The local real ilu preconditioner, only work for sequential CSR matrix. More...

#include <parallel_gemslr.hpp>

Public Member Functions
int	SetupPartialILUT (VectorType &x, VectorType &rhs)
	Setup the local partial ILUT. More...

	ParallelGemslrClass ()
	The constructor of precondioner class. More...

virtual int	Clear ()
	Free the current precondioner. More...

virtual	~ParallelGemslrClass ()
	The destructor of precondioner class. More...

	ParallelGemslrClass (const ParallelGemslrClass< MatrixType, VectorType, DataType > &precond)
	The copy constructor of ParallelGemslrClass. More...

	ParallelGemslrClass (ParallelGemslrClass< MatrixType, VectorType, DataType > &&precond)
	The move constructor of ParallelGemslrClass. More...

ParallelGemslrClass< MatrixType, VectorType, DataType > &	operator= (const ParallelGemslrClass< MatrixType, VectorType, DataType > &precond)
	The operator = of ParallelGemslrClass. More...

ParallelGemslrClass< MatrixType, VectorType, DataType > &	operator= (ParallelGemslrClass< MatrixType, VectorType, DataType > &&precond)
	The operator = of ParallelGemslrClass. More...

int	CheckParameters ()
	Check the input parameters, avoid invalid settings. More...

virtual int	Setup (VectorType &x, VectorType &rhs)
	Setup the precondioner phase. Will be called by the solver if not called directly. More...

int	SetupPermutation ()
	Setup the permutation of the GeMSLR. More...

int	SetupPermutationRKway (MatrixType &A, int nlev_setup, int &nlev_max, int &nlev_used, vector_int &map_v, vector_int &mapptr_v)
	Setup the Recursive KWay partition of the GeMSLR. More...

int	SetupPermutationND (MatrixType &A, int nlev_setup, int &nlev_max, int &nlev_used, vector_int &map_v, vector_int &mapptr_v)
	Setup the Nested Dissection partition of the GeMSLR. More...

int	SetupPermutationBuildLevelStructure (MatrixType &A, int level_start, vector_int &map_v, vector_int &mapptr_v)
	Setup the level structure of the GeMSLR. More...

int	SetupBSolve (VectorType &x, VectorType &rhs)
	Setup the solve of B matrices of the GeMSLR. More...

int	SetupBSolveILUT (VectorType &x, VectorType &rhs, int level)
	Setup the solve of B matrices of the GeMSLR with ILUT. More...

int	SetupBSolveILUK (VectorType &x, VectorType &rhs, int level)
	Setup the solve of B matrices of the GeMSLR with ILUK. More...

int	SetupBSolveGemslr (VectorType &x, VectorType &rhs, int level)
	Setup the solve of B matrices of the GeMSLR with GeMSLR. More...

int	SetupLowRank (VectorType &x, VectorType &rhs)
	Setup the low-rank part of the GeMSLR. More...

int	SetupLowRankSubspaceIteration (VectorType &x, VectorType &rhs, int level, int option)
	Setup the low-rank part of the GeMSLR with subspace iteration. More...

int	SetupLowRankNoRestart (VectorType &x, VectorType &rhs, int level, int option)
	Setup the low-rank part of the GeMSLR with arnodi no-restart. More...

int	SetupLowRankThickRestart (VectorType &x, VectorType &rhs, int level, int option)
	Setup the low-rank part of the GeMSLR with arnodi thick-restart. More...

int	SetupLowRankBuildLowRank (VectorType &x, VectorType &rhs, DenseMatrixClass< DataType > &V, DenseMatrixClass< DataType > &H, int m, int rank, int level, int option)
	Given the V and H from Arnoldi, compute the final low-rank correction of S^{-1} - C^{-1}. More...

virtual int	Solve (VectorType &x, VectorType &rhs)
	Solve phase. Call this function after Setup. Solve with cusparse if unified memory/device memory is used. More...

int	SolveLevelGemslr (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelGemslrU (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelGemslrMul (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelEsmslr (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelEsmslrU (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelEsmslrMul (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveLevelPslr (VectorType &x_out, VectorType &rhs_in, int level, bool doperm)
	Solve starting from a certain level. More...

int	SolveB (VectorType &x, VectorType &rhs, int option, int level)
	Solve with B on a certain level. More...

int	SolveApplyLowRankLevel (VectorType &x, VectorType &rhs, int level)
	Apply the low-rank update on a certain level. More...

int	GetNumRows (int level)
	Get the local number of rows on certain level for the low-rank part. More...

int	EBFCMatVec (int level, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec function y = Gx = I - Ci(Ci\x) + Ei(UBi(LBi(Fi(Ci\x). Note that alpha and beta are untouched. More...

int	RAPEBFCMatVec (int level, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec function y = G*x = (2EB^{-1}F + EB^{-1}BB^{-1}F)C^{-1}x. Note that alpha and beta are untouched. More...

int	SCMatVec (int level, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec function y = Gx = C\Sx-x. Note that alpha and beta are untouched. More...

int	ACMatVec (int level, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec function y = x-AM^{-1}x. Note that alpha and beta are untouched. More...

int	PCLRMatVec (int level, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec function y = (EsC^{-1})^m*x. Note that alpha and beta are untouched. More...

int	SchurMatVec (int level, int option, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec with the Schur Complement. More...

int	RAPMatVec (int level, int option, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec with the Schur Complement via RAP. More...

int	CMatVec (int level, int option, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec with the C on the current level. More...

int	BMatVec (int level, int option, char trans, const DataType &alpha, VectorType &x, const DataType &beta, VectorType &y)
	The matvec with the B on the current level. More...

int	GetSize ()
	Get the size of the problem. More...

virtual long int	GetNumNonzeros ()
	Get the total number of nonzeros the ILU. More...

