Computes Advective fluxes for a constant velocity. More...

#include <AdvectiveFluxCalculatorConstantVelocity.h>

Inheritance diagram for AdvectiveFluxCalculatorConstantVelocity:

Public Member Functions
	AdvectiveFluxCalculatorConstantVelocity (const InputParameters &parameters)

virtual void	timestepSetup () override

virtual void	meshChanged () override

virtual void	initialize () override

virtual void	threadJoin (const UserObject &uo) override

virtual void	finalize () override

virtual void	execute () override

virtual void	executeOnElement (dof_id_type global_i, dof_id_type global_j, unsigned local_i, unsigned local_j, unsigned qp)
	This is called by multiple times in execute() in a double loop over _current_elem's nodes (local_i and local_j) nested in a loop over each of _current_elem's quadpoints (qp). More...

Real	getFluxOut (dof_id_type node_i) const
	Returns the flux out of lobal node id. More...

const std::map< dof_id_type, Real > &	getdFluxOutdu (dof_id_type node_i) const
	Returns r where r[j] = d(flux out of global node i)/du(global node j) used in Jacobian computations. More...

const std::vector< std::vector< Real > > &	getdFluxOutdKjk (dof_id_type node_i) const
	Returns r where r[j][k] = d(flux out of global node i)/dK[connected node j][connected node k] used in Jacobian computations. More...

unsigned	getValence (dof_id_type node_i) const
	Returns the valence of the global node i Valence is the number of times the node is encountered in a loop over elements (that have appropriate subdomain_id, if the user has employed the "blocks=" parameter) seen by this processor (including ghosted elements) More...

Protected Types
enum	FluxLimiterTypeEnum { FluxLimiterTypeEnum::MinMod, FluxLimiterTypeEnum::VanLeer, FluxLimiterTypeEnum::MC, FluxLimiterTypeEnum::superbee, FluxLimiterTypeEnum::None }
	Determines Flux Limiter type (Page 135 of Kuzmin and Turek) "None" means that limitFlux=0 always, which implies zero antidiffusion will be added. More...

enum	PQPlusMinusEnum { PQPlusMinusEnum::PPlus, PQPlusMinusEnum::PMinus, PQPlusMinusEnum::QPlus, PQPlusMinusEnum::QMinus }
	Signals to the PQPlusMinus method what should be computed. More...

Protected Member Functions
virtual Real	computeVelocity (unsigned i, unsigned j, unsigned qp) const override
	Computes the transfer velocity between current node i and current node j at the current qp in the current element (_current_elem). More...

virtual Real	computeU (unsigned i) const override
	Computes the value of u at the local node id of the current element (_current_elem) More...

virtual void	buildCommLists ()
	When using multiple processors, other processors will compute: More...

virtual void	exchangeGhostedInfo ()
	Sends and receives multi-processor information regarding u_nodal and k_ij. More...

void	limitFlux (Real a, Real b, Real &limited, Real &dlimited_db) const
	flux limiter, L, on Page 135 of Kuzmin and Turek More...

Real	rPlus (dof_id_type sequential_i, std::vector< Real > &dlimited_du, std::vector< Real > &dlimited_dk) const
	Returns the value of R_{i}^{+}, Eqn (49) of KT. More...

Real	rMinus (dof_id_type sequential_i, std::vector< Real > &dlimited_du, std::vector< Real > &dlimited_dk) const
	Returns the value of R_{i}^{-}, Eqn (49) of KT. More...

Real	PQPlusMinus (dof_id_type sequential_i, const PQPlusMinusEnum pq_plus_minus, std::vector< Real > &derivs, std::vector< Real > &dpq_dk) const
	Returns the value of P_{i}^{+}, P_{i}^{-}, Q_{i}^{+} or Q_{i}^{-} (depending on pq_plus_minus) which are defined in Eqns (47) and (48) of KT. More...

void	zeroedConnection (std::map< dof_id_type, Real > &the_map, dof_id_type node_i) const
	Clears the_map, then, using _kij, constructs the_map so that the_map[node_id] = 0.0 for all node_id connected with node_i. More...

Protected Attributes
RealVectorValue	_velocity
	advection velocity More...

const VariableValue &	_u_at_nodes
	the nodal values of u More...

const VariablePhiValue &	_phi
	Kuzmin-Turek shape function. More...

const VariablePhiGradient &	_grad_phi
	grad(Kuzmin-Turek shape function) More...

bool	_resizing_needed
	whether _kij, etc, need to be sized appropriately (and valence recomputed) at the start of the timestep More...

enum AdvectiveFluxCalculatorBase::FluxLimiterTypeEnum	_flux_limiter_type

std::vector< std::vector< Real > >	_kij
	Kuzmin-Turek K_ij matrix. More...

std::vector< Real >	_flux_out
	_flux_out[i] = flux of "heat" from sequential node i More...

std::vector< std::map< dof_id_type, Real > >	_dflux_out_du
	_dflux_out_du[i][j] = d(flux_out[i])/d(u[j]). More...

std::vector< std::vector< std::vector< Real > > >	_dflux_out_dKjk
	_dflux_out_dKjk[sequential_i][j][k] = d(flux_out[sequential_i])/d(K[j][k]). More...

std::vector< unsigned >	_valence
	_valence[i] = number of times, in a loop over elements seen by this processor (viz, including ghost elements) and are part of the block-restricted blocks of this UserObject, that the sequential node i is encountered More...

std::vector< Real >	_u_nodal
	_u_nodal[i] = value of _u at sequential node number i More...

std::vector< bool >	_u_nodal_computed_by_thread
	_u_nodal_computed_by_thread(i) = true if _u_nodal[i] has been computed in execute() by the thread on this processor More...

PorousFlowConnectedNodes	_connections
	Holds the sequential and global nodal IDs, and info regarding mesh connections between them. More...

std::size_t	_number_of_nodes
	Number of nodes held by the _connections object. More...

processor_id_type	_my_pid
	processor ID of this object More...

std::map< processor_id_type, std::vector< dof_id_type > >	_nodes_to_receive
	_nodes_to_receive[proc_id] = list of sequential nodal IDs. More...

std::map< processor_id_type, std::vector< dof_id_type > >	_nodes_to_send
	_nodes_to_send[proc_id] = list of sequential nodal IDs. More...

std::map< processor_id_type, std::vector< std::pair< dof_id_type, dof_id_type > > >	_pairs_to_receive
	_pairs_to_receive[proc_id] indicates the k(i, j) pairs that will be sent to us from proc_id _pairs_to_receive is first built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs and the index of connections is performed, for efficiency. More...

std::map< processor_id_type, std::vector< std::pair< dof_id_type, dof_id_type > > >	_pairs_to_send
	_pairs_to_send[proc_id] indicates the k(i, j) pairs that we will send to proc_id _pairs_to_send is first built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs and the index of connections is performed, for efficiency. More...

const Real	_allowable_MB_wastage
	A mooseWarning is issued if mb_wasted = (_connections.sizeSequential() - _connections.numNodes()) * 4 / 1048576 > _allowable_MB_wastage. More...

std::vector< std::vector< Real > >	_dij
	Vectors used in finalize() More...

std::vector< std::vector< Real > >	_dDij_dKij
	dDij_dKij[i][j] = d(D[i][j])/d(K[i][j]) for i!=j More...

std::vector< std::vector< Real > >	_dDij_dKji
	dDij_dKji[i][j] = d(D[i][j])/d(K[j][i]) for i!=j More...

std::vector< std::vector< Real > >	_dDii_dKij
	dDii_dKij[i][j] = d(D[i][i])/d(K[i][j]) More...

std::vector< std::vector< Real > >	_dDii_dKji
	dDii_dKji[i][j] = d(D[i][i])/d(K[j][i]) More...

std::vector< std::vector< Real > >	_lij

std::vector< Real >	_rP

std::vector< Real >	_rM

std::vector< std::vector< Real > >	_drP
	drP[i][j] = d(rP[i])/d(u[j]). Here j indexes the j^th node connected to i More...

std::vector< std::vector< Real > >	_drM
	drM[i][j] = d(rM[i])/d(u[j]). Here j indexes the j^th node connected to i More...

std::vector< std::vector< Real > >	_drP_dk
	drP_dk[i][j] = d(rP[i])/d(K[i][j]). Here j indexes the j^th node connected to i More...

std::vector< std::vector< Real > >	_drM_dk
	drM_dk[i][j] = d(rM[i])/d(K[i][j]). Here j indexes the j^th node connected to i More...

std::vector< std::vector< Real > >	_fa
	fa[sequential_i][j] sequential_j is the j^th connection to sequential_i More...

std::vector< std::vector< std::map< dof_id_type, Real > > >	_dfa
	dfa[sequential_i][j][global_k] = d(fa[sequential_i][j])/du[global_k]. More...

std::vector< std::vector< std::vector< Real > > >	_dFij_dKik
	dFij_dKik[sequential_i][j][k] = d(fa[sequential_i][j])/d(K[sequential_i][k]). More...

std::vector< std::vector< std::vector< Real > > >	_dFij_dKjk
	dFij_dKjk[sequential_i][j][k] = d(fa[sequential_i][j])/d(K[sequential_j][k]). More...

