mfem::GinkgoWrappers::GinkgoIterativeSolverBase Class Reference

#include <ginkgo.hpp>

Inheritance diagram for mfem::GinkgoWrappers::GinkgoIterativeSolverBase:

Public Member Functions
	GinkgoIterativeSolverBase (const std::string &exec_type, int print_iter, int max_num_iter, double RTOLERANCE, double ATOLERANCE)
virtual	~GinkgoIterativeSolverBase ()=default
void	initialize (const SparseMatrix *matrix)
void	apply (Vector &solution, const Vector &rhs)
void	solve (const SparseMatrix *matrix, Vector &solution, const Vector &rhs)

Protected Attributes
int	print_lvl
int	max_iter
double	rel_tol
double	abs_tol
double	final_norm
int	final_iter
int	converged
std::shared_ptr < gko::LinOpFactory >	solver_gen
std::shared_ptr < gko::stop::ResidualNormReduction <>::Factory >	residual_criterion
std::shared_ptr < gko::log::Convergence<> >	convergence_logger
std::shared_ptr< ResidualLogger<> >	residual_logger
std::shared_ptr < gko::stop::Combined::Factory >	combined_factory
std::shared_ptr< gko::Executor >	executor

Detailed Description

This class forms the base class for all of Ginkgo's iterative solvers. The various derived classes only take the additional data that is specific to them and solve the given linear system. The entire collection of solvers that Ginkgo implements is available at the Ginkgo documentation and manual pages, https://ginkgo-project.github.io/ginkgo/doc/develop.

Definition at line 192 of file ginkgo.hpp.

Constructor & Destructor Documentation

mfem::GinkgoWrappers::GinkgoIterativeSolverBase::GinkgoIterativeSolverBase	(	const std::string &	exec_type,
		int	print_iter,
		int	max_num_iter,
		double	RTOLERANCE,
		double	ATOLERANCE
	)

Constructor.

The exec_type defines the paradigm where the solution is computed. It is a string and the choices are "omp" , "reference" or "cuda". The respective strings create the respective executors as given below.

Ginkgo currently supports three different executor types:

• OmpExecutor specifies that the data should be stored and the associated operations executed on an OpenMP-supporting device (e.g. host CPU); ``` auto omp = gko::create<gko::OmpExecutor>(); ```
• CudaExecutor specifies that the data should be stored and the operations executed on the NVIDIA GPU accelerator; ``` if(gko::CudaExecutor::get_num_devices() > 0 ) { auto cuda = gko::create<gko::CudaExecutor>(); } ```
• ReferenceExecutor executes a non-optimized reference implementation, which can be used to debug the library. ``` auto ref = gko::create<gko::ReferenceExecutor>(); ```

The following code snippet demonstrates the using of the OpenMP executor to create a solver which would use the OpenMP paradigm to the solve the system on the CPU.

``` auto omp = gko::create<gko::OmpExecutor>(); using cg = gko::solver::Cg<>; auto solver_gen = cg::build() .with_criteria( gko::stop::Iteration::build().with_max_iters(20u).on(omp), gko::stop::ResidualNormReduction<>::build() .with_reduction_factor(1e-6) .on(omp)) .on(omp); auto solver = solver_gen->generate(system_matrix);

solver->apply(lend(rhs), lend(solution)); ```

Definition at line 31 of file ginkgo.cpp.

virtual mfem::GinkgoWrappers::GinkgoIterativeSolverBase::~GinkgoIterativeSolverBase ( )

virtualdefault

Destructor.

Member Function Documentation

void mfem::GinkgoWrappers::GinkgoIterativeSolverBase::apply	(	Vector &	solution,
		const Vector &	rhs
	)

Solve the linear system Ax=b. Dependent on the information provided by derived classes one of Ginkgo's linear solvers is chosen.

Definition at line 85 of file ginkgo.cpp.

void mfem::GinkgoWrappers::GinkgoIterativeSolverBase::initialize

(

const SparseMatrix *

matrix

)

Initialize the matrix and copy over its data to Ginkgo's data structures.

Definition at line 225 of file ginkgo.cpp.

void mfem::GinkgoWrappers::GinkgoIterativeSolverBase::solve	(	const SparseMatrix *	matrix,
		Vector &	solution,
		const Vector &	rhs
	)

Solve the linear system Ax=b. Dependent on the information provided by derived classes one of Ginkgo's linear solvers is chosen.

Definition at line 258 of file ginkgo.cpp.

Member Data Documentation

double mfem::GinkgoWrappers::GinkgoIterativeSolverBase::abs_tol

protected

Definition at line 278 of file ginkgo.hpp.

std::shared_ptr<gko::stop::Combined::Factory> mfem::GinkgoWrappers::GinkgoIterativeSolverBase::combined_factory

protected

The Ginkgo combined factory object is used to create a combined stopping criterion to be passed to the solver.

Definition at line 311 of file ginkgo.hpp.

int mfem::GinkgoWrappers::GinkgoIterativeSolverBase::converged

mutableprotected

Definition at line 281 of file ginkgo.hpp.

std::shared_ptr<gko::log::Convergence<> > mfem::GinkgoWrappers::GinkgoIterativeSolverBase::convergence_logger

protected

The Ginkgo convergence logger used to check for convergence and other solver data if needed.

Definition at line 299 of file ginkgo.hpp.

std::shared_ptr<gko::Executor> mfem::GinkgoWrappers::GinkgoIterativeSolverBase::executor

protected

The execution paradigm in Ginkgo. The choices are between gko::OmpExecutor, gko::CudaExecutor and gko::ReferenceExecutor and more details can be found in Ginkgo's documentation.

Definition at line 318 of file ginkgo.hpp.

int mfem::GinkgoWrappers::GinkgoIterativeSolverBase::final_iter

mutableprotected

Definition at line 280 of file ginkgo.hpp.

double mfem::GinkgoWrappers::GinkgoIterativeSolverBase::final_norm

mutableprotected

Definition at line 279 of file ginkgo.hpp.

int mfem::GinkgoWrappers::GinkgoIterativeSolverBase::max_iter

protected

Definition at line 276 of file ginkgo.hpp.

int mfem::GinkgoWrappers::GinkgoIterativeSolverBase::print_lvl

protected

Definition at line 275 of file ginkgo.hpp.

double mfem::GinkgoWrappers::GinkgoIterativeSolverBase::rel_tol

protected

Definition at line 277 of file ginkgo.hpp.

std::shared_ptr<gko::stop::ResidualNormReduction<>::Factory> mfem::GinkgoWrappers::GinkgoIterativeSolverBase::residual_criterion

protected

The residual criterion object that controls the reduction of the residual based on the tolerance set in the solver_control member.

Definition at line 293 of file ginkgo.hpp.

std::shared_ptr<ResidualLogger<> > mfem::GinkgoWrappers::GinkgoIterativeSolverBase::residual_logger

protected

The residual logger object used to check for convergence and other solver data if needed.

Definition at line 305 of file ginkgo.hpp.

std::shared_ptr<gko::LinOpFactory> mfem::GinkgoWrappers::GinkgoIterativeSolverBase::solver_gen

protected

The Ginkgo generated solver factory object.

Definition at line 286 of file ginkgo.hpp.

---

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

• linalg/ginkgo.hpp
• linalg/ginkgo.cpp