HybridADRSolver::MatrixBasedSolver< dim > Class Template Reference

Matrix-based solver with hybrid MPI+threading parallelization. More...

#include <matrix_based_solver.h>

Inheritance diagram for HybridADRSolver::MatrixBasedSolver< dim >:

Collaboration diagram for HybridADRSolver::MatrixBasedSolver< dim >:

Public Member Functions
	MatrixBasedSolver (const ProblemInterface< dim > &problem, unsigned int degree, MPI_Comm comm, const SolverParameters &params=SolverParameters())
void	run (unsigned int n_refinements) override
SolverType	get_solver_type () const override
std::string	get_name () const override
const LADistributed::MPI::SparseMatrix &	get_system_matrix () const
double	compute_memory_usage () const
	Compute memory usage in MB.
double	compute_l2_error ()
	Compute the L2 error of the solution.
Public Member Functions inherited from HybridADRSolver::ParallelSolverBase< dim >
	ParallelSolverBase (MPI_Comm comm, SolverParameters params, bool construct_mg_hierarchy=false)
virtual	~ParallelSolverBase ()=default
const TimingResults &	get_timing_results () const

Protected Member Functions
void	setup_dofs () override
void	assemble_system () override
void	solve () override
void	output_results (unsigned int cycle) const override
Protected Member Functions inherited from HybridADRSolver::ParallelSolverBase< dim >
virtual void	setup_grid (unsigned int n_refinements)

Additional Inherited Members
Protected Attributes inherited from HybridADRSolver::ParallelSolverBase< dim >
MPI_Comm	mpi_communicator
unsigned int	n_mpi_processes
unsigned int	this_mpi_process
parallel::distributed::Triangulation< dim >	triangulation
std::unique_ptr< FiniteElement< dim > >	fe
DoFHandler< dim >	dof_handler
std::unique_ptr< Mapping< dim > >	mapping
AffineConstraints< double >	constraints
IndexSet	locally_owned_dofs
IndexSet	locally_relevant_dofs
SolverParameters	parameters
TimingResults	timing_results
ConditionalOStream	pcout
TimerOutput	computing_timer
bool	has_mg_hierarchy

Detailed Description

template<int dim>
class HybridADRSolver::MatrixBasedSolver< dim >

Matrix-based solver with hybrid MPI+threading parallelization.

• This solver:
• Explicitly assembles and stores the sparse system matrix
• Uses WorkStream for thread-parallel assembly
• Uses PETSc for distributed linear algebra
• Supports ILU and AMG preconditioning

• Template Parameters

  dim	Spatial dimension

Constructor & Destructor Documentation

MatrixBasedSolver()

template<int dim>

HybridADRSolver::MatrixBasedSolver< dim >::MatrixBasedSolver	(	const ProblemInterface< dim > &	problem,
		unsigned int	degree,
		MPI_Comm	comm,
		const SolverParameters &	params = SolverParameters() )

Constructor

Parameters

problem	The PDE problem to solve
degree	Polynomial degree of finite elements
comm	MPI communicator
params	Solver parameters

Member Function Documentation

assemble_system()

template<int dim>

void HybridADRSolver::MatrixBasedSolver< dim >::assemble_system ( )

overrideprotectedvirtual

Assemble the system (matrix and/or RHS)

Implements HybridADRSolver::ParallelSolverBase< dim >.

compute_l2_error()

template<int dim>

double HybridADRSolver::MatrixBasedSolver< dim >::compute_l2_error

(

)

Compute the L2 error of the solution.

compute_memory_usage()

template<int dim>

double HybridADRSolver::MatrixBasedSolver< dim >::compute_memory_usage

(

)

const

Compute memory usage in MB.

get_name()

template<int dim>

std::string HybridADRSolver::MatrixBasedSolver< dim >::get_name ( ) const

inlineoverridevirtual

Get a descriptive name for the solver

Implements HybridADRSolver::ParallelSolverBase< dim >.

get_solver_type()

template<int dim>

SolverType HybridADRSolver::MatrixBasedSolver< dim >::get_solver_type ( ) const

inlineoverridevirtual

Get the solver type (matrix-based or matrix-free)

Implements HybridADRSolver::ParallelSolverBase< dim >.

get_system_matrix()

template<int dim>

const LADistributed::MPI::SparseMatrix & HybridADRSolver::MatrixBasedSolver< dim >::get_system_matrix ( ) const

inline

Get the system matrix (for analysis/debugging)

output_results()

template<int dim>

void HybridADRSolver::MatrixBasedSolver< dim >::output_results ( unsigned int cycle ) const

overrideprotectedvirtual

Output results to files

Parameters

cycle

Current refinement cycle

Implements HybridADRSolver::ParallelSolverBase< dim >.

run()

template<int dim>

void HybridADRSolver::MatrixBasedSolver< dim >::run ( unsigned int n_refinements )

overridevirtual

Run the complete solve cycle

Parameters

n_refinements

Number of global mesh refinements

Implements HybridADRSolver::ParallelSolverBase< dim >.

setup_dofs()

template<int dim>

void HybridADRSolver::MatrixBasedSolver< dim >::setup_dofs ( )

overrideprotectedvirtual

Distribute degrees of freedom and setup constraints

Implements HybridADRSolver::ParallelSolverBase< dim >.

solve()

template<int dim>

void HybridADRSolver::MatrixBasedSolver< dim >::solve ( )

overrideprotectedvirtual

Solve the linear system

Implements HybridADRSolver::ParallelSolverBase< dim >.

---

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

• src/include/matrix_based/matrix_based_solver.h
• src/include/matrix_based/matrix_based_solver.cpp