long int	GetNumNonzeros (long int &nnz_bsolver, long int &nnz_lr)
	Get the total number of nonzeros the ILU. More...

int	PlotPatternGnuPlot (const char *datafilename)
	Plot the pattern of the parallel gemslr using gnuplot. Similar to spy in the MATLAB. Function for testing purpose. More...

virtual int	SetLocalGemslr (GemslrClass< CsrMatrixClass< DataType >, SequentialVectorClass< DataType >, DataType > &gemslr)
	Setup with parameter array. This is the helper function to set the local gemslr for B solve. More...

virtual int	SetLowerGemslr (ParallelGemslrClass< MatrixType, VectorType, DataType > &gemslr)
	Setup with parameter array. This is the helper function to set the lower level gemslr for explicite Schur solve. More...

virtual int	SetWithParameterArray (double *params)
	Setup with parameter array. More...

int	SetGlobalPrecondOption (int option)
	Set the global precond option. 0: BJ; 1: ISCHUR; 2: ESCHUR; 3: MLEV. More...

int	SetPartitionOption (int option)
	Set the global partition option. 0: ND; 1: RKway. More...

int	SetBPartitionOption (int option)
	Set the B part partition option. 0: ND; 1: RKway. More...

int	SetPermutationOption (int option)
	Set the global permutation option. 0: No; 1: RCM; 2: AMD. More...

int	SetBPermutationOption (int option)
	Set the B part permutation option. 0: No; 1: RCM; 2: AMD. More...

int	SetNumLevels (int option)
	Set the global target number of levels of GeMSLR. More...

int	SetBNumLevels (int option)
	Set the B part target number of levels of GeMSLR. More...

int	SetNumSubdomains (int option)
	Set the global target number of subdomains on each level of GeMSLR. More...

int	SetBNumSubdomains (int option)
	Set the B part target number of subdomains on each level of GeMSLR. More...

int	SetMinimalNumberSubdomains (int option)
	Set the global minimal number of subdomains on each level of GeMSLR. More...

int	SetBMinimalNumberSubdomains (int option)
	Set the B part minimal number of subdomains on each level of GeMSLR. More...

int	SetNumberSubdomainsReduceFactor (int option)
	Set the global reduce factor of subdomains on each level of GeMSLR. More...

int	SetBNumberSubdomainsReduceFactor (int option)
	Set the B part reduce factor of subdomains on each level of GeMSLR. More...

int	SetSeperatorOption (bool option)
	Set the global separator option of GeMSLR. More...

int	SetBSeperatorOption (bool option)
	Set the B part separator option of GeMSLR. More...

int	SetGlobalPartitionOption (bool option)
	Set the global partition option of GeMSLR. More...

int	SetInnerIterationOption (bool option)
	Set the inner iteration option of GeMSLR. More...

template<typename T >
int	SetInnerIterationThreshold (T option)
	Set the stop threshold of inner iteration of GeMSLR. More...

int	SetInnerIterationMaxNumberIterations (int option)
	Set the max number of iterations of inner iteration of GeMSLR. More...

int	SetSolveOption (int option)
	Set the solve option. 0: additive LU solve; 1: additive L solve; 2: multi-solve. More...

int	SetLowRankOptionTopLevel (int option)
	Set the low-rank option on the top level. 0: Standard. 1: Thick-restart. More...

int	SetLowRankOptionOtherLevels (int option)
	Set the low-rank option on other levels. 0: Standard. 1: Thick-restart. More...

int	SetLowRankOptionA (int option)
	Set the low-rank option on A. 0: Standard. 1: Thick-restart. More...

int	SetBLowRankOptionTopLevel (int option)
	Set the B part low-rank option on the top level. 0: Standard. 1: Thick-restart. More...

int	SetBLowRankOptionOtherLevels (int option)
	Set the B part low-rank option on other levels. 0: Standard. 1: Thick-restart. More...

int	SetLowRankRandomInitGuess (bool option)
	Set if we use random initial guess for Arnoldi. Otherwise we use 1 as initial guess. More...

int	SetBLowRankRandomInitGuess (bool option)
	Set if we use random initial guess for Arnoldi in the B part. Otherwise we use 1 as initial guess. More...

template<typename T >
int	SetLowRankFactorTopLevel (T option)
	Set the low-rank factor on the top level. The actuall computed number of low-rank terms is rank * factor >= rank. More...

template<typename T >
int	SetLowRankFactorOtherLevels (T option)
	Set the low-rank factor on the other levels. The actuall computed number of low-rank terms is rank * factor >= rank. More...

template<typename T >
int	SetLowRankFactorA (T option)
	Set the low-rank factor on A. The actuall computed number of low-rank terms is rank * factor >= rank. More...

template<typename T >
int	SetBLowRankFactorTopLevel (T option)
	Set the B part low-rank factor on the top level. The actuall computed number of low-rank terms is rank * factor >= rank. More...

template<typename T >
int	SetBLowRankFactorOtherLevels (T option)
	Set the B part low-rank factor on the other levels. The actuall computed number of low-rank terms is rank * factor >= rank. More...

template<typename T >
int	SetLowRankRanksTopLevel (T option)
	Set the target number of low-rank terms on the top level. More...

template<typename T >
int	SetLowRankRanksOtherLevels (T option)
	Set the target number of low-rank terms on the other levels. More...

template<typename T >
int	SetLowRankA (T option)
	Set the target number of low-rank terms for A. More...

template<typename T >
int	SetBLowRankRanksTopLevel (T option)
	Set the B part target number of low-rank terms on the top level. More...

template<typename T >
int	SetBLowRankRanksOtherLevels (T option)
	Set the B part target number of low-rank terms on the other levels. More...

