Public Member Functions | |
FEFieldFunction (const DH &dh, const VECTOR &data_vector, const Mapping< dim > &mapping=StaticMappingQ1< dim >::mapping) | |
void | set_active_cell (typename DH::active_cell_iterator &newcell) |
virtual void | vector_value (const Point< dim > &p, Vector< double > &values) const |
virtual double | value (const Point< dim > &p, const unsigned int component=0) const |
virtual void | value_list (const std::vector< Point< dim > > &points, std::vector< double > &values, const unsigned int component=0) const |
virtual void | vector_value_list (const std::vector< Point< dim > > &points, std::vector< Vector< double > > &values) const |
virtual void | vector_gradient (const Point< dim > &p, std::vector< Tensor< 1, dim > > &gradients) const |
virtual Tensor< 1, dim > | gradient (const Point< dim > &p, const unsigned int component=0) const |
virtual void | vector_gradient_list (const std::vector< Point< dim > > &p, std::vector< std::vector< Tensor< 1, dim > > > &gradients) const |
virtual void | gradient_list (const std::vector< Point< dim > > &p, std::vector< Tensor< 1, dim > > &gradients, const unsigned int component=0) const |
unsigned int | compute_point_locations (const std::vector< Point< dim > > &points, std::vector< typename DH::active_cell_iterator > &cells, std::vector< std::vector< Point< dim > > > &qpoints, std::vector< std::vector< unsigned int > > &maps) const |
Private Attributes | |
SmartPointer< const DH > | dh |
const VECTOR & | data_vector |
const Mapping< dim > & | mapping |
DH::active_cell_iterator | cell |
const unsigned int | n_components |
This is an interpolation function for the given dof handler and the given solution vector. The points at which this function can be evaluated MUST be inside the domain of the dof handler, but except from this, no other requirement is given. This function is rather slow, as it needs to construct a quadrature object for the point (or set of points) where you want to evaluate your finite element function. In order to do so, it needs to find out where the points lie.
If you know in advance in which cell your points lye, you can accelerate things a bit, by calling set_active_cell before asking for values or gradients of the function. If you don't do this, and your points don't lie in the cell that is currently stored, the function GridTools::find_cell_around_point is called to find out where the point is. You can specify an optional mapping to use when looking for points in the grid. If you don't do so, this function uses a Q1 mapping.
Once the FEFieldFunction knows where the points lie, it creates a quadrature formula for those points, and calls FEValues::get_function_values or FEValues::get_function_grads with the given quadrature points.
If you only need the quadrature points but not the values of the finite element function (you might want this for the adjoint interpolation), you can also use the function compute_point_locations
alone.
An example of how to use this function is the following:
// Generate two triangulations Triangulation<dim> tria_1; Triangulation<dim> tria_2; // Read the triangulations from files, or build them up, or get // them from some place... Assume that tria_2 is *entirely* // included in tria_1 ... // Associate a dofhandler and a solution to the first // triangulation DoFHandler<dim> dh1(tria_1); Vector<double> solution_1; // Do the same with the second DoFHandler<dim> dh2; Vector<double> solution_2; // Setup the system, assemble matrices, solve problems and get the // nobel prize on the first domain... ... // Now project it to the second domain FEFieldFunction<dim> fe_function_1 (dh_1, solution_1); VectorTools::project(dh_2, constraints_2, quad, fe_function_1, solution_2); // Or interpolate it... Vector<double> solution_3; VectorTools::interpolate(dh_2, fe_function_1, solution_3);
The snippet of code above will work assuming that the second triangulation is entirely included in the first one.
FEFieldFunction is designed to be an easy way to get the results of your computations across different, possibly non matching, grids. No knowledge of the location of the points is assumed in this class, which makes it rely entirely on the GridTools::find_active_cell_around_point utility for its job. However the class can be fed an "educated guess" of where the points that will be computed actually are by using the FEFieldFunction::set_active_cell method, so if you have a smart way to tell where your points are, you will save a lot of computational time by letting this class know.
An optimization based on a caching mechanism was used by the author of this class for the implementation of a Finite Element Immersed Boundary Method.
Functions::FEFieldFunction< dim, DH, VECTOR >::FEFieldFunction | ( | const DH & | dh, | |
const VECTOR & | data_vector, | |||
const Mapping< dim > & | mapping = StaticMappingQ1< dim >::mapping | |||
) |
Construct a vector function. A smart pointers is stored to the dof handler, so you have to make sure that it make sense for the entire lifetime of this object. The number of components of this functions is equal to the number of components of the finite element object. If a mapping is specified, that is what is used to find out where the points lay. Otherwise the standard Q1 mapping is used.
void Functions::FEFieldFunction< dim, DH, VECTOR >::set_active_cell | ( | typename DH::active_cell_iterator & | newcell | ) |
Set the current cell. If you know in advance where your points lie, you can tell this object by calling this function. This will speed things up a little.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::vector_value | ( | const Point< dim > & | p, | |
Vector< double > & | values | |||
) | const [virtual] |
Get ONE vector value at the given point. It is inefficient to use single points. If you need more than one at a time, use the vector_value_list() function. For efficiency reasons, it is better if all the points lie on the same cell. This is not mandatory, however it does speed things up.