Detailed Description

Computes Advective fluxes for a constant velocity.

Definition at line 22 of file AdvectiveFluxCalculatorConstantVelocity.h.

Member Enumeration Documentation

◆ FluxLimiterTypeEnum

enum AdvectiveFluxCalculatorBase::FluxLimiterTypeEnum

strongprotectedinherited

Determines Flux Limiter type (Page 135 of Kuzmin and Turek) "None" means that limitFlux=0 always, which implies zero antidiffusion will be added.

Enumerator
MinMod
VanLeer
MC
superbee
None

Definition at line 169 of file AdvectiveFluxCalculatorBase.h.

169 { MinMod, VanLeer, MC, superbee, None } _flux_limiter_type;

◆ PQPlusMinusEnum

enum AdvectiveFluxCalculatorBase::PQPlusMinusEnum

strongprotectedinherited

Signals to the PQPlusMinus method what should be computed.

Enumerator
PPlus
PMinus
QPlus
QMinus

Definition at line 255 of file AdvectiveFluxCalculatorBase.h.

   {
     PPlus,
     PMinus,
     QPlus,
     QMinus
   };

Constructor & Destructor Documentation

◆ AdvectiveFluxCalculatorConstantVelocity()

AdvectiveFluxCalculatorConstantVelocity::AdvectiveFluxCalculatorConstantVelocity ( const InputParameters & parameters )

Definition at line 29 of file AdvectiveFluxCalculatorConstantVelocity.C.

   : AdvectiveFluxCalculatorBase(parameters),
     _velocity(getParam<RealVectorValue>("velocity")),
     _u_at_nodes(coupledDofValues("u")),
     _phi(_assembly.fePhi<Real>(getVar("u", 0)->feType())),
     _grad_phi(_assembly.feGradPhi<Real>(getVar("u", 0)->feType()))
 {
 }

Member Function Documentation

◆ buildCommLists()

void AdvectiveFluxCalculatorBase::buildCommLists ( )

protectedvirtualinherited

When using multiple processors, other processors will compute:

u_nodal for nodes that we don't "own", but that we need when doing the stabilization
k_ij for node pairs that we don't "own", and contributions to node pairs that we do "own" (on the boundary of our set of elements), that are used in the stabilization This method builds _nodes_to_receive and _pairs_to_receive that describe which processors we are going to receive this info from, and similarly it builds _nodes_to_send and _pairs_to_send.

Build the multi-processor communication lists.

(A) We will have to send _u_nodal information to other processors. This is because although we can Evaluate Variables at all elements in _fe_problem.getEvaluableElementRange(), in the PorousFlow setting _u_nodal could depend on Material Properties within the elements, and we can't access those Properties within the ghosted elements.

(B) In a similar way, we need to send _kij information to other processors. A similar strategy is followed

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 850 of file AdvectiveFluxCalculatorBase.C.

 {
   _nodes_to_receive.clear();
   for (const auto & elem : _fe_problem.getEvaluableElementRange())
     if (this->hasBlocks(elem->subdomain_id()))
     {
       const processor_id_type elem_pid = elem->processor_id();
       if (elem_pid != _my_pid)
       {
         if (_nodes_to_receive.find(elem_pid) == _nodes_to_receive.end())
           _nodes_to_receive[elem_pid] = std::vector<dof_id_type>();
         for (unsigned i = 0; i < elem->n_nodes(); ++i)
           if (std::find(_nodes_to_receive[elem_pid].begin(),
                         _nodes_to_receive[elem_pid].end(),
                         elem->node_id(i)) == _nodes_to_receive[elem_pid].end())
             _nodes_to_receive[elem_pid].push_back(elem->node_id(i));
       }
     }
  
   // exchange this info with other processors, building _nodes_to_send at the same time
   _nodes_to_send.clear();
   auto nodes_action_functor = [this](processor_id_type pid, const std::vector<dof_id_type> & nts) {
     _nodes_to_send[pid] = nts;
   };
   Parallel::push_parallel_vector_data(this->comm(), _nodes_to_receive, nodes_action_functor);
  
   // At the moment,  _nodes_to_send and _nodes_to_receive contain global node numbers
   // It is slightly more efficient to convert to sequential node numbers
   // so that we don't have to keep doing things like
   // _u_nodal[_connections.sequentialID(_nodes_to_send[pid][i])
   // every time we send/receive
   for (auto & kv : _nodes_to_send)
   {
     const processor_id_type pid = kv.first;
     const std::size_t num_nodes = kv.second.size();
     for (unsigned i = 0; i < num_nodes; ++i)
       _nodes_to_send[pid][i] = _connections.sequentialID(_nodes_to_send[pid][i]);
   }
   for (auto & kv : _nodes_to_receive)
   {
     const processor_id_type pid = kv.first;
     const std::size_t num_nodes = kv.second.size();
     for (unsigned i = 0; i < num_nodes; ++i)
       _nodes_to_receive[pid][i] = _connections.sequentialID(_nodes_to_receive[pid][i]);
   }
  
   // Build pairs_to_receive
   _pairs_to_receive.clear();
   for (const auto & elem : _fe_problem.getEvaluableElementRange())
     if (this->hasBlocks(elem->subdomain_id()))
     {
       const processor_id_type elem_pid = elem->processor_id();
       if (elem_pid != _my_pid)
       {
         if (_pairs_to_receive.find(elem_pid) == _pairs_to_receive.end())
           _pairs_to_receive[elem_pid] = std::vector<std::pair<dof_id_type, dof_id_type>>();
         for (unsigned i = 0; i < elem->n_nodes(); ++i)
           for (unsigned j = 0; j < elem->n_nodes(); ++j)
           {
             std::pair<dof_id_type, dof_id_type> the_pair(elem->node_id(i), elem->node_id(j));
             if (std::find(_pairs_to_receive[elem_pid].begin(),
                           _pairs_to_receive[elem_pid].end(),
                           the_pair) == _pairs_to_receive[elem_pid].end())
               _pairs_to_receive[elem_pid].push_back(the_pair);
           }
       }
     }
  
   _pairs_to_send.clear();
   auto pairs_action_functor = [this](processor_id_type pid,
                                      const std::vector<std::pair<dof_id_type, dof_id_type>> & pts) {
     _pairs_to_send[pid] = pts;
   };
   Parallel::push_parallel_vector_data(this->comm(), _pairs_to_receive, pairs_action_functor);
  
   // _pairs_to_send and _pairs_to_receive have been built using global node IDs
   // since all processors know about that.  However, using global IDs means
   // that every time we send/receive, we keep having to do things like
   // _kij[_connections.sequentialID(_pairs_to_send[pid][i].first)][_connections.indexOfGlobalConnection(_pairs_to_send[pid][i].first,
   // _pairs_to_send[pid][i].second)] which is quite inefficient.  So:
   for (auto & kv : _pairs_to_send)
   {
     const processor_id_type pid = kv.first;
     const std::size_t num_pairs = kv.second.size();
     for (unsigned i = 0; i < num_pairs; ++i)
     {
       _pairs_to_send[pid][i].second = _connections.indexOfGlobalConnection(
           _pairs_to_send[pid][i].first, _pairs_to_send[pid][i].second);
       _pairs_to_send[pid][i].first = _connections.sequentialID(_pairs_to_send[pid][i].first);
     }
   }
   for (auto & kv : _pairs_to_receive)
   {
     const processor_id_type pid = kv.first;
     const std::size_t num_pairs = kv.second.size();
     for (unsigned i = 0; i < num_pairs; ++i)
     {
       _pairs_to_receive[pid][i].second = _connections.indexOfGlobalConnection(
           _pairs_to_receive[pid][i].first, _pairs_to_receive[pid][i].second);
       _pairs_to_receive[pid][i].first = _connections.sequentialID(_pairs_to_receive[pid][i].first);
     }
   }
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::buildCommLists(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ computeU()

Real AdvectiveFluxCalculatorConstantVelocity::computeU ( unsigned i ) const

overrideprotectedvirtual

Computes the value of u at the local node id of the current element (_current_elem)

Parameters

i	local node id of the current element

Implements AdvectiveFluxCalculatorBase.

Definition at line 46 of file AdvectiveFluxCalculatorConstantVelocity.C.

 {
   return _u_at_nodes[i];
 }

◆ computeVelocity()

Real AdvectiveFluxCalculatorConstantVelocity::computeVelocity	(	unsigned	i,
		unsigned	j,
		unsigned	qp
	)		const

overrideprotectedvirtual

Computes the transfer velocity between current node i and current node j at the current qp in the current element (_current_elem).

For instance, (_grad_phi[i][qp] * _velocity) * _phi[j][qp];

Parameters

i	node number in the current element
j	node number in the current element
qp	quadpoint number in the current element

Implements AdvectiveFluxCalculatorBase.

Definition at line 40 of file AdvectiveFluxCalculatorConstantVelocity.C.

