MatrixBatchEll< ExecSpace > Class Template Reference

Matrix class which is able to manage and solve a batch of sparse linear systems. More...

Inheritance diagram for MatrixBatchEll< ExecSpace >:

Public Member Functions
	MatrixBatchEll (const int batch_size, const int mat_size, const int non_zeros_per_row, std::optional< int > max_iter=std::nullopt, std::optional< double > res_tol=std::nullopt, std::optional< bool > logger=std::nullopt)
	The constructor for MatrixBatchEll class.
	MatrixBatchEll (Kokkos::View< int **, Kokkos::LayoutLeft, ExecSpace > cols_idx, Kokkos::View< double ***, Kokkos::LayoutStride, ExecSpace > batch_values, std::optional< int > max_iter=std::nullopt, std::optional< double > res_tol=std::nullopt, std::optional< bool > logger=std::nullopt)
	Constructor for MatrixBatchEll class.
std::pair< Kokkos::View< int **, Kokkos::LayoutLeft, ExecSpace >, Kokkos::View< double ***, Kokkos::LayoutStride, ExecSpace > >	get_batch_ell ()
	A function to get information about values and indices for the whole batch.
double	get_ell_element (int batch_idx, int line_idx, int non_zero_col_idx) const
	A getter function for a value located at a specified place.
void	set_ell_element (int batch_idx, int line_idx, int non_zero_col_idx, double aij)
	A setter function to modify a value located at a specified place.
void	setup_solver () final
	Perform a pre-process operation on the solver.
void	solve (BatchedRHS const b) const final
	Solve the batched linear problem Ax=b.
double	norm (int batch_idx) const
	A function returns the norm of a matrix located at batch_idx.
Public Member Functions inherited from MatrixBatch< ExecSpace >
virtual	~MatrixBatch ()=default
	Destruct.
std::size_t	size () const
	Get the size of the square matrix corresponding to a single batch in one of its dimensions.
std::size_t	batch_size () const
	Get the batch size of the linear problem.

Additional Inherited Members
Public Types inherited from MatrixBatch< ExecSpace >
using	BatchedRHS = Kokkos::View< double **, Kokkos::LayoutRight, ExecSpace >
	The type of a Kokkos::View storing batched right-hand sides. Second dimenion is batch dimension.
Protected Member Functions inherited from MatrixBatch< ExecSpace >
	MatrixBatch (const std::size_t batch_size, const std::size_t mat_size)
	The constructor for MatrixBatch class.

Detailed Description

template<class ExecSpace>
class MatrixBatchEll< ExecSpace >

Matrix class which is able to manage and solve a batch of sparse linear systems.

Executes on either CPU or GPU. It takes advantage of the sparse structure, and the only batched solver available in Ginkgo : Stabilized Bicg. The sparsity pattern is assumed to be the same for all matrices. ie the non-zero components are located at the same places for all matrices. This class uses the ELL storage format which needs two 1D arrays, one stores values the other column indices. The class returns these arrays (as Kokkos views) with the get_batch_idx_and_vals function, it is then possible to fill them outside the class. Tolerance and maximal number of iterations, which are parameters for the iterative solver, are set in the constructor. It is possibile to get convergence information by activating the logger at constructor call.

Template Parameters

ExecSpace

Execution space,needed by Kokkos for allocations and parallelism. The simplest choice is to follow Kokkos, for that: specify Kokkos::DefaultExecutionSpace

Constructor & Destructor Documentation

MatrixBatchEll() [1/2]

template<class ExecSpace >

MatrixBatchEll< ExecSpace >::MatrixBatchEll ( const int  batch_size, const int  mat_size, const int  non_zeros_per_row, std::optional< int >  max_iter = std::nullopt, std::optional< double >  res_tol = std::nullopt, std::optional< bool >  logger = std::nullopt  )

inlineexplicit

The constructor for MatrixBatchEll class.