template<typename T >
int	SetLowRankArnoldiFactorTopLevel (T option)
	Set the Arnoldi factor on the top level. m steps for arnoldi is rank * rank_factor * arnoldi_factor. More...

template<typename T >
int	SetLowRankArnoldiFactorOtherLevels (T option)
	Set the Arnoldi factor on other levels. m steps for arnoldi is rank * rank_factor * arnoldi_factor. More...

template<typename T >
int	SetLowRankArnoldiFactorA (T option)
	Set the Arnoldi factor on A. m steps for arnoldi is rank * rank_factor * arnoldi_factor. More...

template<typename T >
int	SetBLowRankArnoldiFactorTopLevel (T option)
	Set the B part Arnoldi factor on the top level. m steps for arnoldi is rank * rank_factor * arnoldi_factor. More...

template<typename T >
int	SetBLowRankArnoldiFactorOtherLevels (T option)
	Set the B part Arnoldi factor on other levels. m steps for arnoldi is rank * rank_factor * arnoldi_factor. More...

int	SetLowRankMaxNumberIterationsTopLevel (int option)
	Set max restarts of thick-restart Arnoldi on the top level. More...

int	SetLowRankMaxNumberIterationsOtherLevels (int option)
	Set max restarts of thick-restart Arnoldi on other levels. More...

int	SetLowRankMaxNumberIterationsA (int option)
	Set max restarts of thick-restart Arnoldi on A. More...

int	SetBLowRankMaxNumberIterationsTopLevel (int option)
	Set B part max restarts of thick-restart Arnoldi on the top level. More...

int	SetBLowRankMaxNumberIterationsOtherLevels (int option)
	Set B part max restarts of thick-restart Arnoldi on other levels. More...

template<typename T >
int	SetLowRankThresholdTopLevel (T option)
	Set max restarts of thick-restart Arnoldi on the top level. More...

template<typename T >
int	SetLowRankThresholdOtherLevels (T option)
	Set max restarts of thick-restart Arnoldi on other levels. More...

template<typename T >
int	SetLowRankThresholdA (T option)
	Set max restarts of thick-restart Arnoldi on A. More...

template<typename T >
int	SetBLowRankThresholdTopLevel (T option)
	Set B part max restarts of thick-restart Arnoldi on the top level. More...

template<typename T >
int	SetBLowRankThresholdOtherLevels (T option)
	Set B part max restarts of thick-restart Arnoldi on other levels. More...

template<typename T >
int	SetIluDropTolB (T option)
	Set the threshold for ILUT of the B part. More...

template<typename T >
int	SetIluDropTolC (T option)
	Set the threshold for ILUT of the last level. More...

template<typename T >
int	SetIluDropTolS (T option)
	Set the threshold for ILUT of the S part. More...

template<typename T >
int	SetIluDropTolEF (T option)
	Set the threshold for ILUT of the EF part. More...

template<typename T >
int	SetBIluDropTolB (T option)
	Set the recursive GeMSLR B part threshold for ILUT of the global B part. More...

template<typename T >
int	SetBIluDropTolC (T option)
	Set the recursive GeMSLR last level threshold for ILUT of the global B part. More...

int	SetIluMaxRowNnzB (int option)
	Set the maxinum number of nonzeros for ILUT of the B part. More...

int	SetIluMaxRowNnzC (int option)
	Set the maxinum number of nonzeros ILUT of the last level. More...

int	SetIluMaxRowNnzS (int option)
	Set the maxinum number of nonzeros ILUT of the S part. More...

int	SetBIluMaxRowNnzB (int option)
	Set the recursive GeMSLR B part maxinum number of nonzeros ILUT. More...

int	SetBIluMaxRowNnzC (int option)
	Set the recursive GeMSLR last level maxinum number of nonzeros ILUT. More...

int	SetIluFillLevelB (int option)
	Set the fill level for ILUK of the B part. More...

int	SetIluFillLevelC (int option)
	Set the fill level for ILUK of the last level. More...

int	SetBIluFillLevelB (int option)
	Set the fill level for ILUK of the B part. More...

int	SetBIluFillLevelC (int option)
	Set the fill level for ILUK of the last level. More...

int	SetPolyOrder (int option)
	Set poly order for Poly solve of the B part. More...

int	SetBPolyOrder (int option)
	Set poly order for Poly solve of the B part of the recursive GeMSLR. More...

int	SetPreconditionerOption1 (int option)
	Set top level preconditioner. More...

int	SetPreconditionerOption1Levels (int option)
	Set top level preconditioner apply levels. More...

int	SetPreconditionerOption2 (int option)
	Set mid level preconditioner. More...

int	SetPreconditionerOptionB (int option)
	Set all top levels preconditioner. More...

int	SetSmoothOptionB (int option)
	Set all top levels smoother. More...

int	SetPreconditionerOptionC (int option)
	Set last level preconditioner. More...

int	SetIluResidualIters (int residual_iters)
	Set the number of residual iterations of B solves in the setup phase. Set to <= 1 to turn off. More...

int	SetIluComplexShift (bool complex_shift)
	Set if we turn on the complex shift or not (complex version only). More...

virtual int	SetSolveLocation (const int &location)
	Set the data location that the preconditioner apply to. More...

virtual int	MoveData (const int &location)
	Move the preconditioner to another location. Only can be called after Setup. More...

int	GetSolvePhase ()
	Get the solve phase. More...

Public Attributes
int	_global_precond_option
	The Schur complement option. More...

std::vector< ParallelGemslrLevelClass< MatrixType, VectorType, DataType > >	_levs_v
	Vector holding the level struct for all levels starting from the second level. The size of this vector is equal to _nlev_used, and is the number of true levels. More...

ParallelGemslrLevelClass< MatrixType, VectorType, DataType >	_lev_A
	The structure for the A level low-rank correction. More...