 {
   return (_grad_phi[i][qp] * _velocity) * _phi[j][qp];
 }

◆ exchangeGhostedInfo()

void AdvectiveFluxCalculatorBase::exchangeGhostedInfo ( )

protectedvirtualinherited

Sends and receives multi-processor information regarding u_nodal and k_ij.

See buildCommLists for some more explanation.

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 968 of file AdvectiveFluxCalculatorBase.C.

 {
   // Exchange ghosted _u_nodal information with other processors
   std::map<processor_id_type, std::vector<Real>> unodal_to_send;
   for (const auto & kv : _nodes_to_send)
   {
     const processor_id_type pid = kv.first;
     unodal_to_send[pid] = std::vector<Real>();
     for (const auto & nd : kv.second)
       unodal_to_send[pid].push_back(_u_nodal[nd]);
   }
  
   auto unodal_action_functor = [this](processor_id_type pid,
                                       const std::vector<Real> & unodal_received) {
     const std::size_t msg_size = unodal_received.size();
     mooseAssert(msg_size == _nodes_to_receive[pid].size(),
                 "Message size, " << msg_size
                                  << ", incompatible with nodes_to_receive, which has size "
                                  << _nodes_to_receive[pid].size());
     for (unsigned i = 0; i < msg_size; ++i)
       _u_nodal[_nodes_to_receive[pid][i]] = unodal_received[i];
   };
   Parallel::push_parallel_vector_data(this->comm(), unodal_to_send, unodal_action_functor);
  
   // Exchange _kij information with other processors
   std::map<processor_id_type, std::vector<Real>> kij_to_send;
   for (const auto & kv : _pairs_to_send)
   {
     const processor_id_type pid = kv.first;
     kij_to_send[pid] = std::vector<Real>();
     for (const auto & pr : kv.second)
       kij_to_send[pid].push_back(_kij[pr.first][pr.second]);
   }
  
   auto kij_action_functor = [this](processor_id_type pid, const std::vector<Real> & kij_received) {
     const std::size_t msg_size = kij_received.size();
     mooseAssert(msg_size == _pairs_to_receive[pid].size(),
                 "Message size, " << msg_size
                                  << ", incompatible with pairs_to_receive, which has size "
                                  << _pairs_to_receive[pid].size());
     for (unsigned i = 0; i < msg_size; ++i)
       _kij[_pairs_to_receive[pid][i].first][_pairs_to_receive[pid][i].second] += kij_received[i];
   };
   Parallel::push_parallel_vector_data(this->comm(), kij_to_send, kij_action_functor);
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::exchangeGhostedInfo(), and AdvectiveFluxCalculatorBase::finalize().

◆ execute()

void AdvectiveFluxCalculatorBase::execute ( )

overridevirtualinherited

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 218 of file AdvectiveFluxCalculatorBase.C.

 {
   // compute _kij contributions from this element that is local to this processor
   // and record _u_nodal
   for (unsigned i = 0; i < _current_elem->n_nodes(); ++i)
   {
     const dof_id_type node_i = _current_elem->node_id(i);
     const dof_id_type sequential_i = _connections.sequentialID(node_i);
     if (!_u_nodal_computed_by_thread[sequential_i])
     {
       _u_nodal[sequential_i] = computeU(i);
       _u_nodal_computed_by_thread[sequential_i] = true;
     }
     for (unsigned j = 0; j < _current_elem->n_nodes(); ++j)
     {
       const dof_id_type node_j = _current_elem->node_id(j);
       for (unsigned qp = 0; qp < _qrule->n_points(); ++qp)
         executeOnElement(node_i, node_j, i, j, qp);
     }
   }
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::execute().

◆ executeOnElement()

void AdvectiveFluxCalculatorBase::executeOnElement	(	dof_id_type	global_i,
		dof_id_type	global_j,
		unsigned	local_i,
		unsigned	local_j,
		unsigned	qp
	)

virtualinherited

This is called by multiple times in execute() in a double loop over _current_elem's nodes (local_i and local_j) nested in a loop over each of _current_elem's quadpoints (qp).

It is used to compute _kij and its derivatives

Parameters

global_i	global node id corresponding to the local node local_i
global_j	global node id corresponding to the local node local_j
local_i	local node number of the _current_elem
local_j	local node number of the _current_elem
qp	quadpoint number of the _current_elem

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 241 of file AdvectiveFluxCalculatorBase.C.

 {
   // KT Eqn (20)
   const dof_id_type sequential_i = _connections.sequentialID(global_i);
   const unsigned index_i_to_j = _connections.indexOfGlobalConnection(global_i, global_j);
   _kij[sequential_i][index_i_to_j] += _JxW[qp] * _coord[qp] * computeVelocity(local_i, local_j, qp);
 }

Referenced by AdvectiveFluxCalculatorBase::execute(), and PorousFlowAdvectiveFluxCalculatorBase::executeOnElement().

◆ finalize()

void AdvectiveFluxCalculatorBase::finalize ( )

overridevirtualinherited

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 270 of file AdvectiveFluxCalculatorBase.C.

 {
   // Overall: ensure _kij is fully built, then compute Kuzmin-Turek D, L, f^a and
   // relevant Jacobian information, and then the relevant quantities into _flux_out and
   // _dflux_out_du, _dflux_out_dKjk
  
   if (_app.n_processors() > 1)
     exchangeGhostedInfo();
  
   // Calculate KuzminTurek D matrix
   // See Eqn (32)
   // This adds artificial diffusion, which eliminates any spurious oscillations
   // The idea is that D will remove all negative off-diagonal elements when it is added to K
   // This is identical to full upwinding
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const std::size_t num_con_i = con_i.size();
     _dij[sequential_i].assign(num_con_i, 0.0);
     _dDij_dKij[sequential_i].assign(num_con_i, 0.0);
     _dDij_dKji[sequential_i].assign(num_con_i, 0.0);
     _dDii_dKij[sequential_i].assign(num_con_i, 0.0);
     _dDii_dKji[sequential_i].assign(num_con_i, 0.0);
     const unsigned index_i_to_i =
         _connections.indexOfSequentialConnection(sequential_i, sequential_i);
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       const dof_id_type sequential_j = con_i[j];
       if (sequential_i == sequential_j)
         continue;
       const unsigned index_j_to_i =
           _connections.indexOfSequentialConnection(sequential_j, sequential_i);
       const Real kij = _kij[sequential_i][j];
       const Real kji = _kij[sequential_j][index_j_to_i];
       if ((kij <= kji) && (kij < 0.0))
       {
         _dij[sequential_i][j] = -kij;
         _dDij_dKij[sequential_i][j] = -1.0;
         _dDii_dKij[sequential_i][j] += 1.0;
       }
       else if ((kji <= kij) && (kji < 0.0))
       {
         _dij[sequential_i][j] = -kji;
         _dDij_dKji[sequential_i][j] = -1.0;
         _dDii_dKji[sequential_i][j] += 1.0;
       }
       else
         _dij[sequential_i][j] = 0.0;
       _dij[sequential_i][index_i_to_i] -= _dij[sequential_i][j];
     }
   }
  
   // Calculate KuzminTurek L matrix
   // See Fig 2: L = K + D
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const std::size_t num_con_i = con_i.size();
     _lij[sequential_i].assign(num_con_i, 0.0);
     for (std::size_t j = 0; j < num_con_i; ++j)
       _lij[sequential_i][j] = _kij[sequential_i][j] + _dij[sequential_i][j];
   }
  
   // Compute KuzminTurek R matrices
   // See Eqns (49) and (12)
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     _rP[sequential_i] = rPlus(sequential_i, _drP[sequential_i], _drP_dk[sequential_i]);
     _rM[sequential_i] = rMinus(sequential_i, _drM[sequential_i], _drM_dk[sequential_i]);
   }
  
   // Calculate KuzminTurek f^{a} matrix
   // This is the antidiffusive flux
   // See Eqn (50)
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.globalConnectionsToSequentialID(sequential_i);
     const unsigned num_con_i = con_i.size();
     _fa[sequential_i].assign(num_con_i, 0.0);
     _dfa[sequential_i].resize(num_con_i);
     _dFij_dKik[sequential_i].resize(num_con_i);
     _dFij_dKjk[sequential_i].resize(num_con_i);
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       for (const auto & global_k : con_i)
         _dfa[sequential_i][j][global_k] = 0;
       const dof_id_type global_j = con_i[j];
       const std::vector<dof_id_type> & con_j = _connections.globalConnectionsToGlobalID(global_j);
       const unsigned num_con_j = con_j.size();
       for (const auto & global_k : con_j)
         _dfa[sequential_i][j][global_k] = 0;
       _dFij_dKik[sequential_i][j].assign(num_con_i, 0.0);
       _dFij_dKjk[sequential_i][j].assign(num_con_j, 0.0);
     }
   }
  
