elem	- active element

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 118 of file ComputeUserObjectsThread.C.

{
  _fe_problem.prepare(elem, _tid);
  _fe_problem.reinitElem(elem, _tid);

  // Set up Sentinel class so that, even if reinitMaterials() throws, we
  // still remember to swap back during stack unwinding.
  SwapBackSentinel sentinel(_fe_problem, &FEProblem::swapBackMaterials, _tid);
  _fe_problem.reinitMaterials(_subdomain, _tid);

  for (const auto & uo : _element_objs)
  {
    uo->execute();

    // update the aux solution vector if writable coupled variables are used
    if (uo->hasWritableCoupledVariables())
      for (auto * var : uo->getWritableCoupledVariables())
        var->insert(_aux_sys.solution());
  }

  for (auto & uo : _domain_objs)
  {
    uo->preExecuteOnElement();
    uo->executeOnElement();
  }

  // UserObject Jacobians
  if (_fe_problem.currentlyComputingJacobian() && _shape_element_objs.size() > 0)
  {
    // Prepare shape functions for ShapeElementUserObjects
    const auto & jacobian_moose_vars = _fe_problem.getUserObjectJacobianVariables(_tid);
    for (const auto & jvar : jacobian_moose_vars)
    {
      unsigned int jvar_id = jvar->number();
      auto && dof_indices = jvar->dofIndices();

      _fe_problem.prepareShapes(jvar_id, _tid);
      for (const auto uo : _shape_element_objs)
        uo->executeJacobianWrapper(jvar_id, dof_indices);
    }
  }
}

onExternalSide()

void ComputeUserObjectsThread::onExternalSide ( const Elem *  elem, unsigned int  side  )

overridevirtual

Called when iterating over external sides (no side neighbor)

Parameters

elem	- Element we are on
side	- local side number of the element 'elem'

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 252 of file ComputeUserObjectsThread.C.

{
  // We are not initializing any materials here because objects that perform calculations should
  // run onBoundary. onExternalSide should be used for mesh updates (e.g. adding/removing
  // boundaries). Note that _current_elem / _current_side are not getting updated either.
  for (auto & uo : _domain_objs)
    uo->executeOnExternalSide(elem, side);
}

onInterface()

void ComputeUserObjectsThread::onInterface ( const Elem *  elem, unsigned int  side, BoundaryID  bnd_id  )

overridevirtual

Called when doing interface assembling.

Parameters

elem	- Element we are on
side	- local side number of the element 'elem'
bnd_id	- ID of the interface we are at

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 262 of file ComputeUserObjectsThread.C.

{
  // Pointer to the neighbor we are currently working on.
  const Elem * neighbor = elem->neighbor_ptr(side);
  if (!(neighbor->active()))
    return;

  std::vector<UserObject *> interface_objs;
  queryBoundary(Interfaces::InterfaceUserObject, bnd_id, interface_objs);

  bool has_domain_objs = false;
  // we need to check all domain user objects because a domain user object may not be active
  // on the current subdomain but should be executed on the interface that it attaches to
  for (const auto * const domain_uo : _all_domain_objs)
    if (domain_uo->shouldExecuteOnInterface())
    {
      has_domain_objs = true;
      break;
    }

  // if we do not have any interface user objects and domain user objects on the current
  // interface
  if (interface_objs.empty() && !has_domain_objs)
    return;

  _fe_problem.prepareFace(elem, _tid);
  _fe_problem.reinitNeighbor(elem, side, _tid);

  // Set up Sentinels so that, even if one of the reinitMaterialsXXX() calls throws, we
  // still remember to swap back during stack unwinding.

  SwapBackSentinel face_sentinel(_fe_problem, &FEProblem::swapBackMaterialsFace, _tid);
  _fe_problem.reinitMaterialsFaceOnBoundary(bnd_id, elem->subdomain_id(), _tid);
  _fe_problem.reinitMaterialsBoundary(bnd_id, _tid);

  SwapBackSentinel neighbor_sentinel(_fe_problem, &FEProblem::swapBackMaterialsNeighbor, _tid);
  _fe_problem.reinitMaterialsNeighbor(neighbor->subdomain_id(), _tid);

  // Has to happen after face and neighbor properties have been computed. Note that we don't use
  // a sentinel here because FEProblem::swapBackMaterialsFace is going to handle face materials,
  // boundary materials, and interface materials (e.g. it queries the boundary material data
  // with the current element and side
  _fe_problem.reinitMaterialsInterface(bnd_id, _tid);

  for (const auto & uo : interface_objs)
    uo->execute();

  for (auto & uo : _all_domain_objs)
    if (uo->shouldExecuteOnInterface())
    {
      uo->preExecuteOnInterface();
      uo->executeOnInterface();
    }
}

onInternalSide()

void ComputeUserObjectsThread::onInternalSide ( const Elem *  elem, unsigned int  side  )

overridevirtual

Called when doing internal edge assembling.