Detailed Description

template<class MatrixType, class VectorType, typename DataType>
class pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >

The local real ilu preconditioner, only work for sequential CSR matrix. Matrix type is the type of the matrix, VectorType is the type of the vector. DataType is the data type.

Constructor & Destructor Documentation

◆ ParallelGemslrClass() [1/3]

template<class MatrixType , class VectorType , typename DataType >

template pargemslr::precond_gemslr_csr_par_complexd::ParallelGemslrClass ( )

The constructor of precondioner class.

◆ ~ParallelGemslrClass()

template<class MatrixType , class VectorType , typename DataType >

template pargemslr::precond_gemslr_csr_par_complexd::~ParallelGemslrClass ( )

virtual

The destructor of precondioner class.

◆ ParallelGemslrClass() [2/3]

template<class MatrixType , class VectorType , typename DataType >

pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::ParallelGemslrClass ( const ParallelGemslrClass< MatrixType, VectorType, DataType > & precond )

The copy constructor of ParallelGemslrClass.

◆ ParallelGemslrClass() [3/3]

template<class MatrixType , class VectorType , typename DataType >

pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::ParallelGemslrClass ( ParallelGemslrClass< MatrixType, VectorType, DataType > && precond )

The move constructor of ParallelGemslrClass.

Member Function Documentation

◆ ACMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::ACMatVec	(	int	level,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec function y = x-A*M^{-1}*x. Note that alpha and beta are untouched.
We are finding AM^{-1}(I-X)^{-1} = I -> X = I - A^M{-1}. After that, we can apply the preconditioner as M^{-1}(I-X)^{-1}x.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ BMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::BMatVec	(	int	level,
		int	option,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec with the B on the current level.

Parameters

[in]	level	The matvec level.
[in]	option	The matvec option. 0: standard; 1: use -C_offd instad of C.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ CheckParameters()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::CheckParameters ( )

Check the input parameters, avoid invalid settings.

◆ Clear()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::Clear ( )

virtual

Free the current precondioner.

Returns: Return error message.

◆ CMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::CMatVec	(	int	level,
		int	option,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec with the C on the current level.

Parameters

[in]	level	The matvec level.
[in]	option	The matvec option. 0: standard; 1: use -C_offd instad of C.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ EBFCMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::EBFCMatVec	(	int	level,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec function y = G*x = I - Ci*(Ci\x) + Ei*(UBi(LBi(Fi*(Ci\x). Note that alpha and beta are untouched.

Parameters

[in]	level	The matvec level.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ GetNumNonzeros() [1/2]

template<class MatrixType , class VectorType , typename DataType >

template long int pargemslr::precond_gemslr_csr_par_complexd::GetNumNonzeros ( )

virtual

Get the total number of nonzeros the ILU.

Returns: Return the total number of nonzeros the ILU.

◆ GetNumNonzeros() [2/2]

template<class MatrixType , class VectorType , typename DataType >

template long int pargemslr::precond_gemslr_csr_par_complexd::GetNumNonzeros	(	long int &	nnz_bsolver,
		long int &	nnz_lr
	)

Get the total number of nonzeros the ILU.

Returns: Return the total number of nonzeros the ILU.

◆ GetNumRows()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::GetNumRows ( int level )

Get the local number of rows on certain level for the low-rank part.

Returns: Return the length of the matrix.

◆ GetSize()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::GetSize ( )

Get the size of the problem.

Returns: Return the problem size.

◆ GetSolvePhase()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::GetSolvePhase ( )

inline

Get the solve phase.

Returns: Return the solve phase.

◆ MoveData()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::MoveData ( const int & location )

virtual

Move the preconditioner to another location. Only can be called after Setup.

Parameters

[in] location The target solver location.

Returns: Return error message.

◆ operator=() [1/2]

template<class MatrixType , class VectorType , typename DataType >

ParallelGemslrClass< MatrixType, VectorType, DataType > & pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::operator= ( const ParallelGemslrClass< MatrixType, VectorType, DataType > & precond )

The operator = of ParallelGemslrClass.

◆ operator=() [2/2]

template<class MatrixType , class VectorType , typename DataType >

ParallelGemslrClass< MatrixType, VectorType, DataType > & pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::operator= ( ParallelGemslrClass< MatrixType, VectorType, DataType > && precond )

The operator = of ParallelGemslrClass.

◆ PCLRMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::PCLRMatVec	(	int	level,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec function y = (EsC^{-1})^m*x. Note that alpha and beta are untouched.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ PlotPatternGnuPlot()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::PlotPatternGnuPlot ( const char * datafilename )

Plot the pattern of the parallel gemslr using gnuplot. Similar to spy in the MATLAB. Function for testing purpose.

Parameters

[in] datafilename The filename of the temp file holding the data.

Returns: Return error message.

◆ RAPEBFCMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::RAPEBFCMatVec	(	int	level,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec function y = G*x = (2EB^{-1}F + EB^{-1}BB^{-1}F)C^{-1}x. Note that alpha and beta are untouched.

Parameters

[in]	level	The matvec level.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ RAPMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::RAPMatVec	(	int	level,
		int	option,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec with the Schur Complement via RAP.

Parameters

[in]	level	The matvec level.
[in]	option	The matvec option. 0: use B_solve; 1: use B_precond.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ SchurMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SchurMatVec	(	int	level,
		int	option,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec with the Schur Complement.