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const dof_id_type global_i = _connections.globalID(sequential_i);
     const Real u_i = _u_nodal[sequential_i];
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const std::size_t num_con_i = con_i.size();
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       const dof_id_type sequential_j = con_i[j];
       if (sequential_i == sequential_j)
         continue;
       const unsigned index_j_to_i =
           _connections.indexOfSequentialConnection(sequential_j, sequential_i);
       const Real Lij = _lij[sequential_i][j];
       const Real Lji = _lij[sequential_j][index_j_to_i];
       if (Lji >= Lij) // node i is upwind of node j.
       {
         const Real Dij = _dij[sequential_i][j];
         const dof_id_type global_j = _connections.globalID(sequential_j);
         const Real u_j = _u_nodal[sequential_j];
         Real prefactor = 0.0;
         std::vector<Real> dprefactor_du(num_con_i,
                                         0.0); // dprefactor_du[j] = d(prefactor)/du[sequential_j]
         std::vector<Real> dprefactor_dKij(
             num_con_i, 0.0);      // dprefactor_dKij[j] = d(prefactor)/dKij[sequential_i][j].
         Real dprefactor_dKji = 0; // dprefactor_dKji = d(prefactor)/dKij[sequential_j][index_j_to_i]
         if (u_i >= u_j)
         {
           if (Lji <= _rP[sequential_i] * Dij)
           {
             prefactor = Lji;
             dprefactor_dKij[j] += _dDij_dKji[sequential_j][index_j_to_i];
             dprefactor_dKji += 1.0 + _dDij_dKij[sequential_j][index_j_to_i];
           }
           else
           {
             prefactor = _rP[sequential_i] * Dij;
             for (std::size_t ind_j = 0; ind_j < num_con_i; ++ind_j)
               dprefactor_du[ind_j] = _drP[sequential_i][ind_j] * Dij;
             dprefactor_dKij[j] += _rP[sequential_i] * _dDij_dKij[sequential_i][j];
             dprefactor_dKji += _rP[sequential_i] * _dDij_dKji[sequential_i][j];
             for (std::size_t ind_j = 0; ind_j < num_con_i; ++ind_j)
               dprefactor_dKij[ind_j] += _drP_dk[sequential_i][ind_j] * Dij;
           }
         }
         else
         {
           if (Lji <= _rM[sequential_i] * Dij)
           {
             prefactor = Lji;
             dprefactor_dKij[j] += _dDij_dKji[sequential_j][index_j_to_i];
             dprefactor_dKji += 1.0 + _dDij_dKij[sequential_j][index_j_to_i];
           }
           else
           {
             prefactor = _rM[sequential_i] * Dij;
             for (std::size_t ind_j = 0; ind_j < num_con_i; ++ind_j)
               dprefactor_du[ind_j] = _drM[sequential_i][ind_j] * Dij;
             dprefactor_dKij[j] += _rM[sequential_i] * _dDij_dKij[sequential_i][j];
             dprefactor_dKji += _rM[sequential_i] * _dDij_dKji[sequential_i][j];
             for (std::size_t ind_j = 0; ind_j < num_con_i; ++ind_j)
               dprefactor_dKij[ind_j] += _drM_dk[sequential_i][ind_j] * Dij;
           }
         }
         _fa[sequential_i][j] = prefactor * (u_i - u_j);
         _dfa[sequential_i][j][global_i] = prefactor;
         _dfa[sequential_i][j][global_j] = -prefactor;
         for (std::size_t ind_j = 0; ind_j < num_con_i; ++ind_j)
         {
           _dfa[sequential_i][j][_connections.globalID(con_i[ind_j])] +=
               dprefactor_du[ind_j] * (u_i - u_j);
           _dFij_dKik[sequential_i][j][ind_j] += dprefactor_dKij[ind_j] * (u_i - u_j);
         }
         _dFij_dKjk[sequential_i][j][index_j_to_i] += dprefactor_dKji * (u_i - u_j);
       }
     }
   }
  
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const std::size_t num_con_i = con_i.size();
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       const dof_id_type sequential_j = con_i[j];
       if (sequential_i == sequential_j)
         continue;
       const unsigned index_j_to_i =
           _connections.indexOfSequentialConnection(sequential_j, sequential_i);
       if (_lij[sequential_j][index_j_to_i] < _lij[sequential_i][j]) // node_i is downwind of node_j.
       {
         _fa[sequential_i][j] = -_fa[sequential_j][index_j_to_i];
         for (const auto & dof_deriv : _dfa[sequential_j][index_j_to_i])
           _dfa[sequential_i][j][dof_deriv.first] = -dof_deriv.second;
         for (std::size_t k = 0; k < num_con_i; ++k)
           _dFij_dKik[sequential_i][j][k] = -_dFij_dKjk[sequential_j][index_j_to_i][k];
         const std::size_t num_con_j =
             _connections.sequentialConnectionsToSequentialID(sequential_j).size();
         for (std::size_t k = 0; k < num_con_j; ++k)
           _dFij_dKjk[sequential_i][j][k] = -_dFij_dKik[sequential_j][index_j_to_i][k];
       }
     }
   }
  
   // zero _flux_out and its derivatives
   _flux_out.assign(_number_of_nodes, 0.0);
   // The derivatives are a bit complicated.
   // If i is upwind of a node "j" then _flux_out[i] depends on all nodes connected to i.
   // But if i is downwind of a node "j" then _flux_out depends on all nodes connected with node
   // j.
   _dflux_out_du.assign(_number_of_nodes, std::map<dof_id_type, Real>());
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
     for (const auto & j : _connections.globalConnectionsToSequentialID(sequential_i))
     {
       _dflux_out_du[sequential_i][j] = 0.0;
       for (const auto & neighbors_j : _connections.globalConnectionsToGlobalID(j))
         _dflux_out_du[sequential_i][neighbors_j] = 0.0;
     }
   _dflux_out_dKjk.resize(_number_of_nodes);
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const std::size_t num_con_i = con_i.size();
     _dflux_out_dKjk[sequential_i].resize(num_con_i);
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       const dof_id_type sequential_j = con_i[j];
       const std::vector<dof_id_type> & con_j =
           _connections.sequentialConnectionsToSequentialID(sequential_j);
       _dflux_out_dKjk[sequential_i][j].assign(con_j.size(), 0.0);
     }
   }
  
   // Add everything together
   // See step 3 in Fig 2, noting Eqn (36)
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::vector<dof_id_type> & con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i);
     const size_t num_con_i = con_i.size();
     const dof_id_type index_i_to_i =
         _connections.indexOfSequentialConnection(sequential_i, sequential_i);
     for (std::size_t j = 0; j < num_con_i; ++j)
     {
       const dof_id_type sequential_j = con_i[j];
       const dof_id_type global_j = _connections.globalID(sequential_j);
       const Real u_j = _u_nodal[sequential_j];
  
       // negative sign because residual = -Lu (KT equation (19))
       _flux_out[sequential_i] -= _lij[sequential_i][j] * u_j + _fa[sequential_i][j];
  
       _dflux_out_du[sequential_i][global_j] -= _lij[sequential_i][j];
       for (const auto & dof_deriv : _dfa[sequential_i][j])
         _dflux_out_du[sequential_i][dof_deriv.first] -= dof_deriv.second;
  
       _dflux_out_dKjk[sequential_i][index_i_to_i][j] -= 1.0 * u_j; // from the K in L = K + D
  
       if (sequential_j == sequential_i)
         for (dof_id_type k = 0; k < num_con_i; ++k)
           _dflux_out_dKjk[sequential_i][index_i_to_i][k] -= _dDii_dKij[sequential_i][k] * u_j;
       else
         _dflux_out_dKjk[sequential_i][index_i_to_i][j] -= _dDij_dKij[sequential_i][j] * u_j;
       for (dof_id_type k = 0; k < num_con_i; ++k)
         _dflux_out_dKjk[sequential_i][index_i_to_i][k] -= _dFij_dKik[sequential_i][j][k];
  
       if (sequential_j == sequential_i)
         for (unsigned k = 0; k < con_i.size(); ++k)
         {
           const unsigned index_k_to_i =
               _connections.indexOfSequentialConnection(con_i[k], sequential_i);
           _dflux_out_dKjk[sequential_i][k][index_k_to_i] -= _dDii_dKji[sequential_i][k] * u_j;
         }
       else
       {
         const unsigned index_j_to_i =
             _connections.indexOfSequentialConnection(sequential_j, sequential_i);
         _dflux_out_dKjk[sequential_i][j][index_j_to_i] -= _dDij_dKji[sequential_i][j] * u_j;
       }
       for (unsigned k = 0;
            k < _connections.sequentialConnectionsToSequentialID(sequential_j).size();
            ++k)
         _dflux_out_dKjk[sequential_i][j][k] -= _dFij_dKjk[sequential_i][j][k];
     }
   }
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::finalize().

◆ getdFluxOutdKjk()

const std::vector< std::vector< Real > > & AdvectiveFluxCalculatorBase::getdFluxOutdKjk ( dof_id_type node_i ) const

inherited

Returns r where r[j][k] = d(flux out of global node i)/dK[connected node j][connected node k] used in Jacobian computations.

Parameters

node_i global id of node

Returns: the derivatives (after applying the KT procedure)

Definition at line 737 of file AdvectiveFluxCalculatorBase.C.