Parameters

[in]	batch_size	Number of linear systems to solve.
[in]	mat_size	Common matrix size for all the systems.
[in]	non_zeros_per_row	number of non zero components per line.
[in]	max_iter	maximal number of iterations for the solver
[in]	res_tol	residual tolerance parameter, to ensure convergence. Be careful! the relative residual provided here, will be used as "implicit residual" in ginkgo solver.
[in]	logger	boolean parameter for saving log informations such residual and interations count.

MatrixBatchEll() [2/2]

template<class ExecSpace >

MatrixBatchEll< ExecSpace >::MatrixBatchEll ( Kokkos::View< int **, Kokkos::LayoutLeft, ExecSpace >  cols_idx, Kokkos::View< double ***, Kokkos::LayoutStride, ExecSpace >  batch_values, std::optional< int >  max_iter = std::nullopt, std::optional< double >  res_tol = std::nullopt, std::optional< bool >  logger = std::nullopt  )

inlineexplicit

Constructor for MatrixBatchEll class.

Parameters

[in]	cols_idx	A Kokkos view which stores the column indices of non-zero components.
[in]	batch_values	A Kokkos view which stores the values of non-zero elements.
[in]	max_iter	maximal number of iterations for the solver, default 500.
[in]	res_tol	residual tolerance parameter, to ensure convergence. Be careful! The residual provided here, set as relative residual, will be used as "implicit residual" in ginkgo solver. Default value is set to 1e-15.
[in]	logger	bolean parameter to save logger information. Default value false.

Member Function Documentation

get_batch_ell()

template<class ExecSpace >

std::pair< Kokkos::View< int **, Kokkos::LayoutLeft, ExecSpace >, Kokkos::View< double ***, Kokkos::LayoutStride, ExecSpace > > MatrixBatchEll< ExecSpace >::get_batch_ell ( )

inline

A function to get information about values and indices for the whole batch.

Data is managed by two Kokkos Views stored on the host.

Returns

idx_view Column indices for the non-zero values.

vals_view The non-zero values.

get_ell_element()

template<class ExecSpace >

double MatrixBatchEll< ExecSpace >::get_ell_element ( int  batch_idx, int  line_idx, int  non_zero_col_idx  ) const

inline

A getter function for a value located at a specified place.

Parameters

[in]	batch_idx	Index in the batch.
[in]	line_idx	Line index inside the matrix.
[in]	non_zero_col_idx	Non-zero index element in the line.

Returns

value of the component.

set_ell_element()

template<class ExecSpace >

void MatrixBatchEll< ExecSpace >::set_ell_element ( int  batch_idx, int  line_idx, int  non_zero_col_idx, double  aij  )

inline

A setter function to modify a value located at a specified place.

Parameters

[in]	batch_idx	Index in the batch.
[in]	line_idx	Line index inside the matrix.
[in]	non_zero_col_idx	Non-zero index element in the line.
[in]	aij	New value.

setup_solver()

template<class ExecSpace >

void MatrixBatchEll< ExecSpace >::setup_solver ( )

inlinefinalvirtual

Perform a pre-process operation on the solver.

Must be called after filling the matrix.

It uses parameters like maximum number of iterations and tolerance are used to instantiate a Ginkgo solver.

The stopping criterion is a reduction factor ||Ax-b||/||b||<tol with max_iter maximum iterations.

Implements MatrixBatch< ExecSpace >.

solve()

template<class ExecSpace >

void MatrixBatchEll< ExecSpace >::solve ( BatchedRHS const  b ) const

inlinefinalvirtual

Solve the batched linear problem Ax=b.

Parameters

[in,out]

b

A 2D Kokkos::View storing the batched right-hand sides of the problem and receiving the corresponding solutions.

Implements MatrixBatch< ExecSpace >.

norm()

template<class ExecSpace >

double MatrixBatchEll< ExecSpace >::norm ( int  batch_idx ) const

inline

A function returns the norm of a matrix located at batch_idx.

Parameters

[in]

batch_idx

integer, index of the matrix in the batch.

Returns

value of the matrix infinite-norm.

---

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

• /home/runner/work/gyselalibxx/gyselalibxx/code_branch/vendor/sll/include/sll/matrix_batch_ell.hpp