Parameters

[in]	level	The matvec level.
[in]	option	The matvec option. 0: standard; 1: use -C_offd instad of C.
[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ SCMatVec()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SCMatVec	(	int	level,
		char	trans,
		const DataType &	alpha,
		VectorType &	x,
		const DataType &	beta,
		VectorType &	y
	)

The matvec function y = G*x = C\S*x-x. Note that alpha and beta are untouched.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ SetBIluDropTolB()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluDropTolB ( T option )

inline

Set the recursive GeMSLR B part threshold for ILUT of the global B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBIluDropTolC()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluDropTolC ( T option )

inline

Set the recursive GeMSLR last level threshold for ILUT of the global B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBIluFillLevelB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluFillLevelB ( int option )

inline

Set the fill level for ILUK of the B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBIluFillLevelC()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluFillLevelC ( int option )

inline

Set the fill level for ILUK of the last level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBIluMaxRowNnzB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluMaxRowNnzB ( int option )

inline

Set the recursive GeMSLR B part maxinum number of nonzeros ILUT.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBIluMaxRowNnzC()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBIluMaxRowNnzC ( int option )

inline

Set the recursive GeMSLR last level maxinum number of nonzeros ILUT.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankArnoldiFactorOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankArnoldiFactorOtherLevels ( T option )

inline

Set the B part Arnoldi factor on other levels. m steps for arnoldi is rank * rank_factor * arnoldi_factor

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankArnoldiFactorTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankArnoldiFactorTopLevel ( T option )

inline

Set the B part Arnoldi factor on the top level. m steps for arnoldi is rank * rank_factor * arnoldi_factor

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankFactorOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankFactorOtherLevels ( T option )

inline

Set the B part low-rank factor on the other levels. The actuall computed number of low-rank terms is rank * factor >= rank.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankFactorTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankFactorTopLevel ( T option )

inline

Set the B part low-rank factor on the top level. The actuall computed number of low-rank terms is rank * factor >= rank.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankMaxNumberIterationsOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankMaxNumberIterationsOtherLevels ( int option )

inline

Set B part max restarts of thick-restart Arnoldi on other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankMaxNumberIterationsTopLevel()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankMaxNumberIterationsTopLevel ( int option )

inline

Set B part max restarts of thick-restart Arnoldi on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankOptionOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankOptionOtherLevels ( int option )

inline

Set the B part low-rank option on other levels. 0: Standard. 1: Thick-restart.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankOptionTopLevel()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankOptionTopLevel ( int option )

inline

Set the B part low-rank option on the top level. 0: Standard. 1: Thick-restart.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankRandomInitGuess()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankRandomInitGuess ( bool option )

inline

Set if we use random initial guess for Arnoldi in the B part. Otherwise we use 1 as initial guess.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankRanksOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankRanksOtherLevels ( T option )

inline

Set the B part target number of low-rank terms on the other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankRanksTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankRanksTopLevel ( T option )

inline

Set the B part target number of low-rank terms on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankThresholdOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankThresholdOtherLevels ( T option )

inline

Set B part max restarts of thick-restart Arnoldi on other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBLowRankThresholdTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBLowRankThresholdTopLevel ( T option )

inline

Set B part max restarts of thick-restart Arnoldi on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBMinimalNumberSubdomains()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBMinimalNumberSubdomains ( int option )

inline

Set the B part minimal number of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBNumberSubdomainsReduceFactor()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBNumberSubdomainsReduceFactor ( int option )

inline

Set the B part reduce factor of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBNumLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBNumLevels ( int option )

inline

Set the B part target number of levels of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBNumSubdomains()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBNumSubdomains ( int option )

inline

Set the B part target number of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBPartitionOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBPartitionOption ( int option )

inline

Set the B part partition option. 0: ND; 1: RKway.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBPermutationOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBPermutationOption ( int option )

inline

Set the B part permutation option. 0: No; 1: RCM; 2: AMD.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBPolyOrder()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBPolyOrder ( int option )

inline

Set poly order for Poly solve of the B part of the recursive GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetBSeperatorOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetBSeperatorOption ( bool option )

inline

Set the B part separator option of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetGlobalPartitionOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetGlobalPartitionOption ( bool option )

inline

Set the global partition option of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetGlobalPrecondOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetGlobalPrecondOption ( int option )

inline

Set the global precond option. 0: BJ; 1: ISCHUR; 2: ESCHUR; 3: MLEV.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluComplexShift()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluComplexShift ( bool complex_shift )

inline

Set if we turn on the complex shift or not (complex version only).

Parameters

[in] complex_shift The new drop tol for ILUT.

Returns: Return error message.

◆ SetIluDropTolB()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluDropTolB ( T option )

inline

Set the threshold for ILUT of the B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluDropTolC()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluDropTolC ( T option )

inline

Set the threshold for ILUT of the last level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluDropTolEF()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluDropTolEF ( T option )

inline

Set the threshold for ILUT of the EF part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluDropTolS()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluDropTolS ( T option )

inline

Set the threshold for ILUT of the S part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluFillLevelB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluFillLevelB ( int option )

inline

Set the fill level for ILUK of the B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluFillLevelC()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluFillLevelC ( int option )

inline

Set the fill level for ILUK of the last level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluMaxRowNnzB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluMaxRowNnzB ( int option )

inline

Set the maxinum number of nonzeros for ILUT of the B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluMaxRowNnzC()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluMaxRowNnzC ( int option )

inline

Set the maxinum number of nonzeros ILUT of the last level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluMaxRowNnzS()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluMaxRowNnzS ( int option )

inline

Set the maxinum number of nonzeros ILUT of the S part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetIluResidualIters()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetIluResidualIters ( int residual_iters )

inline

Set the number of residual iterations of B solves in the setup phase. Set to <= 1 to turn off.

Parameters

[in] residual_iters The number of residual iterations of B solves.

Returns: Return error message.

