solid_preconditioners.h
Go to the documentation of this file.
1 // LIC// ====================================================================
2 // LIC// This file forms part of oomph-lib, the object-oriented,
3 // LIC// multi-physics finite-element library, available
4 // LIC// at http://www.oomph-lib.org.
5 // LIC//
6 // LIC// Copyright (C) 2006-2024 Matthias Heil and Andrew Hazel
7 // LIC//
8 // LIC// This library is free software; you can redistribute it and/or
9 // LIC// modify it under the terms of the GNU Lesser General Public
10 // LIC// License as published by the Free Software Foundation; either
11 // LIC// version 2.1 of the License, or (at your option) any later version.
12 // LIC//
13 // LIC// This library is distributed in the hope that it will be useful,
14 // LIC// but WITHOUT ANY WARRANTY; without even the implied warranty of
15 // LIC// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 // LIC// Lesser General Public License for more details.
17 // LIC//
18 // LIC// You should have received a copy of the GNU Lesser General Public
19 // LIC// License along with this library; if not, write to the Free Software
20 // LIC// Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
21 // LIC// 02110-1301 USA.
22 // LIC//
23 // LIC// The authors may be contacted at oomph-lib@maths.man.ac.uk.
24 // LIC//
25 // LIC//====================================================================
26 #ifndef OOMPH_SOLID_PRECONDITIONERS_HEADER
27 #define OOMPH_SOLID_PRECONDITIONERS_HEADER
28 
29 
30 // Config header generated by autoconfig
31 #ifdef HAVE_CONFIG_H
32 #include <oomph-lib-config.h>
33 #endif
34 
35 
36 // oomphlib headers
37 #include "../generic/matrices.h"
38 #include "../generic/assembly_handler.h"
39 #include "../generic/problem.h"
40 #include "../generic/block_preconditioner.h"
41 #include "../generic/preconditioner.h"
42 #include "../generic/SuperLU_preconditioner.h"
43 #include "../generic/matrix_vector_product.h"
44 
45 
46 namespace oomph
47 {
48  //===========================================================================
49  /// The least-squares commutator (LSC; formerly BFBT)
50  /// preconditioner. It uses blocks corresponding to the displacement/position
51  /// and pressure unknowns, i.e. there are a total of 2x2 blocks,
52  /// and all displacement/position components are treated as a
53  /// single block of unknowns.
54  ///
55  /// Here are the details: An "ideal" preconditioner
56  /// would solve the saddle point system
57  /// \f[ \left( \begin{array}{cc} {\bf F} & {\bf G} \\ {\bf D} & {\bf 0} \end{array} \right) \left( \begin{array}{c} {\bf z}_u \\ {\bf z}_p \end{array} \right) = \left( \begin{array}{c} {\bf r}_u \\ {\bf r}_p \end{array} \right) \f]
58  /// where \f$ {\bf F}\f$, \f$ {\bf G} \f$, and \f$ {\bf D}\f$ are
59  /// the blocks that arise in the Jacobian of the pressure-based
60  /// equations of linear and nonlinear elasticity (with dofs in order
61  /// of displacement/position and pressure).
62  /// The use of this preconditioner would ensure the convergence
63  /// of any iterative linear solver in a single iteration but its
64  /// application is, of course, exactly as expensive as a direct solve.
65  /// The LSC/BFBT preconditioner replaces the exact Jacobian by
66  /// a block-triangular approximation
67  /// \f[ \left( \begin{array}{cc} {\bf F} & {\bf G} \\ {\bf 0} & -{\bf M}_s \end{array} \right) \left( \begin{array}{c} {\bf z}_u \\ {\bf z}_p \end{array} \right) = \left( \begin{array}{c} {\bf r}_u \\ {\bf r}_p \end{array} \right), \f]
68  /// where \f${\bf M}_s\f$ is an approximation to the pressure
69  /// Schur-complement \f$ {\bf S} = {\bf D} {\bf F}^{-1}{\bf G}. \f$
70  /// This system can be solved in two steps:
71  /// -# Solve the second row for \f$ {\bf z}_p\f$ via
72  /// \f[ {\bf z}_p = - {\bf M}_s^{-1} {\bf r}_p \f]
73  /// -# Given \f$ {\bf z}_p \f$ , solve the first row for \f$ {\bf z}_u\f$ via
74  /// \f[ {\bf z}_u = {\bf F}^{-1} \big( {\bf r}_u - {\bf G} {\bf z}_p \big) \f].
