Class grouping different direct solvers. More...
#include <SparseDirectSolver.hxx>
Inheritance diagram for Seldon::SparseDirectSolver< T >:
Public Member Functions | |
| SparseDirectSolver () | |
| Default constructor. | |
| int | GetM () const |
| returns the number of rows of the factorised matrix | |
| int | GetN () const |
| returns the number of rows of the factorised matrix | |
| int | GetTypeOrdering () const |
| returns the ordering algorithm to use | |
| void | SetPermutation (const IVect &) |
| sets directly the new ordering (by giving a permutation vector) | |
| void | SelectOrdering (int) |
| modifies the ordering algorithm to use | |
| void | HideMessages () |
| hiding all messages | |
| void | ShowMessages () |
| displaying basic messages | |
| void | ShowFullHistory () |
| displaying all the messages | |
| void | SetPivotThreshold (const double &) |
| sets the threshold used for pivoting | |
| void | SetNumberOfThreadPerNode (int m) |
| modifies the number of threads per node (for Pastix only) | |
| int | GetNumberOfThreadPerNode () const |
| modifies the number of threads per node (for Pastix only) | |
| void | SelectDirectSolver (int) |
| modifies the direct solver to use | |
| void | SetNonSymmetricIlut () |
| enforces the use of unsymmetric algorithm for ilut solver | |
| int | GetDirectSolver () |
| returns the direct solver to use | |
| void | RefineSolution () |
| void | DoNotRefineSolution () |
| void | SetCoefficientEstimationNeededMemory (double) |
| void | SetMaximumCoefficientEstimationNeededMemory (double) |
| void | SetIncreaseCoefficientEstimationNeededMemory (double) |
| double | GetThresholdMatrix () const |
| returns threshold used for ilut (if this solver is selected) | |
| ~SparseDirectSolver () | |
| Destructor. | |
| void | Clear () |
| clearing factorisation | |
| void | SetThresholdMatrix (const double &) |
| modifies threshold used for ilut | |
| template<class Prop , class Storage , class Allocator > | |
| void | Factorize (Matrix< T, Prop, Storage, Allocator > &A, bool keep_matrix=false) |
| factorisation of matrix A More... | |
| template<class Prop , class Storage , class Allocator > | |
| void | PerformAnalysis (Matrix< T, Prop, Storage, Allocator > &A) |
| Performs the analysis of the matrix before numerical factorization. | |
| template<class Prop , class Storage , class Allocator > | |
| void | PerformFactorization (Matrix< T, Prop, Storage, Allocator > &A) |
| Performs the numerical factorization. | |
| int | GetInfoFactorization (int &ierr) const |
| Returns error code of the direct solver. | |
| void | Solve (Vector< T > &x) |
| x_solution is overwritten by solution of A x = b More... | |
| void | Solve (const SeldonTranspose &TransA, Vector< T > &x, bool assemble=true) |
| x_solution is overwritten with solution of A x = b or A^T x = b | |
| void | Solve (Matrix< T, General, ColMajor > &x) |
| x_solution is overwritten by solution of A x = b More... | |
| void | Solve (const SeldonTranspose &, Matrix< T, General, ColMajor > &x) |
| x_solution is overwritten by solution of A x = b More... | |
Static Public Member Functions | |
| static bool | IsAvailableSolver (int type) |
| returns true if the solver type is available | |
Protected Member Functions | |
| bool | AffectOrdering () |
| template<class T0 , class Prop , class Storage , class Alloc > | |
| void | ComputeOrdering (Matrix< T0, Prop, Storage, Alloc > &A) |
| computation of the permutation vector in order to reduce fill-in | |
| void | InitSolver () |
| initializes the solver (internal method) | |
Protected Attributes | |
| int | type_ordering |
| ordering to use | |
| int | type_solver |
| solver to use | |
| int | nb_threads_per_node |
| number of threads (for Pastix) | |
| IVect | permut |
| ordering (if supplied by the user) | |
| int | n |
| size of factorized linear system | |
| VirtualSparseDirectSolver< T > * | solver |
| pointer to the used solver | |
| double | threshold_matrix |
| threshold for ilut solver | |
| double | pivot_threshold |
| bool | refine_solution |
| int | print_level |
| bool | enforce_unsym_ilut |
| use of non-symmetric ilut ? | |
Detailed Description
template<class T>
class Seldon::SparseDirectSolver< T >
Class grouping different direct solvers.
Definition at line 26 of file SparseDirectSolver.hxx.
Member Function Documentation
◆ Factorize()
template<class T >
template<class Prop , class Storage , class Allocator >
| SELDON_EXTERN template void Seldon::SparseDirectSolver< T >::Factorize | ( | Matrix< T, Prop, Storage, Allocator > & | A, |
| bool | keep_matrix = false |
||
| ) |
factorisation of matrix A
LU factorisation is stored in the current object. You can ask to clear the matrix given on input (to spare memory)
Definition at line 503 of file SparseDirectSolver.cxx.
◆ Solve() [1/3]
template<class T >
| void Seldon::SparseDirectSolver< T >::Solve | ( | const SeldonTranspose & | trans, |
| Matrix< T, General, ColMajor > & | x_sol | ||
| ) |
x_solution is overwritten by solution of A x = b
Multiple right hand sides We assume that Factorize has been called previously
Definition at line 827 of file SparseDirectSolver.cxx.
◆ Solve() [2/3]
template<class T >
| void Seldon::SparseDirectSolver< T >::Solve | ( | Matrix< T, General, ColMajor > & | x_solution | ) |
x_solution is overwritten by solution of A x = b
Multiple right hand sides We assume that Factorize has been called previously
Definition at line 814 of file SparseDirectSolver.cxx.
◆ Solve() [3/3]
template<class T >
| void Seldon::SparseDirectSolver< T >::Solve | ( | Vector< T > & | x_solution | ) |
x_solution is overwritten by solution of A x = b
We assume that Factorize has been called previously
Definition at line 783 of file SparseDirectSolver.cxx.
The documentation for this class was generated from the following files:
- computation/interfaces/direct/SparseDirectSolver.hxx
- computation/interfaces/direct/SparseDirectSolver.cxx
- computation/interfaces/direct/SparseDirectSolverInline.cxx
- lib/Compil/Seldon/SparseDirectSolver.cpp