Parameters

elem	- Element we are on
side	- local side number of the element 'elem'

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 214 of file ComputeUserObjectsThread.C.

{
  // Pointer to the neighbor we are currently working on.
  const Elem * neighbor = elem->neighbor_ptr(side);

  // Get the global id of the element and the neighbor
  const dof_id_type elem_id = elem->id(), neighbor_id = neighbor->id();

  if (_internal_side_objs.size() == 0 && _domain_objs.size() == 0)
    return;
  if (!((neighbor->active() && (neighbor->level() == elem->level()) && (elem_id < neighbor_id)) ||
        (neighbor->level() < elem->level())))
    return;

  _fe_problem.prepareFace(elem, _tid);
  _fe_problem.reinitNeighbor(elem, side, _tid);

  // Set up Sentinels so that, even if one of the reinitMaterialsXXX() calls throws, we
  // still remember to swap back during stack unwinding.
  SwapBackSentinel face_sentinel(_fe_problem, &FEProblem::swapBackMaterialsFace, _tid);
  _fe_problem.reinitMaterialsFace(elem->subdomain_id(), _tid);

  SwapBackSentinel neighbor_sentinel(_fe_problem, &FEProblem::swapBackMaterialsNeighbor, _tid);
  _fe_problem.reinitMaterialsNeighbor(neighbor->subdomain_id(), _tid);

  for (const auto & uo : _internal_side_objs)
    if (!uo->blockRestricted() || uo->hasBlocks(neighbor->subdomain_id()))
      uo->execute();

  for (auto & uo : _domain_objs)
    if (!uo->blockRestricted() || uo->hasBlocks(neighbor->subdomain_id()))
    {
      uo->preExecuteOnInternalSide();
      uo->executeOnInternalSide();
    }
}

operator()()

void ThreadedElementLoopBase< ConstElemRange >::operator() ( const ConstElemRange &  range, bool  bypass_threading = false  )

virtualinherited

Reimplemented in NonlinearThread, and ComputeJacobianForScalingThread.

Definition at line 226 of file ThreadedElementLoopBase.h.

{
  try
  {
    try
    {
      ParallelUniqueId puid;
      _tid = bypass_threading ? 0 : puid.id;

      pre();
      printGeneralExecutionInformation();

      _subdomain = Moose::INVALID_BLOCK_ID;
      _neighbor_subdomain = Moose::INVALID_BLOCK_ID;
      typename RangeType::const_iterator el = range.begin();
      for (el = range.begin(); el != range.end(); ++el)
      {
        if (!keepGoing())
          break;

        const Elem * elem = *el;

        preElement(elem);

        _old_subdomain = _subdomain;
        _subdomain = elem->subdomain_id();
        if (_subdomain != _old_subdomain)
        {
          subdomainChanged();
          printBlockExecutionInformation();
        }

        onElement(elem);

        if (_mesh.interiorLowerDBlocks().count(elem->subdomain_id()) > 0 ||
            _mesh.boundaryLowerDBlocks().count(elem->subdomain_id()) > 0)
        {
          postElement(elem);
          continue;
        }

        for (unsigned int side = 0; side < elem->n_sides(); side++)
        {
          std::vector<BoundaryID> boundary_ids = _mesh.getBoundaryIDs(elem, side);
          const Elem * lower_d_elem = _mesh.getLowerDElem(elem, side);

          if (boundary_ids.size() > 0)
            for (std::vector<BoundaryID>::iterator it = boundary_ids.begin();
                 it != boundary_ids.end();
                 ++it)
            {
              preBoundary(elem, side, *it, lower_d_elem);
              printBoundaryExecutionInformation(*it);
              onBoundary(elem, side, *it, lower_d_elem);
            }

          const Elem * neighbor = elem->neighbor_ptr(side);
          if (neighbor)
          {
            preInternalSide(elem, side);

            _old_neighbor_subdomain = _neighbor_subdomain;
            _neighbor_subdomain = neighbor->subdomain_id();
            if (_neighbor_subdomain != _old_neighbor_subdomain)
              neighborSubdomainChanged();

            if (shouldComputeInternalSide(*elem, *neighbor))
              onInternalSide(elem, side);

            if (boundary_ids.size() > 0)
              for (std::vector<BoundaryID>::iterator it = boundary_ids.begin();
                   it != boundary_ids.end();
                   ++it)
                onInterface(elem, side, *it);

            postInternalSide(elem, side);
          }
          else
            onExternalSide(elem, side);
        } // sides

        postElement(elem);
      } // range

      post();
      resetExecPrintedSets();
    }
    catch (MetaPhysicL::LogicError & e)
    {
      moose::translateMetaPhysicLError(e);
    }
    catch (std::exception & e)
    {
      // Continue if we find a libMesh degenerate map exception, but
      // just re-throw for any real error
      if (!strstr(e.what(), "Jacobian") && !strstr(e.what(), "singular") &&
          !strstr(e.what(), "det != 0"))
        throw; // not "throw e;" - that destroys type info!

      mooseException("We caught a libMesh degeneracy exception in ThreadedElementLoopBase:\n",
                     e.what());
    }
  }
  catch (MooseException & e)
  {
    caughtMooseException(e);
  }
}

post()

void ComputeUserObjectsThread::post ( )

overridevirtual

Called after the element range loop.