◆ SetInnerIterationMaxNumberIterations()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetInnerIterationMaxNumberIterations ( int option )

inline

Set the max number of iterations of inner iteration of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetInnerIterationOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetInnerIterationOption ( bool option )

inline

Set the inner iteration option of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetInnerIterationThreshold()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetInnerIterationThreshold ( T option )

inline

Set the stop threshold of inner iteration of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLocalGemslr()

template<class MatrixType , class VectorType , typename DataType >

virtual int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLocalGemslr ( GemslrClass< CsrMatrixClass< DataType >, SequentialVectorClass< DataType >, DataType > & gemslr )

inlinevirtual

Setup with parameter array. This is the helper function to set the local gemslr for B solve.

Parameters

[in] params The parameter array.

Returns: Return error message.

◆ SetLowerGemslr()

template<class MatrixType , class VectorType , typename DataType >

virtual int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowerGemslr ( ParallelGemslrClass< MatrixType, VectorType, DataType > & gemslr )

inlinevirtual

Setup with parameter array. This is the helper function to set the lower level gemslr for explicite Schur solve.

Parameters

[in] params The parameter array.

Returns: Return error message.

◆ SetLowRankA()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankA ( T option )

inline

Set the target number of low-rank terms for A.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankArnoldiFactorA()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankArnoldiFactorA ( T option )

inline

Set the Arnoldi factor on A. m steps for arnoldi is rank * rank_factor * arnoldi_factor

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankArnoldiFactorOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankArnoldiFactorOtherLevels ( T option )

inline

Set the Arnoldi factor on other levels. m steps for arnoldi is rank * rank_factor * arnoldi_factor

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankArnoldiFactorTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankArnoldiFactorTopLevel ( T option )

inline

Set the Arnoldi factor on the top level. m steps for arnoldi is rank * rank_factor * arnoldi_factor

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankFactorA()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankFactorA ( T option )

inline

Set the low-rank factor on A. The actuall computed number of low-rank terms is rank * factor >= rank.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankFactorOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankFactorOtherLevels ( T option )

inline

Set the low-rank factor on the other levels. The actuall computed number of low-rank terms is rank * factor >= rank.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankFactorTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankFactorTopLevel ( T option )

inline

Set the low-rank factor on the top level. The actuall computed number of low-rank terms is rank * factor >= rank.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankMaxNumberIterationsA()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankMaxNumberIterationsA ( int option )

inline

Set max restarts of thick-restart Arnoldi on A.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankMaxNumberIterationsOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankMaxNumberIterationsOtherLevels ( int option )

inline

Set max restarts of thick-restart Arnoldi on other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankMaxNumberIterationsTopLevel()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankMaxNumberIterationsTopLevel ( int option )

inline

Set max restarts of thick-restart Arnoldi on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankOptionA()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankOptionA ( int option )

inline

Set the low-rank option on A. 0: Standard. 1: Thick-restart.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankOptionOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankOptionOtherLevels ( int option )

inline

Set the low-rank option on other levels. 0: Standard. 1: Thick-restart.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankOptionTopLevel()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankOptionTopLevel ( int option )

inline

Set the low-rank option on the top level. 0: Standard. 1: Thick-restart.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankRandomInitGuess()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankRandomInitGuess ( bool option )

inline

Set if we use random initial guess for Arnoldi. Otherwise we use 1 as initial guess.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankRanksOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankRanksOtherLevels ( T option )

inline

Set the target number of low-rank terms on the other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankRanksTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankRanksTopLevel ( T option )

inline

Set the target number of low-rank terms on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankThresholdA()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankThresholdA ( T option )

inline

Set max restarts of thick-restart Arnoldi on A.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankThresholdOtherLevels()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankThresholdOtherLevels ( T option )

inline

Set max restarts of thick-restart Arnoldi on other levels.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetLowRankThresholdTopLevel()

template<class MatrixType , class VectorType , typename DataType >

template<typename T >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetLowRankThresholdTopLevel ( T option )

inline

Set max restarts of thick-restart Arnoldi on the top level.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetMinimalNumberSubdomains()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetMinimalNumberSubdomains ( int option )

inline

Set the global minimal number of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetNumberSubdomainsReduceFactor()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetNumberSubdomainsReduceFactor ( int option )

inline

Set the global reduce factor of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetNumLevels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetNumLevels ( int option )

inline

Set the global target number of levels of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetNumSubdomains()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetNumSubdomains ( int option )

inline

Set the global target number of subdomains on each level of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPartitionOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPartitionOption ( int option )

inline

Set the global partition option. 0: ND; 1: RKway.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPermutationOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPermutationOption ( int option )

inline

Set the global permutation option. 0: No; 1: RCM; 2: AMD.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPolyOrder()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPolyOrder ( int option )

inline

Set poly order for Poly solve of the B part.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPreconditionerOption1()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPreconditionerOption1 ( int option )

inline

Set top level preconditioner.

Note: ILU, GeMSLR, Poly.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPreconditionerOption1Levels()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPreconditionerOption1Levels ( int option )

inline

Set top level preconditioner apply levels.

Note: ILU, GeMSLR, Poly.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPreconditionerOption2()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPreconditionerOption2 ( int option )

inline

Set mid level preconditioner.

Note: ILU, GeMSLR, Poly.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPreconditionerOptionB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPreconditionerOptionB ( int option )

inline

Set all top levels preconditioner.

Note: ILU, BJ-ILU. Not used in the sequential version.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetPreconditionerOptionC()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetPreconditionerOptionC ( int option )

inline

Set last level preconditioner.

Note: ILU, BJ-ILU. Not used in the sequential version.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetSeperatorOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetSeperatorOption ( bool option )

inline

Set the global separator option of GeMSLR.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetSmoothOptionB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetSmoothOptionB ( int option )

inline

Set all top levels smoother.