75  /// In the LSC/BFBT preconditioner, the action of the inverse pressure
76  /// Schur complement
77  /// \f[ {\bf z}_p = - {\bf M}_s^{-1} {\bf r}_p \f]
78  /// is approximated by
79  /// \f[ {\bf z}_p = - \big({\bf D} \widehat{\bf Q}^{-1}{\bf G} \big)^{-1} \big({\bf D} \widehat{\bf Q}^{-1}{\bf F} \widehat{\bf Q}^{-1}{\bf G}\big) \big({\bf D} \widehat{\bf Q}^{-1}{\bf G} \big)^{-1} {\bf r}_p, \f]
80  /// where \f$ \widehat{\bf Q} \f$ is the diagonal of the
81  /// displacement/position mass matrix. The evaluation of this expression
82  /// involves two linear solves involving the matrix
83  /// \f[ {\bf P} = \big({\bf D} \widehat{\bf Q}^{-1}{\bf G} \big) \f]
84  /// which has the character of a matrix arising from the discretisation
85  /// of a Poisson problem on the pressure space. We also have
86  /// to evaluate matrix-vector products with the matrix
87  /// \f[ {\bf E}={\bf D}\widehat{\bf Q}^{-1}{\bf F}\widehat{\bf Q}^{-1}{\bf G} \f]
88  /// Details of the theory can be found in "Finite Elements and
89  /// Fast Iterative Solvers with Applications in Incompressible Fluid
90  /// Dynamics" by Howard C. Elman, David J. Silvester, and Andrew J. Wathen,
91  /// published by Oxford University Press, 2006.
92  ///
93  /// In our implementation of the preconditioner, the linear systems
94  /// can either be solved "exactly", using SuperLU (in its incarnation
95  /// as an exact preconditioner; this is the default) or by any
96  /// other Preconditioner (inexact solver) specified via the access functions
97  /// \code
98  /// PressureBasedSolidLSCPreconditioner::set_f_preconditioner(...)
99  /// \endcode
100  /// or
101  /// \code
102  /// PressureBasedSolidLSCPreconditioner::set_p_preconditioner(...)
103  /// \endcode
104  //===========================================================================
106  : public BlockPreconditioner<CRDoubleMatrix>
107  {
108  public:
109  /// Constructor - sets defaults for control flags
112  {
113  // Flag to indicate that the preconditioner has been setup
114  // previously -- if setup() is called again, data can
115  // be wiped.
117 
118  // By default we use SuperLU for both p and f blocks
121 
122  // resize the mesh pt
123  // note: meaningless if subsidiary preconditioner
124  this->set_nmesh(1);
125  Solid_mesh_pt = 0;
126 
127  // Set default preconditioners (inexact solvers) -- they are
128  // members of this class!
131 
132  // Flag to determine if mass matrix diagonal Q^{-1}
133  // is used for scaling.
134  P_matrix_using_scaling = true;
135 
136  // set Doc_time to false
137  Doc_time = false;
138 
139  // null the off diagonal Block matrix pt
140  Bt_mat_vec_pt = 0;
141 
142  // null the F matrix vector product helper
143  F_mat_vec_pt = 0;
144 
145  // null the QBt matrix vector product pt
146  QBt_mat_vec_pt = 0;
147 
148  // null the E matrix vector product helper
149  E_mat_vec_pt = 0;
150 
151  // by default we do not form the E matrix (BQFQBt)
152  Form_BFBt_product = false;
153  }
154 
155  /// Destructor
157  {
158  clean_up_memory();
159  }
160 
161  /// Broken copy constructor
163  const PressureBasedSolidLSCPreconditioner&) = delete;
164 
165  /// Broken assignment operator
166  // Commented out broken assignment operator because this can lead to a
167  // conflict warning when used in the virtual inheritence hierarchy.