 {
   return _dflux_out_dKjk[_connections.sequentialID(node_i)];
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::finalize().

◆ getdFluxOutdu()

const std::map< dof_id_type, Real > & AdvectiveFluxCalculatorBase::getdFluxOutdu ( dof_id_type node_i ) const

inherited

Returns r where r[j] = d(flux out of global node i)/du(global node j) used in Jacobian computations.

Parameters

node_i global id of node

Returns: the derivatives (after applying the KT procedure)

Definition at line 731 of file AdvectiveFluxCalculatorBase.C.

 {
   return _dflux_out_du[_connections.sequentialID(node_i)];
 }

Referenced by FluxLimitedTVDAdvection::computeJacobian(), and PorousFlowAdvectiveFluxCalculatorBase::finalize().

◆ getFluxOut()

Real AdvectiveFluxCalculatorBase::getFluxOut ( dof_id_type node_i ) const

inherited

Returns the flux out of lobal node id.

Parameters

node_i id of node

Returns: advective flux out of node after applying the KT procedure

Definition at line 743 of file AdvectiveFluxCalculatorBase.C.

 {
   return _flux_out[_connections.sequentialID(node_i)];
 }

Referenced by FluxLimitedTVDAdvection::computeResidual(), and PorousFlowFluxLimitedTVDAdvection::computeResidual().

◆ getValence()

unsigned AdvectiveFluxCalculatorBase::getValence ( dof_id_type node_i ) const

inherited

Returns the valence of the global node i Valence is the number of times the node is encountered in a loop over elements (that have appropriate subdomain_id, if the user has employed the "blocks=" parameter) seen by this processor (including ghosted elements)

Parameters

node_i gloal id of i^th node

Returns: valence of the node

Definition at line 749 of file AdvectiveFluxCalculatorBase.C.

 {
   return _valence[_connections.sequentialID(node_i)];
 }

Referenced by FluxLimitedTVDAdvection::computeJacobian(), PorousFlowFluxLimitedTVDAdvection::computeJacobian(), FluxLimitedTVDAdvection::computeResidual(), and PorousFlowFluxLimitedTVDAdvection::computeResidual().

◆ initialize()

void AdvectiveFluxCalculatorBase::initialize ( )

overridevirtualinherited

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 207 of file AdvectiveFluxCalculatorBase.C.

 {
   // Zero _kij and falsify _u_nodal_computed_by_thread ready for building in execute() and
   // finalize()
   _u_nodal_computed_by_thread.assign(_number_of_nodes, false);
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
     _kij[sequential_i].assign(_connections.sequentialConnectionsToSequentialID(sequential_i).size(),
                               0.0);
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::initialize().

◆ limitFlux()

void AdvectiveFluxCalculatorBase::limitFlux	(	Real	a,
		Real	b,
		Real &	limited,
		Real &	dlimited_db
	)		const

protectedinherited

flux limiter, L, on Page 135 of Kuzmin and Turek

Parameters

a	KT's "a" parameter
b	KT's "b" parameter
limited[out]	The value of the flux limiter, L
dlimited_db[out]	The derivative dL/db

Definition at line 631 of file AdvectiveFluxCalculatorBase.C.

 {
   limited = 0.0;
   dlimited_db = 0.0;
   if (_flux_limiter_type == FluxLimiterTypeEnum::None)
     return;
  
   if ((a >= 0.0 && b <= 0.0) || (a <= 0.0 && b >= 0.0))
     return;
   const Real s = (a > 0.0 ? 1.0 : -1.0);
  
   const Real lal = std::abs(a);
   const Real lbl = std::abs(b);
   const Real dlbl = (b >= 0.0 ? 1.0 : -1.0); // d(lbl)/db
   switch (_flux_limiter_type)
   {
     case FluxLimiterTypeEnum::MinMod:
     {
       if (lal <= lbl)
       {
         limited = s * lal;
         dlimited_db = 0.0;
       }
       else
       {
         limited = s * lbl;
         dlimited_db = s * dlbl;
       }
       return;
     }
     case FluxLimiterTypeEnum::VanLeer:
     {
       limited = s * 2 * lal * lbl / (lal + lbl);
       dlimited_db = s * 2 * lal * (dlbl / (lal + lbl) - lbl * dlbl / std::pow(lal + lbl, 2));
       return;
     }
     case FluxLimiterTypeEnum::MC:
     {
       const Real av = 0.5 * std::abs(a + b);
       if (2 * lal <= av && lal <= lbl)
       {
         // 2 * lal is the smallest
         limited = s * 2.0 * lal;
         dlimited_db = 0.0;
       }
       else if (2 * lbl <= av && lbl <= lal)
       {
         // 2 * lbl is the smallest
         limited = s * 2.0 * lbl;
         dlimited_db = s * 2.0 * dlbl;
       }
       else
       {
         // av is the smallest
         limited = s * av;
         // if (a>0 and b>0) then d(av)/db = 0.5 = 0.5 * dlbl
         // if (a<0 and b<0) then d(av)/db = -0.5 = 0.5 * dlbl
         // if a and b have different sign then limited=0, above
         dlimited_db = s * 0.5 * dlbl;
       }
       return;
     }
     case FluxLimiterTypeEnum::superbee:
     {
       const Real term1 = std::min(2.0 * lal, lbl);
       const Real term2 = std::min(lal, 2.0 * lbl);
       if (term1 >= term2)
       {
         if (2.0 * lal <= lbl)
         {
           limited = s * 2 * lal;
           dlimited_db = 0.0;
         }
         else
         {
           limited = s * lbl;
           dlimited_db = s * dlbl;
         }
       }
       else
       {
         if (lal <= 2.0 * lbl)
         {
           limited = s * lal;
           dlimited_db = 0.0;
         }
         else
         {
           limited = s * 2.0 * lbl;
           dlimited_db = s * 2.0 * dlbl;
         }
       }
       return;
     }
     default:
       return;
   }
 }

Referenced by AdvectiveFluxCalculatorBase::rMinus(), and AdvectiveFluxCalculatorBase::rPlus().

◆ meshChanged()

void AdvectiveFluxCalculatorBase::meshChanged ( )

overridevirtualinherited

Definition at line 198 of file AdvectiveFluxCalculatorBase.C.

 {
   ElementUserObject::meshChanged();
  
   // Signal that _kij, _valence, etc need to be rebuilt
   _resizing_needed = true;
 }

◆ PQPlusMinus()

Real AdvectiveFluxCalculatorBase::PQPlusMinus	(	dof_id_type	sequential_i,
		const PQPlusMinusEnum	pq_plus_minus,
		std::vector< Real > &	derivs,
		std::vector< Real > &	dpq_dk
	)		const

protectedinherited

Returns the value of P_{i}^{+}, P_{i}^{-}, Q_{i}^{+} or Q_{i}^{-} (depending on pq_plus_minus) which are defined in Eqns (47) and (48) of KT.

Parameters

sequential_i	sequential nodal ID
pq_plus_minus	indicates whether P_{i}^{+}, P_{i}^{-}, Q_{i}^{+} or Q_{i}^{-} should be returned
derivs[out]	derivs[j] = d(result)/d(u[sequential_j]). Here sequential_j is the j^th connection to sequential_i
dpq_dk[out]	dpq_dk[j] = d(result)/d(K[node_i][j]). Here j indexes a connection to sequential_i. Recall that d(result)/d(K[l][m]) are zero unless l=sequential_i

Definition at line 764 of file AdvectiveFluxCalculatorBase.C.

 {
   // Find the value of u at sequential_i
   const Real u_i = _u_nodal[sequential_i];
  
   // Connections to sequential_i
   const std::vector<dof_id_type> con_i =
       _connections.sequentialConnectionsToSequentialID(sequential_i);
   const std::size_t num_con = con_i.size();
   // The neighbor number of sequential_i to sequential_i
   const unsigned i_index_i = _connections.indexOfSequentialConnection(sequential_i, sequential_i);
  
   // Initialize the results
   Real result = 0.0;
   derivs.assign(num_con, 0.0);
   dpqdk.assign(num_con, 0.0);
  
   // Sum over all nodes connected with node_i.
   for (std::size_t j = 0; j < num_con; ++j)
   {
     const dof_id_type sequential_j = con_i[j];
     if (sequential_j == sequential_i)
       continue;
     const Real kentry = _kij[sequential_i][j];
  
     // Find the value of u at node_j
     const Real u_j = _u_nodal[sequential_j];
     const Real ujminusi = u_j - u_i;
  
     // Evaluate the i-j contribution to the result
     switch (pq_plus_minus)
     {
       case PQPlusMinusEnum::PPlus:
       {
         if (ujminusi < 0.0 && kentry < 0.0)
         {
           result += kentry * ujminusi;
           derivs[j] += kentry;
           derivs[i_index_i] -= kentry;
           dpqdk[j] += ujminusi;
         }
         break;
       }
       case PQPlusMinusEnum::PMinus:
       {
         if (ujminusi > 0.0 && kentry < 0.0)
         {
           result += kentry * ujminusi;
           derivs[j] += kentry;
           derivs[i_index_i] -= kentry;
           dpqdk[j] += ujminusi;
         }
         break;
       }
       case PQPlusMinusEnum::QPlus:
       {
         if (ujminusi > 0.0 && kentry > 0.0)
         {
           result += kentry * ujminusi;
           derivs[j] += kentry;
           derivs[i_index_i] -= kentry;
           dpqdk[j] += ujminusi;
         }
         break;
       }
       case PQPlusMinusEnum::QMinus:
       {
         if (ujminusi < 0.0 && kentry > 0.0)
         {
           result += kentry * ujminusi;
           derivs[j] += kentry;
           derivs[i_index_i] -= kentry;
           dpqdk[j] += ujminusi;
         }
         break;
       }
     }
   }
  
   return result;
 }

Referenced by AdvectiveFluxCalculatorBase::rMinus(), and AdvectiveFluxCalculatorBase::rPlus().