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 318 of file ComputeUserObjectsThread.C.

{
  _fe_problem.clearActiveElementalMooseVariables(_tid);
  _fe_problem.clearActiveMaterialProperties(_tid);
}

postElement()

void ThreadedElementLoopBase< ConstElemRange >::postElement ( const Elem *  elem )

virtualinherited

Called after the element assembly is done (including surface assembling)

Parameters

elem	- active element

Reimplemented in ComputeJacobianBlocksThread, NonlinearThread, ComputeIndicatorThread, ComputeJacobianThread, and ComputeMarkerThread.

Definition at line 361 of file ThreadedElementLoopBase.h.

{
}

postInternalSide()

void ThreadedElementLoopBase< ConstElemRange >::postInternalSide ( const Elem *  elem, unsigned int  side  )

virtualinherited

Called after evaluations on an element internal side.

Parameters

elem	- Element we are on
side	- local side number of the element 'elem'

Reimplemented in ComputeJacobianBlocksThread.

Definition at line 391 of file ThreadedElementLoopBase.h.

{
}

pre()

void ThreadedElementLoopBase< ConstElemRange >::pre ( )

virtualinherited

Called before the element range loop.

Definition at line 337 of file ThreadedElementLoopBase.h.

Referenced by ComputeJacobianForScalingThread::operator()().

{
}

preBoundary()

void ThreadedElementLoop< ConstElemRange >::preBoundary ( const Elem *  elem, unsigned int  side, BoundaryID  bnd_id, const Elem *  lower_d_elem = nullptr  )

overridevirtualinherited

Called before the boundary assembly.

Parameters

elem	- The element we are checking is on the boundary.
side	- The side of the element in question.
bnd_id	- ID of the boundary we are at
lower_d_elem	- Lower dimensional element (e.g. Mortar)

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 129 of file ThreadedElementLoop.h.

{
  _fe_problem.setCurrentBoundaryID(bnd_id, ThreadedElementLoopBase<RangeType>::_tid);
}

preElement()

void ThreadedElementLoop< ConstElemRange >::preElement ( const Elem *  elem )

overridevirtualinherited

Called before the element assembly.

Parameters

elem	- active element

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 115 of file ThreadedElementLoop.h.

Referenced by ComputeJacobianForScalingThread::operator()().

{
  _fe_problem.setCurrentSubdomainID(el, ThreadedElementLoopBase<RangeType>::_tid);
}

preInternalSide()

void ThreadedElementLoop< ConstElemRange >::preInternalSide ( const Elem *  elem, unsigned int  side  )

overridevirtualinherited

Called before evaluations on an element internal side.

Parameters

elem	- Element we are on
side	- local side number of the element 'elem'

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 122 of file ThreadedElementLoop.h.

{
  _fe_problem.setNeighborSubdomainID(el, side, ThreadedElementLoopBase<RangeType>::_tid);
}

prepareElement()

void ThreadedElementLoop< ConstElemRange >::prepareElement ( const Elem *  elem )

protectedinherited

Definition at line 186 of file ThreadedElementLoop.h.

Referenced by NonlinearThread::onElement().

{
  _fe_problem.prepare(elem, this->_tid);
  _fe_problem.reinitElem(elem, this->_tid);
  _fe_problem.reinitMaterials(this->_subdomain, this->_tid);
}

printBlockExecutionInformation()

void ComputeUserObjectsThread::printBlockExecutionInformation ( ) const

overrideprotectedvirtual

Print information about the loop, mostly order of execution of particular objects.

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 360 of file ComputeUserObjectsThread.C.

{
  if (!_fe_problem.shouldPrintExecution(_tid))
    return;

  // Gather all user objects that may execute
  // TODO: restrict this gathering of boundary objects to boundaries that are present
  // in the current block
  std::vector<ShapeSideUserObject *> shapers;
  const_cast<ComputeUserObjectsThread *>(this)->queryBoundary(
      Interfaces::ShapeSideUserObject, Moose::ANY_BOUNDARY_ID, shapers);

  std::vector<SideUserObject *> side_uos;
  const_cast<ComputeUserObjectsThread *>(this)->queryBoundary(
      Interfaces::SideUserObject, Moose::ANY_BOUNDARY_ID, side_uos);

  std::vector<InterfaceUserObject *> interface_objs;
  const_cast<ComputeUserObjectsThread *>(this)->queryBoundary(
      Interfaces::InterfaceUserObject, Moose::ANY_BOUNDARY_ID, interface_objs);

  std::vector<const DomainUserObject *> domain_interface_uos;
  for (const auto * const domain_uo : _domain_objs)
    if (domain_uo->shouldExecuteOnInterface())
      domain_interface_uos.push_back(domain_uo);

  // Approximation of the number of user objects currently executing
  const auto num_objects = _element_objs.size() + _domain_objs.size() + _shape_element_objs.size() +
                           side_uos.size() + shapers.size() + _internal_side_objs.size() +
                           interface_objs.size() + domain_interface_uos.size();

  const auto & console = _fe_problem.console();
  const auto & execute_on = _fe_problem.getCurrentExecuteOnFlag();

  if (num_objects > 0)
  {
    if (_blocks_exec_printed.count(_subdomain))
      return;

    console << "[DBG] Ordering of User Objects on block " << _subdomain << std::endl;
    // Output specific ordering of objects
    printExecutionOrdering<ElementUserObject>(_element_objs, "element user objects");
    printExecutionOrdering<DomainUserObject>(_domain_objs, "domain user objects");
    if (_fe_problem.currentlyComputingJacobian())
      printExecutionOrdering<ShapeElementUserObject>(
          _shape_element_objs, "element user objects contributing to the Jacobian");
    printExecutionOrdering<SideUserObject>(side_uos, "side user objects");
    if (_fe_problem.currentlyComputingJacobian())
      printExecutionOrdering<ShapeSideUserObject>(shapers,
                                                  "side user objects contributing to the Jacobian");
    printExecutionOrdering<InternalSideUserObject>(_internal_side_objs,
                                                   "internal side user objects");
    printExecutionOrdering<InterfaceUserObject>(interface_objs, "interface user objects");
    console << "[DBG] Only user objects active on local element/sides are executed" << std::endl;
  }
  else if (num_objects == 0 && !_blocks_exec_printed.count(_subdomain))
    console << "[DBG] No User Objects on block " << _subdomain << " on " << execute_on.name()
            << std::endl;

  // Mark subdomain as having printed to avoid printing again
  _blocks_exec_printed.insert(_subdomain);
}

printBoundaryExecutionInformation()

virtual void ThreadedElementLoopBase< ConstElemRange >::printBoundaryExecutionInformation ( const unsigned int  ) const

inlineprotectedvirtualinherited

Print information about the particular ordering of objects on each boundary.

Reimplemented in NonlinearThread.

Definition at line 187 of file ThreadedElementLoopBase.h.

{}

printExecutionOrdering() [1/2]

void ThreadedElementLoop< ConstElemRange >::printExecutionOrdering ( const std::vector< T *> &  objs, const bool  print_header = true, const std::string &  line_prefix = "[DBG]"  ) const

protectedinherited

Routine to output the ordering of objects within a vector of pointers to these objects.

These objects must implement the name() routine, and it must return a string or compatible type.

Template Parameters

T	the object type

Parameters

objs	the vector with all the objects (should be pointers)
objects_type	the name of the type of objects. Defaults to the CPP object name
print_header	whether to print a header about the timing of execution and the type of objects

Definition at line 148 of file ThreadedElementLoop.h.

{
  if (!objs.size())
    return;

  auto & console = _fe_problem.console();
  const auto objects_type = MooseUtils::prettyCppType(objs[0]);
  std::vector<MooseObject *> moose_objs;
  for (auto obj_ptr : objs)
    moose_objs.push_back(dynamic_cast<MooseObject *>(obj_ptr));
  const auto names = ConsoleUtils::mooseObjectVectorToString(moose_objs);

  // Print string with a DBG prefix and with sufficient line breaks
  std::string message = print_header ? "Executing " + objects_type + " on " +
                                           _fe_problem.getCurrentExecuteOnFlag().name() + "\n"
                                     : "";
  message += (print_header ? "Order of execution:\n" : "") + names;
  console << ConsoleUtils::formatString(message, line_prefix) << std::endl;
}

printExecutionOrdering() [2/2]

void ThreadedElementLoop< ConstElemRange >::printExecutionOrdering ( const std::vector< std::shared_ptr< T >> &  objs_ptrs, const bool  print_header = true, const std::string &  line_prefix = "[DBG]"  ) const

protectedinherited

Definition at line 173 of file ThreadedElementLoop.h.

{
  std::vector<T *> regular_ptrs;
  for (auto shared_ptr : objs_ptrs)
    regular_ptrs.push_back(shared_ptr.get());
  printExecutionOrdering<T>(regular_ptrs, print_header, line_prefix);
}

printGeneralExecutionInformation()

void ComputeUserObjectsThread::printGeneralExecutionInformation ( ) const

overrideprotectedvirtual

Print general information about the loop, like the ordering of class of objects.

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 330 of file ComputeUserObjectsThread.C.

{
  if (_fe_problem.shouldPrintExecution(_tid))
  {
    const auto & console = _fe_problem.console();
    const auto & execute_on = _fe_problem.getCurrentExecuteOnFlag();
    console << "[DBG] Computing elemental user objects on " << execute_on << std::endl;
    mooseDoOnce(console << "[DBG] Execution order of objects types on each element then its sides:"
                        << std::endl;
                // onElement
                console << "[DBG] - element user objects" << std::endl;
                console << "[DBG] - domain user objects" << std::endl;
                console << "[DBG] - element user objects contributing to the Jacobian" << std::endl;

                // onBoundary
                console << "[DBG] - side user objects" << std::endl;
                console << "[DBG] - domain user objects executing on sides" << std::endl;
                console << "[DBG] - side user objects contributing to the Jacobian" << std::endl;

                // onInternalSide
                console << "[DBG] - internal side user objects" << std::endl;
                console << "[DBG] - domain user objects executing on internal sides" << std::endl;

                // onInterface
                console << "[DBG] - interface user objects" << std::endl;
                console << "[DBG] - domain user objects executing at interfaces" << std::endl;);
  }
}

printVectorOrdering()

template<typename T >

void ComputeUserObjectsThread::printVectorOrdering ( std::vector< T *>  uos, const std::string &  name  ) const

protected

Format output of vector of UOs.

queryBoundary()

template<typename T >

void ComputeUserObjectsThread::queryBoundary ( Interfaces  iface, BoundaryID  bnd, std::vector< T > &  results  )

inlineprivate

Definition at line 77 of file ComputeUserObjectsThread.h.

Referenced by onBoundary(), onInterface(), and printBlockExecutionInformation().

  {
    _query_boundary.queryInto(results, _tid, std::make_tuple(bnd, false), iface);
  }

querySubdomain()

template<typename T >

void ComputeUserObjectsThread::querySubdomain ( Interfaces  iface, std::vector< T > &  results  )

inlineprivate

Definition at line 72 of file ComputeUserObjectsThread.h.

Referenced by subdomainChanged().

  {
    _query_subdomain.queryInto(results, _tid, _subdomain, iface);
  }

resetExecPrintedSets()

void ThreadedElementLoopBase< ConstElemRange >::resetExecPrintedSets ( ) const

protectedinherited

Resets the set of blocks and boundaries visited.

Definition at line 454 of file ThreadedElementLoopBase.h.

{
  _blocks_exec_printed.clear();
  _boundaries_exec_printed.clear();
}

shouldComputeInternalSide()

bool ThreadedElementLoopBase< ConstElemRange >::shouldComputeInternalSide ( const Elem &  elem, const Elem &  neighbor  ) const

protectedvirtualinherited

Whether to compute the internal side for the provided element-neighbor pair.

Typically this will return true if the element id is less than the neighbor id when the elements are equal level, or when the element is more refined than the neighbor, and then false otherwise. One type of loop where the logic will be different is when projecting stateful material properties

Reimplemented in NonlinearThread, and FlagElementsThread.

Definition at line 429 of file ThreadedElementLoopBase.h.

{
  auto level = [this](const auto & elem_arg)
  {
    if (_mesh.doingPRefinement())
      return elem_arg.p_level();
    else
      return elem_arg.level();
  };
  const auto elem_id = elem.id(), neighbor_id = neighbor.id();
  const auto elem_level = level(elem), neighbor_level = level(neighbor);

  // When looping over elements and then sides, we need to make sure that we do not duplicate
  // effort, e.g. if a face is shared by element 1 and element 2, then we do not want to do compute
  // work both when we are visiting element 1 *and* then later when visiting element 2. Our rule is
  // to only compute when we are visiting the element that has the lower element id when element and
  // neighbor are of the same adaptivity level, and then if they are not of the same level, then
  // we only compute when we are visiting the finer element
  return (neighbor.active() && (neighbor_level == elem_level) && (elem_id < neighbor_id)) ||
         (neighbor_level < elem_level);
}

subdomainChanged()

void ComputeUserObjectsThread::subdomainChanged ( )

overridevirtual

Called every time the current subdomain changes (i.e.

the subdomain of this element is not the same as the subdomain of the last element). Beware of over-using this! You might think that you can do some expensive stuff in here and get away with it... but there are applications that have TONS of subdomains....

Reimplemented from ThreadedElementLoopBase< ConstElemRange >.

Definition at line 52 of file ComputeUserObjectsThread.C.