Reimplemented from Function< dim >.
virtual double Functions::FEFieldFunction< dim, DH, VECTOR >::value | ( | const Point< dim > & | p, | |
const unsigned int | component = 0 | |||
) | const [virtual] |
Return the value of the function at the given point. Unless there is only one component (i.e. the function is scalar), you should state the component you want to have evaluated; it defaults to zero, i.e. the first component. It is inefficient to use single points. If you need more than one at a time, use the vector_value_list function. For efficiency reasons, it is better if all the points lie on the same cell. This is not mandatory, however it does speed things up.
Reimplemented from Function< dim >.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::value_list | ( | const std::vector< Point< dim > > & | points, | |
std::vector< double > & | values, | |||
const unsigned int | component = 0 | |||
) | const [virtual] |
Set values
to the point values of the specified component of the function at the points
. It is assumed that values
already has the right size, i.e. the same size as the points array. This is rather efficient if all the points lie on the same cell. If this is not the case, things may slow down a bit.
Reimplemented from Function< dim >.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::vector_value_list | ( | const std::vector< Point< dim > > & | points, | |
std::vector< Vector< double > > & | values | |||
) | const [virtual] |
Set values
to the point values of the function at the points
. It is assumed that values
already has the right size, i.e. the same size as the points array. This is rather efficient if all the points lie on the same cell. If this is not the case, things may slow down a bit.
Reimplemented from Function< dim >.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::vector_gradient | ( | const Point< dim > & | p, | |
std::vector< Tensor< 1, dim > > & | gradients | |||
) | const [virtual] |
Return the gradient of all components of the function at the given point. It is inefficient to use single points. If you need more than one at a time, use the vector_value_list function. For efficiency reasons, it is better if all the points lie on the same cell. This is not mandatory, however it does speed things up.
Reimplemented from Function< dim >.
virtual Tensor<1,dim> Functions::FEFieldFunction< dim, DH, VECTOR >::gradient | ( | const Point< dim > & | p, | |
const unsigned int | component = 0 | |||
) | const [virtual] |
Return the gradient of the specified component of the function at the given point. It is inefficient to use single points. If you need more than one at a time, use the vector_value_list function. For efficiency reasons, it is better if all the points lie on the same cell. This is not mandatory, however it does speed things up.
Reimplemented from Function< dim >.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::vector_gradient_list | ( | const std::vector< Point< dim > > & | p, | |
std::vector< std::vector< Tensor< 1, dim > > > & | gradients | |||
) | const [virtual] |
Return the gradient of all components of the function at all the given points. This is rather efficient if all the points lie on the same cell. If this is not the case, things may slow down a bit.
Reimplemented from Function< dim >.
virtual void Functions::FEFieldFunction< dim, DH, VECTOR >::gradient_list | ( | const std::vector< Point< dim > > & | p, | |
std::vector< Tensor< 1, dim > > & | gradients, | |||
const unsigned int | component = 0 | |||
) | const [virtual] |
Return the gradient of the specified component of the function at all the given points. This is rather efficient if all the points lie on the same cell. If this is not the case, things may slow down a bit.
Reimplemented from Function< dim >.
unsigned int Functions::FEFieldFunction< dim, DH, VECTOR >::compute_point_locations | ( | const std::vector< Point< dim > > & | points, | |
std::vector< typename DH::active_cell_iterator > & | cells, | |||
std::vector< std::vector< Point< dim > > > & | qpoints, | |||
std::vector< std::vector< unsigned int > > & | maps | |||
) | const |
Create quadrature rules. This function groups the points into blocks that live in the same cell, and fills up three vectors: cells
, qpoints
and maps
. The first is a list of the cells that contain the points, the second is a list of quadrature points matching each cell of the first list, and the third contains the index of the given quadrature points, i.e., points
[maps[3][4]] ends up as the 5th quadrature point in the 4th cell. This is where optimization would help. This function returns the number of cells that contain the given set of points.
SmartPointer<const DH> Functions::FEFieldFunction< dim, DH, VECTOR >::dh [private] |
Pointer to the dof handler.
const VECTOR& Functions::FEFieldFunction< dim, DH, VECTOR >::data_vector [private] |
A reference to the actual data vector.
const Mapping<dim>& Functions::FEFieldFunction< dim, DH, VECTOR >::mapping [private] |
A reference to the mapping being used.
DH::active_cell_iterator Functions::FEFieldFunction< dim, DH, VECTOR >::cell [mutable, private] |
The current cell in which we are evaluating.
const unsigned int Functions::FEFieldFunction< dim, DH, VECTOR >::n_components [private] |
Store the number of components of this function.
Reimplemented from Function< dim >.