◆ rMinus()

Real AdvectiveFluxCalculatorBase::rMinus	(	dof_id_type	sequential_i,
		std::vector< Real > &	dlimited_du,
		std::vector< Real > &	dlimited_dk
	)		const

protectedinherited

Returns the value of R_{i}^{-}, Eqn (49) of KT.

Parameters

sequential_i	Sequential nodal ID
dlimited_du[out]	dlimited_du[j] = d(R_{sequential_i}^{-})/du[sequential_j]. Here sequential_j is the j^th connection to sequential_i
dlimited_dk[out]	dlimited_dk[j] = d(R_{sequential_i}^{-})/d(K[sequential_i][j]). Note Derivatives w.r.t. K[l][m] with l!=i are zero

Definition at line 595 of file AdvectiveFluxCalculatorBase.C.

 {
   const std::size_t num_con = _connections.sequentialConnectionsToSequentialID(sequential_i).size();
   dlimited_du.assign(num_con, 0.0);
   dlimited_dk.assign(num_con, 0.0);
   if (_flux_limiter_type == FluxLimiterTypeEnum::None)
     return 0.0;
   std::vector<Real> dp_du;
   std::vector<Real> dp_dk;
   const Real p = PQPlusMinus(sequential_i, PQPlusMinusEnum::PMinus, dp_du, dp_dk);
   if (p == 0.0)
     // Comment after Eqn (49): if P=0 then there's no antidiffusion, so no need to remove it
     return 1.0;
   std::vector<Real> dq_du;
   std::vector<Real> dq_dk;
   const Real q = PQPlusMinus(sequential_i, PQPlusMinusEnum::QMinus, dq_du, dq_dk);
  
   const Real r = q / p;
   Real limited;
   Real dlimited_dr;
   limitFlux(1.0, r, limited, dlimited_dr);
  
   const Real p2 = std::pow(p, 2);
   for (std::size_t j = 0; j < num_con; ++j)
   {
     const Real dr_du = dq_du[j] / p - q * dp_du[j] / p2;
     const Real dr_dk = dq_dk[j] / p - q * dp_dk[j] / p2;
     dlimited_du[j] = dlimited_dr * dr_du;
     dlimited_dk[j] = dlimited_dr * dr_dk;
   }
   return limited;
 }

Referenced by AdvectiveFluxCalculatorBase::finalize().

◆ rPlus()

Real AdvectiveFluxCalculatorBase::rPlus	(	dof_id_type	sequential_i,
		std::vector< Real > &	dlimited_du,
		std::vector< Real > &	dlimited_dk
	)		const

protectedinherited

Returns the value of R_{i}^{+}, Eqn (49) of KT.

Parameters

node_i	nodal id
dlimited_du[out]	dlimited_du[j] = d(R_{sequential_i}^{+})/du[sequential_j]. Here sequential_j is the j^th connection to sequential_i
dlimited_dk[out]	dlimited_dk[j] = d(R_{sequential_i}^{+})/d(K[sequential_i][j]). Note Derivatives w.r.t. K[l][m] with l!=i are zero

Definition at line 559 of file AdvectiveFluxCalculatorBase.C.

 {
   const std::size_t num_con = _connections.sequentialConnectionsToSequentialID(sequential_i).size();
   dlimited_du.assign(num_con, 0.0);
   dlimited_dk.assign(num_con, 0.0);
   if (_flux_limiter_type == FluxLimiterTypeEnum::None)
     return 0.0;
   std::vector<Real> dp_du;
   std::vector<Real> dp_dk;
   const Real p = PQPlusMinus(sequential_i, PQPlusMinusEnum::PPlus, dp_du, dp_dk);
   if (p == 0.0)
     // Comment after Eqn (49): if P=0 then there's no antidiffusion, so no need to remove it
     return 1.0;
   std::vector<Real> dq_du;
   std::vector<Real> dq_dk;
   const Real q = PQPlusMinus(sequential_i, PQPlusMinusEnum::QPlus, dq_du, dq_dk);
  
   const Real r = q / p;
   Real limited;
   Real dlimited_dr;
   limitFlux(1.0, r, limited, dlimited_dr);
  
   const Real p2 = std::pow(p, 2);
   for (std::size_t j = 0; j < num_con; ++j)
   {
     const Real dr_du = dq_du[j] / p - q * dp_du[j] / p2;
     const Real dr_dk = dq_dk[j] / p - q * dp_dk[j] / p2;
     dlimited_du[j] = dlimited_dr * dr_du;
     dlimited_dk[j] = dlimited_dr * dr_dk;
   }
   return limited;
 }

Referenced by AdvectiveFluxCalculatorBase::finalize().

◆ threadJoin()

void AdvectiveFluxCalculatorBase::threadJoin ( const UserObject & uo )

overridevirtualinherited

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 251 of file AdvectiveFluxCalculatorBase.C.

 {
   const AdvectiveFluxCalculatorBase & afc = static_cast<const AdvectiveFluxCalculatorBase &>(uo);
   // add the values of _kij computed by different threads
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
   {
     const std::size_t num_con_i =
         _connections.sequentialConnectionsToSequentialID(sequential_i).size();
     for (std::size_t j = 0; j < num_con_i; ++j)
       _kij[sequential_i][j] += afc._kij[sequential_i][j];
   }
  
   // gather the values of _u_nodal computed by different threads
   for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
     if (!_u_nodal_computed_by_thread[sequential_i] && afc._u_nodal_computed_by_thread[sequential_i])
       _u_nodal[sequential_i] = afc._u_nodal[sequential_i];
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::threadJoin().

◆ timestepSetup()

void AdvectiveFluxCalculatorBase::timestepSetup ( )

overridevirtualinherited

Reimplemented in PorousFlowAdvectiveFluxCalculatorBase.

Definition at line 81 of file AdvectiveFluxCalculatorBase.C.

 {
   // If needed, size and initialize quantities appropriately, and compute _valence
   if (_resizing_needed)
   {
     /*
      * Populate _connections for all nodes that can be seen by this processor and on relevant
      * blocks
      *
      * MULTIPROC NOTE: this must loop over local elements and 2 layers of ghosted elements.
      * The Kernel will only loop over local elements, so will only use _kij, etc, for
      * linked node-node pairs that appear in the local elements.  Nevertheless, we
      * need to build _kij, etc, for the nodes in the ghosted elements in order to simplify
      * Jacobian computations
      */
     _connections.clear();
     for (const auto & elem : _fe_problem.getEvaluableElementRange())
       if (this->hasBlocks(elem->subdomain_id()))
         for (unsigned i = 0; i < elem->n_nodes(); ++i)
           _connections.addGlobalNode(elem->node_id(i));
     _connections.finalizeAddingGlobalNodes();
     for (const auto & elem : _fe_problem.getEvaluableElementRange())
       if (this->hasBlocks(elem->subdomain_id()))
         for (unsigned i = 0; i < elem->n_nodes(); ++i)
           for (unsigned j = 0; j < elem->n_nodes(); ++j)
             _connections.addConnection(elem->node_id(i), elem->node_id(j));
     _connections.finalizeAddingConnections();
  
     _number_of_nodes = _connections.numNodes();
  
     Real mb_wasted = (_connections.sizeSequential() - _number_of_nodes) * 4.0 / 1048576;
     gatherMax(mb_wasted);
     if (mb_wasted > _allowable_MB_wastage)
       mooseWarning(
           "In at least one processor, the sequential node-numbering internal data structure used "
           "in " +
           name() + " is using memory inefficiently.\nThe memory wasted is " +
           Moose::stringify(mb_wasted) +
           " megabytes.\n  The node-numbering data structure has been optimized for speed at the "
           "expense of memory, but that may not be an appropriate optimization for your case, "
           "because the node numbering is not close to sequential in your case.\n");
  
     // initialize _kij
     _kij.resize(_number_of_nodes);
     for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
       _kij[sequential_i].assign(
           _connections.sequentialConnectionsToSequentialID(sequential_i).size(), 0.0);
  
     /*
      * Build _valence[i], which is the number of times the sequential node i is encountered when
      * looping over the entire mesh (and on relevant blocks)
      *
      * MULTIPROC NOTE: this must loop over local elements and >=1 layer of ghosted elements.
      * The Kernel will only loop over local elements, so will only use _valence for
      * linked node-node pairs that appear in the local elements.  But other processors will
      * loop over neighboring elements, so avoid multiple counting of the residual and Jacobian
      * this processor must record how many times each node-node link of its local elements
      * appears in the local elements and >=1 layer, and pass that info to the Kernel
      */
     _valence.assign(_number_of_nodes, 0);
     for (const auto & elem : _fe_problem.getEvaluableElementRange())
       if (this->hasBlocks(elem->subdomain_id()))
         for (unsigned i = 0; i < elem->n_nodes(); ++i)
         {
           const dof_id_type node_i = elem->node_id(i);
           const dof_id_type sequential_i = _connections.sequentialID(node_i);
           _valence[sequential_i] += 1;
         }
  
     _u_nodal.assign(_number_of_nodes, 0.0);
     _u_nodal_computed_by_thread.assign(_number_of_nodes, false);
     _flux_out.assign(_number_of_nodes, 0.0);
     _dflux_out_du.assign(_number_of_nodes, std::map<dof_id_type, Real>());
     for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
       zeroedConnection(_dflux_out_du[sequential_i], _connections.globalID(sequential_i));
     _dflux_out_dKjk.resize(_number_of_nodes);
     for (dof_id_type sequential_i = 0; sequential_i < _number_of_nodes; ++sequential_i)
     {
       const std::vector<dof_id_type> con_i =
           _connections.sequentialConnectionsToSequentialID(sequential_i);
       const std::size_t num_con_i = con_i.size();
       _dflux_out_dKjk[sequential_i].resize(num_con_i);
       for (std::size_t j = 0; j < con_i.size(); ++j)
       {
         const dof_id_type sequential_j = con_i[j];
         const std::size_t num_con_j =
             _connections.sequentialConnectionsToSequentialID(sequential_j).size();
         _dflux_out_dKjk[sequential_i][j].assign(num_con_j, 0.0);
       }
     }
  
     if (_app.n_processors() > 1)
       buildCommLists();
  
     _resizing_needed = false;
  
     // Clear and size all member vectors
     _dij.assign(_number_of_nodes, {});
     _dDij_dKij.assign(_number_of_nodes, {});
     _dDij_dKji.assign(_number_of_nodes, {});
     _dDii_dKij.assign(_number_of_nodes, {});
     _dDii_dKji.assign(_number_of_nodes, {});
     _lij.assign(_number_of_nodes, {});
     _rP.assign(_number_of_nodes, 0.0);
     _rM.assign(_number_of_nodes, 0.0);
     _drP.assign(_number_of_nodes, {});
     _drM.assign(_number_of_nodes, {});
     _drP_dk.assign(_number_of_nodes, {});
     _drM_dk.assign(_number_of_nodes, {});
     _fa.assign(_number_of_nodes, {});
     _dfa.assign(_number_of_nodes, {});
     _dFij_dKik.assign(_number_of_nodes, {});
     _dFij_dKjk.assign(_number_of_nodes, {});
   }
 }

Referenced by PorousFlowAdvectiveFluxCalculatorBase::timestepSetup().

◆ zeroedConnection()

void AdvectiveFluxCalculatorBase::zeroedConnection	(	std::map< dof_id_type, Real > &	the_map,
		dof_id_type	node_i
	)		const

protectedinherited

Clears the_map, then, using _kij, constructs the_map so that the_map[node_id] = 0.0 for all node_id connected with node_i.

Parameters

[out]	the_map	the map to be zeroed appropriately
[in]	node_i	nodal id

Definition at line 755 of file AdvectiveFluxCalculatorBase.C.

 {
   the_map.clear();
   for (const auto & node_j : _connections.globalConnectionsToGlobalID(node_i))
     the_map[node_j] = 0.0;
 }

Referenced by AdvectiveFluxCalculatorBase::timestepSetup().

Member Data Documentation

◆ _allowable_MB_wastage

const Real AdvectiveFluxCalculatorBase::_allowable_MB_wastage

protectedinherited

A mooseWarning is issued if mb_wasted = (_connections.sizeSequential() - _connections.numNodes()) * 4 / 1048576 > _allowable_MB_wastage.

The _connections object uses sequential node numbering for computational efficiency, but this leads to memory being used inefficiently: the number of megabytes wasted is mb_wasted.

Definition at line 252 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::timestepSetup().

◆ _connections

PorousFlowConnectedNodes AdvectiveFluxCalculatorBase::_connections

protectedinherited

Holds the sequential and global nodal IDs, and info regarding mesh connections between them.

Definition at line 204 of file AdvectiveFluxCalculatorBase.h.

◆ _dDii_dKij

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_dDii_dKij

protectedinherited

dDii_dKij[i][j] = d(D[i][i])/d(K[i][j])

Definition at line 294 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dDii_dKji

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_dDii_dKji

protectedinherited

dDii_dKji[i][j] = d(D[i][i])/d(K[j][i])

Definition at line 296 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dDij_dKij

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_dDij_dKij

protectedinherited

dDij_dKij[i][j] = d(D[i][j])/d(K[i][j]) for i!=j

Definition at line 290 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dDij_dKji

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_dDij_dKji

protectedinherited

dDij_dKji[i][j] = d(D[i][j])/d(K[j][i]) for i!=j

Definition at line 292 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dfa

std::vector<std::vector<std::map<dof_id_type, Real> > > AdvectiveFluxCalculatorBase::_dfa

protectedinherited

dfa[sequential_i][j][global_k] = d(fa[sequential_i][j])/du[global_k].

Here global_k can be a neighbor to sequential_i or a neighbour to sequential_j (sequential_j is the j^th connection to sequential_i)

Definition at line 316 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dFij_dKik

std::vector<std::vector<std::vector<Real> > > AdvectiveFluxCalculatorBase::_dFij_dKik

protectedinherited

dFij_dKik[sequential_i][j][k] = d(fa[sequential_i][j])/d(K[sequential_i][k]).

Here j denotes the j^th connection to sequential_i, while k denotes the k^th connection to sequential_i

Definition at line 321 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dFij_dKjk

std::vector<std::vector<std::vector<Real> > > AdvectiveFluxCalculatorBase::_dFij_dKjk

protectedinherited

dFij_dKjk[sequential_i][j][k] = d(fa[sequential_i][j])/d(K[sequential_j][k]).

Here sequential_j is the j^th connection to sequential_i, and k denotes the k^th connection to sequential_j (this will include sequential_i itself)

Definition at line 326 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dflux_out_dKjk

std::vector<std::vector<std::vector<Real> > > AdvectiveFluxCalculatorBase::_dflux_out_dKjk

protectedinherited

_dflux_out_dKjk[sequential_i][j][k] = d(flux_out[sequential_i])/d(K[j][k]).

Here sequential_i is a sequential node number according to the _connections object, and j represents the j^th node connected to i according to the _connections object, and k represents the k^th node connected to j according to the _connections object. Here j must be connected to i (this does include (the sequential version of j) == i), and k must be connected to j (this does include (the sequential version of k) = i and (the sequential version of k) == (sequential version of j)))

Definition at line 190 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::getdFluxOutdKjk(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dflux_out_du

std::vector<std::map<dof_id_type, Real> > AdvectiveFluxCalculatorBase::_dflux_out_du

protectedinherited

_dflux_out_du[i][j] = d(flux_out[i])/d(u[j]).

Here i is a sequential node number according to the _connections object, and j (global ID) must be connected to i, or to a node that is connected to i.

Definition at line 183 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::getdFluxOutdu(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _dij

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_dij

protectedinherited

Vectors used in finalize()

Definition at line 288 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _drM

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_drM

protectedinherited

drM[i][j] = d(rM[i])/d(u[j]). Here j indexes the j^th node connected to i

Definition at line 305 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _drM_dk

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_drM_dk

protectedinherited

drM_dk[i][j] = d(rM[i])/d(K[i][j]). Here j indexes the j^th node connected to i

Definition at line 309 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _drP

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_drP

protectedinherited

drP[i][j] = d(rP[i])/d(u[j]). Here j indexes the j^th node connected to i

Definition at line 303 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _drP_dk

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_drP_dk

protectedinherited

drP_dk[i][j] = d(rP[i])/d(K[i][j]). Here j indexes the j^th node connected to i

Definition at line 307 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _fa

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_fa

protectedinherited

fa[sequential_i][j] sequential_j is the j^th connection to sequential_i

Definition at line 312 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _flux_limiter_type

enum AdvectiveFluxCalculatorBase::FluxLimiterTypeEnum AdvectiveFluxCalculatorBase::_flux_limiter_type

protectedinherited

Referenced by AdvectiveFluxCalculatorBase::limitFlux(), AdvectiveFluxCalculatorBase::rMinus(), and AdvectiveFluxCalculatorBase::rPlus().

◆ _flux_out

std::vector<Real> AdvectiveFluxCalculatorBase::_flux_out

protectedinherited

_flux_out[i] = flux of "heat" from sequential node i

Definition at line 179 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::getFluxOut(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _grad_phi

const VariablePhiGradient& AdvectiveFluxCalculatorConstantVelocity::_grad_phi

protected

grad(Kuzmin-Turek shape function)

Definition at line 42 of file AdvectiveFluxCalculatorConstantVelocity.h.