{
  // for the current thread get block objects for the current subdomain and *all* side objects
  std::vector<UserObject *> objs;
  querySubdomain(Interfaces::ElementUserObject | Interfaces::InternalSideUserObject |
                     Interfaces::InterfaceUserObject | Interfaces::DomainUserObject,
                 objs);

  _query.clone()
      .condition<AttribThread>(_tid)
      .condition<AttribInterfaces>(Interfaces::DomainUserObject)
      .queryInto(_all_domain_objs);

  std::vector<UserObject *> side_objs;
  _query.clone()
      .condition<AttribThread>(_tid)
      .condition<AttribInterfaces>(Interfaces::SideUserObject)
      .queryInto(side_objs);

  objs.insert(objs.begin(), side_objs.begin(), side_objs.end());

  // collect dependencies and run subdomain setup
  _fe_problem.subdomainSetup(_subdomain, _tid);

  std::set<MooseVariableFEBase *> needed_moose_vars;
  std::unordered_set<unsigned int> needed_mat_props;
  std::set<TagID> needed_fe_var_vector_tags;
  for (const auto obj : objs)
  {
    auto v_obj = dynamic_cast<MooseVariableDependencyInterface *>(obj);
    if (v_obj)
    {
      const auto & v_deps = v_obj->getMooseVariableDependencies();
      needed_moose_vars.insert(v_deps.begin(), v_deps.end());
    }

    auto m_obj = dynamic_cast<MaterialPropertyInterface *>(obj);
    if (m_obj)
    {
      auto & m_deps = m_obj->getMatPropDependencies();
      needed_mat_props.insert(m_deps.begin(), m_deps.end());
    }

    auto c_obj = dynamic_cast<Coupleable *>(obj);
    if (c_obj)
    {
      const auto & tag_deps = c_obj->getFEVariableCoupleableVectorTags();
      needed_fe_var_vector_tags.insert(tag_deps.begin(), tag_deps.end());
    }

    obj->subdomainSetup();
  }
  _fe_problem.getMaterialWarehouse().updateBlockFEVariableCoupledVectorTagDependency(
      _subdomain, needed_fe_var_vector_tags, _tid);

  _fe_problem.setActiveElementalMooseVariables(needed_moose_vars, _tid);
  _fe_problem.setActiveFEVariableCoupleableVectorTags(needed_fe_var_vector_tags, _tid);
  _fe_problem.prepareMaterials(needed_mat_props, _subdomain, _tid);

  querySubdomain(Interfaces::InternalSideUserObject, _internal_side_objs);
  querySubdomain(Interfaces::ElementUserObject, _element_objs);
  querySubdomain(Interfaces::ShapeElementUserObject, _shape_element_objs);
  querySubdomain(Interfaces::DomainUserObject, _domain_objs);
}

Member Data Documentation

_all_domain_objs

std::vector<DomainUserObject *> ComputeUserObjectsThread::_all_domain_objs

private

Definition at line 90 of file ComputeUserObjectsThread.h.

Referenced by onInterface(), and subdomainChanged().

_aux_sys

AuxiliarySystem& ComputeUserObjectsThread::_aux_sys

private

Definition at line 92 of file ComputeUserObjectsThread.h.

Referenced by onElement().

_blocks_exec_printed

std::set<SubdomainID> ThreadedElementLoopBase< ConstElemRange >::_blocks_exec_printed

mutableprotectedinherited

Keep track of which blocks were visited.

Definition at line 190 of file ThreadedElementLoopBase.h.

Referenced by ComputeMarkerThread::printBlockExecutionInformation(), ComputeIndicatorThread::printBlockExecutionInformation(), printBlockExecutionInformation(), and NonlinearThread::printBlockExecutionInformation().

_boundaries_exec_printed

std::set<BoundaryID> ThreadedElementLoopBase< ConstElemRange >::_boundaries_exec_printed

mutableprotectedinherited

Keep track of which boundaries were visited.

Definition at line 193 of file ThreadedElementLoopBase.h.

Referenced by NonlinearThread::printBoundaryExecutionInformation().

_domain_objs

std::vector<DomainUserObject *> ComputeUserObjectsThread::_domain_objs

private

Definition at line 89 of file ComputeUserObjectsThread.h.

Referenced by onBoundary(), onElement(), onExternalSide(), onInternalSide(), printBlockExecutionInformation(), and subdomainChanged().

_element_objs

std::vector<ElementUserObject *> ComputeUserObjectsThread::_element_objs

private

Definition at line 87 of file ComputeUserObjectsThread.h.

Referenced by onElement(), printBlockExecutionInformation(), and subdomainChanged().

_fe_problem

FEProblemBase& ThreadedElementLoop< ConstElemRange >::_fe_problem

protectedinherited

Definition at line 62 of file ThreadedElementLoop.h.

