Base class for all EigenProblem solves. This simply defines standard interfaces so that different solvers can be used easily. More...
#include <eigen_solver.h>
Inheritance diagram for oomph::EigenSolver:
Public Member Functions | |
| EigenSolver () | |
| Empty constructor. | |
| EigenSolver (const EigenSolver &) | |
| Empty copy constructor. | |
| virtual | ~EigenSolver () |
| Empty destructor. | |
| virtual void | solve_eigenproblem (Problem *const &problem_pt, const int &n_eval, Vector< std::complex< double > > &eigenvalue, Vector< DoubleVector > &eigenvector_real, Vector< DoubleVector > &eigenvector_imag, const bool &do_adjoint_problem=false) |
| Solve the real eigenproblem that is assembled by elements in a mesh in a Problem object. Note that the assembled matrices include the shift and are real. The eigenvalues and eigenvectors are, in general, complex, and the eigenvalues can be NaNs or Infs. This function is therefore merely provided as a convenience, to be used if the user is confident that NaNs don't arise (e.g. in Arnoldi based solvers where typically only a small number of (finite) eigenvalues are computed), or if the users is happy to deal with NaNs in the subsequent post-processing. Function is virtual so it can be overloaded for Arnoldi type solvers that compute the (finite) eigenvalues directly At least n_eval eigenvalues are computed. | |
| virtual void | solve_eigenproblem (Problem *const &problem_pt, const int &n_eval, Vector< std::complex< double > > &alpha, Vector< double > &beta, Vector< DoubleVector > &eigenvector_real, Vector< DoubleVector > &eigenvector_imag, const bool &do_adjoint_problem=false)=0 |
Solve the real eigenproblem that is assembled by elements in a mesh in a Problem object. Note that the assembled matrices include the shift and are real. The eigenvalues and eigenvectors are, in general, complex. Eigenvalues may be infinite and are therefore returned as  where  is complex while  is real. The actual eigenvalues may then be computed by doing the division, checking for zero betas to avoid NaNs. There's a convenience wrapper to this function that simply computes these eigenvalues regardless. That version may die in NaN checking is enabled (via the fenv.h header and the associated feenable function). At least n_eval eigenvalues are computed. | |
| void | set_shift (const double &shift_value) |
| Set the value of the (real) shift. | |
| const double & | get_shift () const |
| Return the value of the (real) shift (const version) | |
| Public Member Functions inherited from oomph::DistributableLinearAlgebraObject | |
| DistributableLinearAlgebraObject () | |
| Default constructor - create a distribution. | |
| DistributableLinearAlgebraObject (const DistributableLinearAlgebraObject &matrix)=delete | |
| Broken copy constructor. | |
| void | operator= (const DistributableLinearAlgebraObject &)=delete |
| Broken assignment operator. | |
| virtual | ~DistributableLinearAlgebraObject () |
| Destructor. | |
| LinearAlgebraDistribution * | distribution_pt () const |
| access to the LinearAlgebraDistribution | |
| unsigned | nrow () const |
| access function to the number of global rows. | |
| unsigned | nrow_local () const |
| access function for the num of local rows on this processor. | |
| unsigned | nrow_local (const unsigned &p) const |
| access function for the num of local rows on this processor. | |
| unsigned | first_row () const |
| access function for the first row on this processor | |
| unsigned | first_row (const unsigned &p) const |
| access function for the first row on this processor | |
| bool | distributed () const |
| distribution is serial or distributed | |
| bool | distribution_built () const |
| if the communicator_pt is null then the distribution is not setup then false is returned, otherwise return true | |
| void | build_distribution (const LinearAlgebraDistribution *const dist_pt) |
| setup the distribution of this distributable linear algebra object | |
| void | build_distribution (const LinearAlgebraDistribution &dist) |
| setup the distribution of this distributable linear algebra object | |
Protected Attributes | |
| double | Sigma_real |
| Double value that represents the real part of the shift in shifted eigensolvers. | |
Additional Inherited Members | |
| Protected Member Functions inherited from oomph::DistributableLinearAlgebraObject | |
| void | clear_distribution () |
| clear the distribution of this distributable linear algebra object | |
Detailed Description
Base class for all EigenProblem solves. This simply defines standard interfaces so that different solvers can be used easily.
Definition at line 60 of file eigen_solver.h.
Constructor & Destructor Documentation
◆ EigenSolver() [1/2]
|
inline |
Empty constructor.
Definition at line 69 of file eigen_solver.h.
◆ EigenSolver() [2/2]
|
inline |
Empty copy constructor.
Definition at line 72 of file eigen_solver.h.
◆ ~EigenSolver()
|
inlinevirtual |
Empty destructor.
Definition at line 75 of file eigen_solver.h.
Member Function Documentation
◆ get_shift()
Return the value of the (real) shift (const version)
Definition at line 144 of file eigen_solver.h.
References Sigma_real.
◆ set_shift()
Set the value of the (real) shift.
Definition at line 138 of file eigen_solver.h.
References Sigma_real.
◆ solve_eigenproblem() [1/2]
|
pure virtual |
Solve the real eigenproblem that is assembled by elements in a mesh in a Problem object. Note that the assembled matrices include the shift and are real. The eigenvalues and eigenvectors are, in general, complex. Eigenvalues may be infinite and are therefore returned as where is complex while is real. The actual eigenvalues may then be computed by doing the division, checking for zero betas to avoid NaNs. There's a convenience wrapper to this function that simply computes these eigenvalues regardless. That version may die in NaN checking is enabled (via the fenv.h header and the associated feenable function). At least n_eval eigenvalues are computed.
Implemented in oomph::ANASAZI, and oomph::LAPACK_QZ.
◆ solve_eigenproblem() [2/2]
|
inlinevirtual |
Solve the real eigenproblem that is assembled by elements in a mesh in a Problem object. Note that the assembled matrices include the shift and are real. The eigenvalues and eigenvectors are, in general, complex, and the eigenvalues can be NaNs or Infs. This function is therefore merely provided as a convenience, to be used if the user is confident that NaNs don't arise (e.g. in Arnoldi based solvers where typically only a small number of (finite) eigenvalues are computed), or if the users is happy to deal with NaNs in the subsequent post-processing. Function is virtual so it can be overloaded for Arnoldi type solvers that compute the (finite) eigenvalues directly At least n_eval eigenvalues are computed.
Reimplemented in oomph::ANASAZI.
Definition at line 89 of file eigen_solver.h.
References i, oomph::FiniteElement::size(), and solve_eigenproblem().
Referenced by oomph::Problem::solve_adjoint_eigenproblem(), oomph::Problem::solve_eigenproblem(), oomph::Problem::solve_eigenproblem(), and solve_eigenproblem().
Member Data Documentation
◆ Sigma_real
|
protected |
Double value that represents the real part of the shift in shifted eigensolvers.
Definition at line 65 of file eigen_solver.h.
Referenced by oomph::LAPACK_QZ::find_eigenvalues(), get_shift(), set_shift(), oomph::ANASAZI::solve_eigenproblem(), and oomph::LAPACK_QZ::solve_eigenproblem_helper().
The documentation for this class was generated from the following file: