Finite element helpers

Collaboration diagram for Finite element helpers:

Namespaces
namespace	ryujin::DoFRenumbering
namespace	ryujin::DoFTools

Functions
template<typename Number >
void	ryujin::transform_to_local_range (const dealii::Utilities::MPI::Partitioner &partitioner, dealii::AffineConstraints< Number > &affine_constraints)
template<typename VECTOR >
void	ryujin::transform_to_local_range (const dealii::Utilities::MPI::Partitioner &partitioner, VECTOR &vector)
template<int dim>
unsigned int	ryujin::DoFRenumbering::export_indices_first (dealii::DoFHandler< dim > &dof_handler, const MPI_Comm &mpi_communicator, const unsigned int n_locally_internal, const std::size_t warp_size)
template<int dim>
unsigned int	ryujin::DoFRenumbering::inconsistent_strides_last (dealii::DoFHandler< dim > &dof_handler, const dealii::DynamicSparsityPattern &sparsity, const unsigned int n_locally_internal, const std::size_t warp_size)
template<int dim>
unsigned int	ryujin::DoFRenumbering::internal_range (dealii::DoFHandler< dim > &dof_handler, const dealii::DynamicSparsityPattern &sparsity, const std::size_t warp_size)
template<int dim, typename Number , typename SPARSITY >
void	ryujin::DoFTools::make_extended_sparsity_pattern_dg (const dealii::DoFHandler< dim > &dof_handler, SPARSITY &dsp, const dealii::AffineConstraints< Number > &affine_constraints, bool keep_constrained)
template<int dim, typename Number , typename Number2 , bool add>
void	ryujin::internal::integrate_add_tensor_product_value_linear (const Number2 &value, Number2 *values, const dealii::Point< dim, Number > &p)

Detailed Description

Some helper functions for local index handling, dof renumbering, sparsity pattern creation, reference cells, and tensor product point kernels (include/ryujin/discretization).

Function Documentation

transform_to_local_range() [1/2]

template<typename Number >

void ryujin::transform_to_local_range	(	const dealii::Utilities::MPI::Partitioner &	partitioner,
		dealii::AffineConstraints< Number > &	affine_constraints
	)

Given a partitioner the function converts an AffineConstraints object affine_constraints from the usual global numbering used in deal.II to an MPI-rank local numbering \([0,n]\), where \(n\) is the number of locally relevant degrees of freedom.

Definition at line 29 of file local_index_handling.h.

transform_to_local_range() [2/2]

template<typename VECTOR >

void ryujin::transform_to_local_range	(	const dealii::Utilities::MPI::Partitioner &	partitioner,
		VECTOR &	vector
	)

Given a partitioner the function translates each element of a given vector vector from the usual global numbering used in deal.II to an MPI-rank local numbering \([0,n]\), where \(n\) is the number of locally relevant degrees of freedom.

Definition at line 69 of file local_index_handling.h.

export_indices_first()

template<int dim>

unsigned int ryujin::DoFRenumbering::export_indices_first	(	dealii::DoFHandler< dim > &	dof_handler,
		const MPI_Comm &	mpi_communicator,
		const unsigned int	n_locally_internal,
		const std::size_t	warp_size
	)

Reorder all (strides of) locally internal indices that contain export indices to the start of the index range.

This renumbering requires MPI communication in order to determine the set of export indices.

Definition at line 103 of file local_index_handling.h.

References ryujin::warp_size.

inconsistent_strides_last()

template<int dim>

unsigned int ryujin::DoFRenumbering::inconsistent_strides_last	(	dealii::DoFHandler< dim > &	dof_handler,
		const dealii::DynamicSparsityPattern &	sparsity,
		const unsigned int	n_locally_internal,
		const std::size_t	warp_size
	)

Reorder all strides of inconsistent locally internal indices to the end of the locally internal index range.

Definition at line 216 of file local_index_handling.h.

References ryujin::warp_size.

internal_range()

template<int dim>

unsigned int ryujin::DoFRenumbering::internal_range	(	dealii::DoFHandler< dim > &	dof_handler,
		const dealii::DynamicSparsityPattern &	sparsity,
		const std::size_t	warp_size
	)

Reorder indices:

In order to traverse over multiple rows of a (to be constructed) sparsity pattern simultaneously (using SIMD instructions, or a warp of threads on the device) we reorder all locally owned degrees of freedom to ensure that a local index range \([0, \text{n_locally_internal_}) \subset [0, \text{n_locally_owned})\) is available that groups dofs with same stencil size in "warps" of warp_size consecutive indices.

Returns the right boundary n_internal of the internal index range.

Definition at line 316 of file local_index_handling.h.

References ryujin::warp_size.

make_extended_sparsity_pattern_dg()

template<int dim, typename Number , typename SPARSITY >

void ryujin::DoFTools::make_extended_sparsity_pattern_dg	(	const dealii::DoFHandler< dim > &	dof_handler,
		SPARSITY &	dsp,
		const dealii::AffineConstraints< Number > &	affine_constraints,
		bool	keep_constrained
	)

Given a dof_handler, and constraints affine_constraints this function creates an extended sparsity pattern for the discontinuous Galerkin formulation that also includes locally relevant to locally relevant couplings.

Definition at line 407 of file local_index_handling.h.

integrate_add_tensor_product_value_linear()

template<int dim, typename Number , typename Number2 , bool add>

void ryujin::internal::integrate_add_tensor_product_value_linear ( const Number2 &  value, Number2 *  values, const dealii::Point< dim, Number > &  p  )

inline

This is a copy of the integrate_add_tensor_product_value_linear() function shipped with deal.II that fixes a compilation error when instantiating with float.

Definition at line 33 of file tensor_product_point_kernels.h.

References ryujin::add().

Referenced by ryujin::internal::integrate_tensor_product_value().