terra::linalg::util Namespace Reference

Functions
template<typename ScalarType >
Eigen::SparseVector< ScalarType >	debug_sparse_assembly_vector_vec_q1_scalar (const VectorQ1Scalar< double > &vec)
	Assembles a VectorQ1Scalar into a Eigen::SparseVector. Likely not efficient - only use for debugging.
template<OperatorLike Operator>
Eigen::SparseMatrix< double >	debug_sparse_assembly_operator_vec_q1_scalar (const grid::shell::DistributedDomain &domain_src, Operator &A, VectorQ1Scalar< double > &tmp_src, VectorQ1Scalar< double > &tmp_dst)
	Brute force sparse matrix assembly for debugging purposes.

Function Documentation

debug_sparse_assembly_operator_vec_q1_scalar()

template<OperatorLike Operator>

Eigen::SparseMatrix< double > terra::linalg::util::debug_sparse_assembly_operator_vec_q1_scalar	(	const grid::shell::DistributedDomain &	domain_src,
		Operator &	A,
		VectorQ1Scalar< double > &	tmp_src,
		VectorQ1Scalar< double > &	tmp_dst
	)

Brute force sparse matrix assembly for debugging purposes.

Let the operator be representable by a nxm matrix. The assembly involves m matrix-vector multiplications, one with each of the standard basis vectors {e^1, ..., e^m} of R^m (per matrix-vector multiplication one column of the operator is 'assembled').

Needless to say, this is a very inefficient but flexible approach as any kind of operator can be assembled easily, but it involves many (n) matrix-vector multiplications.

Note

ONLY FOR TESTING/DEBUGGING PURPOSES - EXPECTED TO BE EXTREMELY SLOW

Only works when running in host space, in serial.

debug_sparse_assembly_vector_vec_q1_scalar()

template<typename ScalarType >

Eigen::SparseVector< ScalarType > terra::linalg::util::debug_sparse_assembly_vector_vec_q1_scalar

(

const VectorQ1Scalar< double > &

vec

)

Assembles a VectorQ1Scalar into a Eigen::SparseVector. Likely not efficient - only use for debugging.

Note

Only works in the serial case and on host space (only for debugging!).