General interface to inversion of large, sparse Matrices. More...
#include <supermatrix.hpp>
Inheritance diagram for super_Matrix:
Collaboration diagram for super_Matrix:
Public Types | |
| typedef STD_TR1::shared_ptr< super_Matrix_iterator_factory > | p_iterator_factory |
| typedef STD_TR1::shared_ptr< const super_Matrix_iterator_factory > | p_const_iterator_factory |
| typedef std::pair< int, int > | indexpair |
| typedef super_Matrix_strategy::status | status |
| typedef STD_TR1::shared_ptr< super_Matrix > | ref |
| typedef STD_TR1::shared_ptr< const super_Matrix > | const_ref |
 Public Types inherited from calculation_Matrix | |
| typedef STD_TR1::shared_ptr< calculation_Matrix > | ref |
| typedef STD_TR1::shared_ptr< const calculation_Matrix > | const_ref |
Public Member Functions | |
| super_Matrix (MPI_Comm comm, const int N) | |
| Construct with a size N of the problem (invert a NxN matrix) | |
| status | get_status () |
| Return the current phase. | |
| p_iterator_factory | provide_iterator () |
| Return an iterator factory over non-zero elements. | |
| p_const_iterator_factory | provide_const_iterator () const |
| void | set_param (const std::string &, double) |
| Store optional parameters. | |
| void | set_param (const std::string &, int) |
| bool | get_double_param (const std::string &, double &) const |
| bool | get_int_param (const std::string &, int &) const |
| void | obtain_Diagonal (complex *) |
| Return the diagonal elements of the inverse. | |
| void | Add_To (Matrix &m) const |
| Add superMatrix to another matrix. | |
| void | Output (std::ostream &os) const |
| Output the matrix. | |
| void | Bake_to_Multiply (bool use_mkl=true) |
| Finish SetUp phase, proceed to Multiplication. | |
| void | Bake_to_Distributed_Multiply () |
| Finish SetUp phase, proceed to Multiplication. | |
| void | Bake_to_Factorize (bool distributed=false) |
| Finish SetUp phase, proceed to Factorization. | |
| void | Factorize () |
| Factorize, will call Bake_to_Factorize() if needed. | |
| void | Reset () |
| Reset, return to SetUp phase. | |
| ref | deep_copy () const |
| Create a deep copy of all matrix elements. | |
| MPI_Comm | get_comm () const |
| Return MPI Communicator. | |
| Public Member Functions inherited from calculation_Matrix | |
| void | Apply (const complex *phi, complex *result) const |
| Apply a vector (stored in a pointer, for legacy code only!) to the matrix. | |
| void | Apply (const Vector &phi, Vector &result) const |
| Apply a vector to the matrix. | |
| void | Apply (complex *phi) const |
| Overload of Apply to directly overwrite the input vector. | |
| void | Apply (Vector &phi) const |
| Overload of Apply to directly overwrite the input vector. | |
| void | Scale (complex v) |
| Scale the entire matrix by a complex constant. | |
| Public Member Functions inherited from Matrix | |
| void | assert_size (row r, column c) |
| assert that the Matrix has a minimum number of rows and columns, if not try to resize | |
| void | assert_size (row r, column c) const |
| assert that the Matrix has a minimum number of rows and columns, if not throw an exception | |
| void | Insert (row r, column c, complex v) |
| Insert an element at a given position (overwritten if already there) | |
| void | Add (row r, column c, complex v) |
| Add an element at a given position. | |
| complex | Get (row r, column c) const |
| Obtain an element at a given position. | |
| row | nrows () const |
| Number of rows. | |
| size_t | num_of_rows () const |
| column | ncols () const |
| Number of Columns. | |
| size_t | num_of_cols () const |
| void | add_diagonal (complex e) |
| Add a number to the entire diagonal of the Matrix. | |
| void | insert_diagonal (complex e) |
| Insert a number to the entire diagonal of the Matrix. | |
| void | multiply_by_number (complex v) |
| multiply whole Matrix by a complex number | |
| void | print (std::ostream &) const |
| Output matrix to the stream. | |
| void | formatted_print (std::ostream &) const |
| bool | check_hermitian () const |
| bool | check_realsymmetric () const |
| Matrix & | operator+= (Matrix const &rhs) |
Detailed Description
General interface to inversion of large, sparse Matrices.
super_Matrix is a general Matrix object, that can interface with different sparse matrix-inversion routines. Currently SuperLU, intelMKL, PETSc, and MUMPS are implemented. MUMPS is used as default on the vsc, see also http://graal.ens-lyon.fr/MUMPS/ .
Super_Matrix works in three phases. A strategy pattern is used to switch between phases. The enum of super_Matrix_strategy gives information about the current status:
enum status {SetUp, Baked, Multiply, Factorized, Factorized_dist};
- SetUp: Matrix buildup During buildup, values may be set using standard Matrix operations. No computational routines are available during buildup.
- Baked: Matrix preprocessing Input complete, preprocess the data. Rescaling and Conditioning is done in this phase. Usually, you do not need to access this phase explicitly.
- Factorized_dist factorized in a distributed MPI environment, Factorize the preprocessed matrix. All threads need to call computational routines. Multiplying with v yield A^-1 . v
- Factorized Matrix factorized Factorize the preprocessed matrix. All computational routines are available now. Multiplying with v yield A^-1 . v
- Multiply Matrix baked for multiplication The preprocessed matrix will be used for efficient multiplication. All computational routines are available now. Multiplying with v yield A . v
The documentation for this class was generated from the following files:
- src/matrix_inverter/inc/supermatrix.hpp
- src/matrix_inverter/src/supermatrix.cpp