Referenced by ComputeResidualThread::accumulate(), ComputeResidualAndJacobianThread::accumulate(), ComputeResidualThread::accumulateLower(), ComputeResidualAndJacobianThread::accumulateLower(), ComputeJacobianThread::accumulateLower(), ComputeResidualThread::accumulateNeighbor(), ComputeResidualAndJacobianThread::accumulateNeighbor(), ComputeJacobianThread::accumulateNeighbor(), ComputeResidualThread::accumulateNeighborLower(), ComputeResidualAndJacobianThread::accumulateNeighborLower(), ComputeJacobianThread::accumulateNeighborLower(), ComputeJacobianThread::compute(), ComputeFullJacobianThread::computeOnBoundary(), ComputeFullJacobianThread::computeOnElement(), NonlinearThread::computeOnElement(), ComputeFullJacobianThread::computeOnInterface(), ComputeFullJacobianThread::computeOnInternalFace(), ComputeResidualThread::determineObjectWarehouses(), ComputeJacobianThread::determineObjectWarehouses(), ComputeResidualAndJacobianThread::determineObjectWarehouses(), NonlinearThread::onBoundary(), onBoundary(), ComputeElemDampingThread::onElement(), NonlinearThread::onElement(), onElement(), NonlinearThread::onInterface(), onInterface(), NonlinearThread::onInternalSide(), onInternalSide(), post(), ComputeJacobianThread::postElement(), ComputeJacobianBlocksThread::postElement(), ComputeJacobianBlocksThread::postInternalSide(), NonlinearThread::prepareFace(), printBlockExecutionInformation(), NonlinearThread::printBlockExecutionInformation(), NonlinearThread::printBoundaryExecutionInformation(), ComputeElemDampingThread::printGeneralExecutionInformation(), printGeneralExecutionInformation(), NonlinearThread::printGeneralExecutionInformation(), NonlinearThread::subdomainChanged(), and subdomainChanged().

_interface_user_objects

std::vector<InterfaceUserObject *> ComputeUserObjectsThread::_interface_user_objects

private

Definition at line 86 of file ComputeUserObjectsThread.h.

_internal_side_objs

std::vector<InternalSideUserObject *> ComputeUserObjectsThread::_internal_side_objs

private

Definition at line 85 of file ComputeUserObjectsThread.h.

Referenced by onInternalSide(), printBlockExecutionInformation(), and subdomainChanged().

_mesh

MooseMesh& ThreadedElementLoopBase< ConstElemRange >::_mesh

protectedinherited

Definition at line 165 of file ThreadedElementLoopBase.h.

Referenced by CacheChangedListsThread::onElement(), NonlinearThread::printBlockExecutionInformation(), and NonlinearThread::printBoundaryExecutionInformation().

_neighbor_subdomain

SubdomainID ThreadedElementLoopBase< ConstElemRange >::_neighbor_subdomain

protectedinherited

The subdomain for the current neighbor.

Definition at line 175 of file ThreadedElementLoopBase.h.

Referenced by ComputeResidualThread::computeOnInternalFace(), ComputeFullJacobianThread::computeOnInternalFace(), and ComputeJacobianForScalingThread::operator()().

_old_neighbor_subdomain

SubdomainID ThreadedElementLoopBase< ConstElemRange >::_old_neighbor_subdomain

protectedinherited

The subdomain for the last neighbor.

Definition at line 178 of file ThreadedElementLoopBase.h.

_old_subdomain

SubdomainID ThreadedElementLoopBase< ConstElemRange >::_old_subdomain

protectedinherited

The subdomain for the last element.

Definition at line 172 of file ThreadedElementLoopBase.h.

Referenced by ComputeJacobianForScalingThread::operator()().

_query

const TheWarehouse::Query ComputeUserObjectsThread::_query

private

Definition at line 82 of file ComputeUserObjectsThread.h.

Referenced by subdomainChanged().

_query_boundary

TheWarehouse::QueryCache<AttribThread, AttribBoundaries, AttribInterfaces> ComputeUserObjectsThread::_query_boundary

private

Definition at line 84 of file ComputeUserObjectsThread.h.

Referenced by queryBoundary().

_query_subdomain

TheWarehouse::QueryCache<AttribThread, AttribSubdomains, AttribInterfaces> ComputeUserObjectsThread::_query_subdomain

private

Definition at line 83 of file ComputeUserObjectsThread.h.

Referenced by querySubdomain().

_shape_element_objs

std::vector<ShapeElementUserObject *> ComputeUserObjectsThread::_shape_element_objs

private

Definition at line 88 of file ComputeUserObjectsThread.h.

Referenced by onElement(), printBlockExecutionInformation(), and subdomainChanged().

_subdomain

SubdomainID ThreadedElementLoopBase< ConstElemRange >::_subdomain

protectedinherited

The subdomain for the current element.

Definition at line 169 of file ThreadedElementLoopBase.h.

