algebra::BlockDiagPrecond< BS > Class Template Reference

#include <block_precond.h>

Public Member Functions
	BlockDiagPrecond (const SparseMatrix &A, const int nbBlocks, const std::vector< int > &ld)
void	apply (const std::vector< double > &x, std::vector< double > &y) const

Private Attributes
std::vector< Eigen::Matrix< double, BS, BS > >	inv

Detailed Description

template<int BS>
class algebra::BlockDiagPrecond< BS >

block diagonal preconditioner: inverse of the BS x BS diagonal block of each node, for a system whose unknowns are numbered node by node (unknown BS*i + d is the component d of the node i). For the spin diffusion problem (BS = 3), this block holds the relaxation and the precession coupling s x u, local to each node, which the diagonal preconditioner misses (same preconditioner as the BLOCK3 of st-feeLLGood). The rows and columns of the Dirichlet components are replaced by the identity before inversion, so that the preconditioner does not mix the imposed components with the free ones.

Constructor & Destructor Documentation

BlockDiagPrecond()

template<int BS>

algebra::BlockDiagPrecond< BS >::BlockDiagPrecond ( const SparseMatrix &  A, const int  nbBlocks, const std::vector< int > &  ld  )

inline

computes the inverse of the diagonal block of each node of A; ld: indices of the Dirichlet components

Member Function Documentation

apply()

template<int BS>

void algebra::BlockDiagPrecond< BS >::apply ( const std::vector< double > &  x, std::vector< double > &  y  ) const

inline

y = M x, M being the block diagonal preconditioner

Member Data Documentation

inv

template<int BS>

std::vector<Eigen::Matrix<double, BS, BS> > algebra::BlockDiagPrecond< BS >::inv

private

inverses of the diagonal blocks

---

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

• src/algebra/block_precond.h