Referenced by computeVelocity().

◆ _kij

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_kij

protectedinherited

Kuzmin-Turek K_ij matrix.

Along with R+ and R-, this is the key quantity computed by this UserObject. _kij[i][j] = k_ij corresponding to the i-j node pair. Here i is a sequential node numbers according to the _connections object, and j represents the j^th node connected to i according to the _connections object.

Definition at line 176 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::exchangeGhostedInfo(), AdvectiveFluxCalculatorBase::executeOnElement(), AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::initialize(), AdvectiveFluxCalculatorBase::PQPlusMinus(), AdvectiveFluxCalculatorBase::threadJoin(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _lij

std::vector<std::vector<Real> > AdvectiveFluxCalculatorBase::_lij

protectedinherited

Definition at line 298 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _my_pid

processor_id_type AdvectiveFluxCalculatorBase::_my_pid

protectedinherited

processor ID of this object

Definition at line 210 of file AdvectiveFluxCalculatorBase.h.

Referenced by PorousFlowAdvectiveFluxCalculatorBase::buildCommLists(), and AdvectiveFluxCalculatorBase::buildCommLists().

◆ _nodes_to_receive

std::map<processor_id_type, std::vector<dof_id_type> > AdvectiveFluxCalculatorBase::_nodes_to_receive

protectedinherited

_nodes_to_receive[proc_id] = list of sequential nodal IDs.

proc_id will send us _u_nodal at those nodes. _nodes_to_receive is built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs is made, for efficiency. The result is: we will receive _u_nodal[_nodes_to_receive[proc_id][i]] from proc_id

Definition at line 218 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::buildCommLists(), PorousFlowAdvectiveFluxCalculatorBase::exchangeGhostedInfo(), and AdvectiveFluxCalculatorBase::exchangeGhostedInfo().

◆ _nodes_to_send

std::map<processor_id_type, std::vector<dof_id_type> > AdvectiveFluxCalculatorBase::_nodes_to_send

protectedinherited

_nodes_to_send[proc_id] = list of sequential nodal IDs.

We will send _u_nodal at those nodes to proc_id _nodes_to_send is built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs is made, for efficiency The result is: we will send _u_nodal[_nodes_to_receive[proc_id][i]] to proc_id

Definition at line 226 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::buildCommLists(), PorousFlowAdvectiveFluxCalculatorBase::exchangeGhostedInfo(), and AdvectiveFluxCalculatorBase::exchangeGhostedInfo().

◆ _number_of_nodes

std::size_t AdvectiveFluxCalculatorBase::_number_of_nodes

protectedinherited

Number of nodes held by the _connections object.

Definition at line 207 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::initialize(), AdvectiveFluxCalculatorBase::threadJoin(), PorousFlowAdvectiveFluxCalculatorBase::threadJoin(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _pairs_to_receive

std::map<processor_id_type, std::vector<std::pair<dof_id_type, dof_id_type> > > AdvectiveFluxCalculatorBase::_pairs_to_receive

protectedinherited

_pairs_to_receive[proc_id] indicates the k(i, j) pairs that will be sent to us from proc_id _pairs_to_receive is first built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs and the index of connections is performed, for efficiency.

The result is we will receive: _kij[_pairs_to_receive[proc_id][i].first][_pairs_to_receive[proc_id][i].second] from proc_id

Definition at line 235 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::buildCommLists(), and AdvectiveFluxCalculatorBase::exchangeGhostedInfo().

◆ _pairs_to_send

std::map<processor_id_type, std::vector<std::pair<dof_id_type, dof_id_type> > > AdvectiveFluxCalculatorBase::_pairs_to_send

protectedinherited

_pairs_to_send[proc_id] indicates the k(i, j) pairs that we will send to proc_id _pairs_to_send is first built (in buildCommLists()) using global node IDs, but after construction, a translation to sequential node IDs and the index of connections is performed, for efficiency.

The result is we will send: _kij[_pairs_to_send[proc_id][i].first][_pairs_to_send[proc_id][i+1].second] to proc_id

Definition at line 244 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::buildCommLists(), and AdvectiveFluxCalculatorBase::exchangeGhostedInfo().

◆ _phi

const VariablePhiValue& AdvectiveFluxCalculatorConstantVelocity::_phi

protected

Kuzmin-Turek shape function.

Definition at line 39 of file AdvectiveFluxCalculatorConstantVelocity.h.

Referenced by computeVelocity().

◆ _resizing_needed

bool AdvectiveFluxCalculatorBase::_resizing_needed

protectedinherited

whether _kij, etc, need to be sized appropriately (and valence recomputed) at the start of the timestep

Definition at line 130 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::meshChanged(), AdvectiveFluxCalculatorBase::timestepSetup(), and PorousFlowAdvectiveFluxCalculatorBase::timestepSetup().

◆ _rM

std::vector<Real> AdvectiveFluxCalculatorBase::_rM

protectedinherited

Definition at line 300 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _rP

std::vector<Real> AdvectiveFluxCalculatorBase::_rP

protectedinherited

Definition at line 299 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::finalize(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _u_at_nodes

const VariableValue& AdvectiveFluxCalculatorConstantVelocity::_u_at_nodes

protected

the nodal values of u

Definition at line 36 of file AdvectiveFluxCalculatorConstantVelocity.h.

Referenced by computeU().

◆ _u_nodal

std::vector<Real> AdvectiveFluxCalculatorBase::_u_nodal

protectedinherited

_u_nodal[i] = value of _u at sequential node number i

Definition at line 198 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::exchangeGhostedInfo(), AdvectiveFluxCalculatorBase::execute(), AdvectiveFluxCalculatorBase::finalize(), AdvectiveFluxCalculatorBase::PQPlusMinus(), AdvectiveFluxCalculatorBase::threadJoin(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _u_nodal_computed_by_thread

std::vector<bool> AdvectiveFluxCalculatorBase::_u_nodal_computed_by_thread

protectedinherited

_u_nodal_computed_by_thread(i) = true if _u_nodal[i] has been computed in execute() by the thread on this processor

Definition at line 201 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::execute(), AdvectiveFluxCalculatorBase::initialize(), AdvectiveFluxCalculatorBase::threadJoin(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _valence

std::vector<unsigned> AdvectiveFluxCalculatorBase::_valence

protectedinherited

_valence[i] = number of times, in a loop over elements seen by this processor (viz, including ghost elements) and are part of the block-restricted blocks of this UserObject, that the sequential node i is encountered

Definition at line 195 of file AdvectiveFluxCalculatorBase.h.

Referenced by AdvectiveFluxCalculatorBase::getValence(), and AdvectiveFluxCalculatorBase::timestepSetup().

◆ _velocity

RealVectorValue AdvectiveFluxCalculatorConstantVelocity::_velocity

protected

advection velocity

Definition at line 33 of file AdvectiveFluxCalculatorConstantVelocity.h.

Referenced by computeVelocity().

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

porous_flow/include/userobjects/AdvectiveFluxCalculatorConstantVelocity.h
porous_flow/src/userobjects/AdvectiveFluxCalculatorConstantVelocity.C

Public Member Functions

Protected Types

Protected Member Functions

Protected Attributes

Detailed Description

Member Enumeration Documentation

◆ FluxLimiterTypeEnum

◆ PQPlusMinusEnum

Constructor & Destructor Documentation

◆ AdvectiveFluxCalculatorConstantVelocity()

Member Function Documentation

◆ buildCommLists()

◆ computeU()

◆ computeVelocity()

◆ exchangeGhostedInfo()

◆ execute()

◆ executeOnElement()

◆ finalize()

◆ getdFluxOutdKjk()

◆ getdFluxOutdu()

◆ getFluxOut()

◆ getValence()

◆ initialize()

◆ limitFlux()

◆ meshChanged()

◆ PQPlusMinus()

◆ rMinus()

◆ rPlus()

◆ threadJoin()

◆ timestepSetup()

◆ zeroedConnection()

Member Data Documentation

◆ _allowable_MB_wastage

◆ _connections

◆ _dDii_dKij

◆ _dDii_dKji

◆ _dDij_dKij

◆ _dDij_dKji

◆ _dfa

◆ _dFij_dKik

◆ _dFij_dKjk

◆ _dflux_out_dKjk

◆ _dflux_out_du

◆ _dij

◆ _drM

◆ _drM_dk

◆ _drP

◆ _drP_dk

◆ _fa

◆ _flux_limiter_type

◆ _flux_out

◆ _grad_phi

◆ _kij

◆ _lij

◆ _my_pid

◆ _nodes_to_receive

◆ _nodes_to_send

◆ _number_of_nodes

◆ _pairs_to_receive

◆ _pairs_to_send

◆ _phi

◆ _resizing_needed

◆ _rM

◆ _rP

◆ _u_at_nodes

◆ _u_nodal

◆ _u_nodal_computed_by_thread

◆ _valence

◆ _velocity