Parameters

[in] option The new option.

Returns: Return error message.

◆ SetSolveLocation()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::SetSolveLocation ( const int & location )

virtual

Set the data location that the preconditioner apply to.

Parameters

[in] location The target solver location.

Returns: Return error message.

◆ SetSolveOption()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetSolveOption ( int option )

inline

Set the solve option. 0: additive LU solve; 1: additive L solve; 2: multi-solve.

Parameters

[in] option The new option.

Returns: Return error message.

◆ Setup()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::Setup	(	VectorType &	x,
		VectorType &	rhs
	)

virtual

Setup the precondioner phase. Will be called by the solver if not called directly.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.

Returns: Return error message.

◆ SetupBSolve()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupBSolve	(	VectorType &	x,
		VectorType &	rhs
	)

Setup the solve of B matrices of the GeMSLR.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.

Returns: Return error message.

◆ SetupBSolveGemslr()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupBSolveGemslr	(	VectorType &	x,
		VectorType &	rhs,
		int	level
	)

Setup the solve of B matrices of the GeMSLR with GeMSLR.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.

Returns: Return error message.

◆ SetupBSolveILUK()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupBSolveILUK	(	VectorType &	x,
		VectorType &	rhs,
		int	level
	)

Setup the solve of B matrices of the GeMSLR with ILUK.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.

Returns: Return error message.

◆ SetupBSolveILUT()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupBSolveILUT	(	VectorType &	x,
		VectorType &	rhs,
		int	level
	)

Setup the solve of B matrices of the GeMSLR with ILUT.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.

Returns: Return error message.

◆ SetupLowRank()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupLowRank	(	VectorType &	x,
		VectorType &	rhs
	)

Setup the low-rank part of the GeMSLR.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.

Returns: Return error message.

◆ SetupLowRankBuildLowRank()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupLowRankBuildLowRank	(	VectorType &	x,
		VectorType &	rhs,
		DenseMatrixClass< DataType > &	V,
		DenseMatrixClass< DataType > &	H,
		int	m,
		int	rank,
		int	level,
		int	option
	)

Given the V and H from Arnoldi, compute the final low-rank correction of S^{-1} - C^{-1}.
For GeMSLR, H -> (I-H)^{-1} - I
For EsMSLR, X -> X^{-1} - I.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	V	Matrix holding the orthogonal basis genrated by Arnoldi iteration.
[in]	H	The Hessenberg matrix generated by Arnoldi iteration.
[in]	m	The number of finished steps in Arnoldi iteration.
[in]	rank	The number of eigenvectors to keep.
[in]	level	The current level.
[in]	option	The setup option. GeMSLR, ESMSLR, or PSLR.

Returns: return # of Schur vectors, if return value < 0 an error occurs

◆ SetupLowRankNoRestart()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupLowRankNoRestart	(	VectorType &	x,
		VectorType &	rhs,
		int	level,
		int	option
	)

Setup the low-rank part of the GeMSLR with arnodi no-restart.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.
[in]	option	The setup option. GeMSLR, ESMSLR, or PSLR.

Returns: Return error message.

◆ SetupLowRankSubspaceIteration()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupLowRankSubspaceIteration	(	VectorType &	x,
		VectorType &	rhs,
		int	level,
		int	option
	)

Setup the low-rank part of the GeMSLR with subspace iteration

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.
[in]	option	The setup option. GeMSLR, ESMSLR, or PSLR.

Returns: Return error message.

◆ SetupLowRankThickRestart()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupLowRankThickRestart	(	VectorType &	x,
		VectorType &	rhs,
		int	level,
		int	option
	)

Setup the low-rank part of the GeMSLR with arnodi thick-restart.

Parameters

[in]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The target level.
[in]	option	The setup option. GeMSLR, ESMSLR, or PSLR.

Returns: Return error message.

◆ SetupPartialILUT()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupPartialILUT	(	VectorType &	x,
		VectorType &	rhs
	)

Setup the local partial ILUT.

◆ SetupPermutation()

template<class MatrixType , class VectorType , typename DataType >

template int pargemslr::precond_gemslr_csr_par_complexd::SetupPermutation ( )

Setup the permutation of the GeMSLR.

Returns: Return error message.

◆ SetupPermutationBuildLevelStructure()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupPermutationBuildLevelStructure	(	MatrixType &	A,
		int	level_start,
		vector_int &	map_v,
		vector_int &	mapptr_v
	)

Setup the level structure of the GeMSLR.

Returns: Return error message.

◆ SetupPermutationND()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupPermutationND	(	MatrixType &	A,
		int	nlev_setup,
		int &	nlev_max,
		int &	nlev_used,
		vector_int &	map_v,
		vector_int &	mapptr_v
	)

Setup the Nested Dissection partition of the GeMSLR. This is a symmetric reordering algorithm work on A+A'.

Returns: Return error message.

◆ SetupPermutationRKway()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetupPermutationRKway	(	MatrixType &	A,
		int	nlev_setup,
		int &	nlev_max,
		int &	nlev_used,
		vector_int &	map_v,
		vector_int &	mapptr_v
	)

Setup the Recursive KWay partition of the GeMSLR. This is a symmetric reordering algorithm work on A+A'.
1) On each level, partition the current adjacency graph with kway partitioning.
2) Find the adjacency graph of the interface matrix.
3) Repeat until we reach the last level or the interface matrix is too small.

Returns: Return error message.

◆ SetWithParameterArray()

template<class MatrixType , class VectorType , typename DataType >

virtual int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SetWithParameterArray ( double * params )

inlinevirtual

Setup with parameter array.

Parameters

[in] params The parameter array.

Returns: Return error message.