Referenced by ComputeFullJacobianThread::computeOnBoundary(), ComputeFullJacobianThread::computeOnElement(), NonlinearThread::computeOnElement(), ComputeResidualThread::computeOnInternalFace(), ComputeFullJacobianThread::computeOnInternalFace(), ComputeResidualAndJacobianThread::computeOnInternalFace(), ComputeJacobianThread::computeOnInternalFace(), NonlinearThread::computeOnInternalFace(), ComputeResidualThread::determineObjectWarehouses(), ComputeResidualAndJacobianThread::determineObjectWarehouses(), ComputeMaterialsObjectThread::onBoundary(), ComputeMarkerThread::onElement(), ComputeIndicatorThread::onElement(), ComputeMaterialsObjectThread::onElement(), onElement(), ComputeMaterialsObjectThread::onInterface(), ComputeMaterialsObjectThread::onInternalSide(), ComputeJacobianForScalingThread::operator()(), ComputeJacobianBlocksThread::postInternalSide(), ComputeMarkerThread::printBlockExecutionInformation(), ComputeIndicatorThread::printBlockExecutionInformation(), printBlockExecutionInformation(), NonlinearThread::printBlockExecutionInformation(), querySubdomain(), ComputeMarkerThread::subdomainChanged(), ComputeMaterialsObjectThread::subdomainChanged(), ComputeIndicatorThread::subdomainChanged(), NonlinearThread::subdomainChanged(), and subdomainChanged().

_tid

THREAD_ID ThreadedElementLoopBase< ConstElemRange >::_tid

protectedinherited

Definition at line 166 of file ThreadedElementLoopBase.h.

Referenced by ComputeResidualThread::accumulate(), ComputeResidualAndJacobianThread::accumulate(), ComputeResidualThread::accumulateLower(), ComputeResidualAndJacobianThread::accumulateLower(), ComputeJacobianThread::accumulateLower(), ComputeResidualThread::accumulateNeighbor(), ComputeResidualAndJacobianThread::accumulateNeighbor(), ComputeJacobianThread::accumulateNeighbor(), ComputeResidualThread::accumulateNeighborLower(), ComputeResidualAndJacobianThread::accumulateNeighborLower(), ComputeJacobianThread::accumulateNeighborLower(), ComputeFullJacobianThread::computeOnBoundary(), NonlinearThread::computeOnBoundary(), ComputeFullJacobianThread::computeOnElement(), NonlinearThread::computeOnElement(), ComputeFullJacobianThread::computeOnInterface(), NonlinearThread::computeOnInterface(), ComputeResidualThread::computeOnInternalFace(), ComputeFullJacobianThread::computeOnInternalFace(), ComputeResidualAndJacobianThread::computeOnInternalFace(), ComputeJacobianThread::computeOnInternalFace(), NonlinearThread::computeOnInternalFace(), ComputeResidualThread::determineObjectWarehouses(), ComputeJacobianThread::determineObjectWarehouses(), ComputeResidualAndJacobianThread::determineObjectWarehouses(), ComputeMaterialsObjectThread::onBoundary(), NonlinearThread::onBoundary(), onBoundary(), ComputeMarkerThread::onElement(), ComputeElemDampingThread::onElement(), ComputeMaterialsObjectThread::onElement(), ComputeIndicatorThread::onElement(), NonlinearThread::onElement(), onElement(), ComputeMaterialsObjectThread::onInterface(), NonlinearThread::onInterface(), onInterface(), ComputeMaterialsObjectThread::onInternalSide(), ComputeIndicatorThread::onInternalSide(), NonlinearThread::onInternalSide(), onInternalSide(), ComputeJacobianForScalingThread::operator()(), ComputeMaterialsObjectThread::post(), ComputeMarkerThread::post(), ComputeIndicatorThread::post(), post(), ComputeJacobianThread::postElement(), ComputeJacobianBlocksThread::postElement(), ComputeJacobianBlocksThread::postInternalSide(), NonlinearThread::prepareFace(), ComputeMarkerThread::printBlockExecutionInformation(), ComputeIndicatorThread::printBlockExecutionInformation(), printBlockExecutionInformation(), NonlinearThread::printBlockExecutionInformation(), NonlinearThread::printBoundaryExecutionInformation(), ComputeElemDampingThread::printGeneralExecutionInformation(), ComputeMarkerThread::printGeneralExecutionInformation(), ComputeIndicatorThread::printGeneralExecutionInformation(), printGeneralExecutionInformation(), NonlinearThread::printGeneralExecutionInformation(), queryBoundary(), querySubdomain(), ComputeMarkerThread::subdomainChanged(), ComputeMaterialsObjectThread::subdomainChanged(), ComputeIndicatorThread::subdomainChanged(), NonlinearThread::subdomainChanged(), and subdomainChanged().

---

The documentation for this class was generated from the following files:

• include/loops/ComputeUserObjectsThread.h
• src/loops/ComputeUserObjectsThread.C