◆ Solve()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::Solve	(	VectorType &	x,
		VectorType &	rhs
	)

virtual

Solve phase. Call this function after Setup. Solve with cusparse if unified memory/device memory is used.

Parameters

[in,out]	x	The initial guess.
[in]	rhs	The right-hand-side.

Returns: Return error message.

◆ SolveApplyLowRankLevel()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveApplyLowRankLevel	(	VectorType &	x,
		VectorType &	rhs,
		int	level
	)

Apply the low-rank update on a certain level.

Parameters

[in,out]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	level	The start level.

Returns: Return error message.

◆ SolveB()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveB	(	VectorType &	x,
		VectorType &	rhs,
		int	option,
		int	level
	)

Solve with B on a certain level.

Parameters

[in,out]	x	The initial guess.
[in]	rhs	The right-hand-side.
[in]	option	0: solve with B_solver; 1: solve with B_precond; 2: apply smoothing.
[in]	level	The start level.

Returns: Return error message.

◆ SolveLevelEsmslr()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelEsmslr	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelEsmslrMul()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelEsmslrMul	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelEsmslrU()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelEsmslrU	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelGemslr()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelGemslr	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelGemslrMul()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelGemslrMul	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelGemslrU()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelGemslrU	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

◆ SolveLevelPslr()

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::SolveLevelPslr	(	VectorType &	x_out,
		VectorType &	rhs_in,
		int	level,
		bool	doperm
	)

Solve starting from a certain level.

Parameters

[in,out]	x_in	The initial guess.
[in]	rhs_out	The right-hand-side.
[in]	level	The start level.
[in]	doperm	Do we apply permutation on this level? (Some time the system is already permuted).

Returns: Return error message.

Member Data Documentation

◆ _global_precond_option

template<class MatrixType , class VectorType , typename DataType >

int pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::_global_precond_option

The Schur complement option.

◆ _lev_A

template<class MatrixType , class VectorType , typename DataType >

ParallelGemslrLevelClass< MatrixType, VectorType, DataType> pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::_lev_A

The structure for the A level low-rank correction.

◆ _levs_v

template<class MatrixType , class VectorType , typename DataType >

std::vector< ParallelGemslrLevelClass< MatrixType, VectorType, DataType> > pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >::_levs_v

Vector holding the level struct for all levels starting from the second level. The size of this vector is equal to _nlev_used, and is the number of true levels.

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

SRC/preconditioners/parallel_gemslr.hpp
SRC/preconditioners/parallel_gemslr.cpp

Public Member Functions

Public Attributes

Detailed Description

template<class MatrixType, class VectorType, typename DataType> class pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >

Constructor & Destructor Documentation

◆ ParallelGemslrClass() [1/3]

◆ ~ParallelGemslrClass()

◆ ParallelGemslrClass() [2/3]

◆ ParallelGemslrClass() [3/3]

Member Function Documentation

◆ ACMatVec()

◆ BMatVec()

◆ CheckParameters()

◆ Clear()

◆ CMatVec()

◆ EBFCMatVec()

◆ GetNumNonzeros() [1/2]

◆ GetNumNonzeros() [2/2]

◆ GetNumRows()

◆ GetSize()

◆ GetSolvePhase()

◆ MoveData()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ PCLRMatVec()

◆ PlotPatternGnuPlot()

◆ RAPEBFCMatVec()

◆ RAPMatVec()

◆ SchurMatVec()

◆ SCMatVec()

◆ SetBIluDropTolB()

◆ SetBIluDropTolC()

◆ SetBIluFillLevelB()

◆ SetBIluFillLevelC()

◆ SetBIluMaxRowNnzB()

◆ SetBIluMaxRowNnzC()

◆ SetBLowRankArnoldiFactorOtherLevels()

◆ SetBLowRankArnoldiFactorTopLevel()

◆ SetBLowRankFactorOtherLevels()

◆ SetBLowRankFactorTopLevel()

◆ SetBLowRankMaxNumberIterationsOtherLevels()

◆ SetBLowRankMaxNumberIterationsTopLevel()

◆ SetBLowRankOptionOtherLevels()

◆ SetBLowRankOptionTopLevel()

◆ SetBLowRankRandomInitGuess()

◆ SetBLowRankRanksOtherLevels()

◆ SetBLowRankRanksTopLevel()

◆ SetBLowRankThresholdOtherLevels()

◆ SetBLowRankThresholdTopLevel()

◆ SetBMinimalNumberSubdomains()

◆ SetBNumberSubdomainsReduceFactor()

◆ SetBNumLevels()

◆ SetBNumSubdomains()

◆ SetBPartitionOption()

◆ SetBPermutationOption()

◆ SetBPolyOrder()

◆ SetBSeperatorOption()

◆ SetGlobalPartitionOption()

◆ SetGlobalPrecondOption()

◆ SetIluComplexShift()

◆ SetIluDropTolB()

◆ SetIluDropTolC()

◆ SetIluDropTolEF()

◆ SetIluDropTolS()

◆ SetIluFillLevelB()

◆ SetIluFillLevelC()

◆ SetIluMaxRowNnzB()

◆ SetIluMaxRowNnzC()

◆ SetIluMaxRowNnzS()

◆ SetIluResidualIters()

◆ SetInnerIterationMaxNumberIterations()

◆ SetInnerIterationOption()

◆ SetInnerIterationThreshold()

◆ SetLocalGemslr()

◆ SetLowerGemslr()

◆ SetLowRankA()

◆ SetLowRankArnoldiFactorA()

◆ SetLowRankArnoldiFactorOtherLevels()

◆ SetLowRankArnoldiFactorTopLevel()

◆ SetLowRankFactorA()

template<class MatrixType, class VectorType, typename DataType>
class pargemslr::ParallelGemslrClass< MatrixType, VectorType, DataType >