168  // Essentially the compiler doesn't realise that two separate
169  // implementations of the broken function are the same and so, quite
170  // rightly, it shouts.
171  /*void operator=(const PressureBasedSolidLSCPreconditioner&) = delete;*/
172 
173  /// Setup the preconditioner
174  void setup();
175 
176  /// Apply preconditioner to Vector r
178 
179  /// specify the mesh containing the mesh containing the
180  /// block-preconditionable solid elements. The dimension of the
181  /// problem must also be specified.
183  {
185  }
186 
187  /// Enable mass matrix diagonal scaling in the
188  /// Schur complement approximation
190  {
191  P_matrix_using_scaling = true;
192  }
193 
194  /// Enable mass matrix diagonal scaling in the
195  /// Schur complement approximation
197  {
198  P_matrix_using_scaling = false;
199  }
200 
201  /// Return whether the mass matrix is using diagonal
202  /// scaling or not
204  {
205  return P_matrix_using_scaling;
206  }
207 
208  /// Function to set a new pressure matrix preconditioner (inexact solver)
209  void set_p_preconditioner(Preconditioner* new_p_preconditioner_pt)
210  {
211  // If the default preconditioner has been used
212  // clean it up now...
214  {
215  delete P_preconditioner_pt;
216  }
217  P_preconditioner_pt = new_p_preconditioner_pt;
219  }
220 
221  /// Function to (re-)set pressure matrix preconditioner (inexact
222  /// solver) to SuperLU
224  {
226  {
229  }
230  }
231 
232  /// Function to set a new momentum matrix preconditioner (inexact solver)
233  void set_f_preconditioner(Preconditioner* new_f_preconditioner_pt)
234  {
235  // If the default preconditioner has been used
236  // clean it up now...
238  {
239  delete F_preconditioner_pt;
240  }
241  F_preconditioner_pt = new_f_preconditioner_pt;
243  }
244 
245  /// Function to (re-)set momentum matrix preconditioner (inexact
246  /// solver) to SuperLU
248  {
250  {
253  }
254  }
255 
256 
257  /// Enable documentation of time
259  {
260  Doc_time = true;
261  }
262 
263  /// Disable documentation of time
265  {
266  Doc_time = false;
267  }
268 
269  /// If this function is called then:
270  /// in setup(...) : BFBt is computed.
271  /// in preconditioner_solve(...) : a single matrix vector product with
272  /// BFBt is performed.
274  {
275  Form_BFBt_product = true;
276  }
277 
278  /// if this function is called then:
279  /// in setup(...) : the matrices B, F are assembled and stored
280  /// (the default behaviour) .
281  /// in preconditioner_solve(...) : a sequence of matrix vector products
282  /// with B, F, and Bt is performed.
283  /// (Note: in this discussion no scaling was considered but B and Bt
284  /// are replaced with BQ and QBt with scaling)
286  {
287  Form_BFBt_product = false;
288  }
289 
290  /// Helper function to delete preconditioner data.
291  void clean_up_memory();
292 
293  private:
294  // oomph-lib objects
295  // -----------------
296 
297  // Pointers to preconditioner (=inexact solver) objects
298  // -----------------------------------------------------
299  /// Pointer to the 'preconditioner' for the pressure matrix
301 
302  /// Pointer to the 'preconditioner' for the F matrix
304 
305  /// flag indicating whether the default F preconditioner is used
307 
308  /// flag indicating whether the default P preconditioner is used
310 
311  /// Control flag is true if the preconditioner has been setup
312  /// (used so we can wipe the data when the preconditioner is
313  /// called again)
315 
316  /// Control flag is true if mass matrix diagonal scaling
317  /// is used in the Schur complement approximation
319 
320  /// Helper function to assemble the diagonal of the
321  /// mass matrix from the elemental contributions defined in
322  /// PressureBasedSolidEquations<DIM>::get_mass_matrix_diagonal(...).
324 
325  /// Boolean indicating whether the momentum system preconditioner
326  /// is a block preconditioner
328 
329  /// Set Doc_time to true for outputting results of timings
330  bool Doc_time;
331 
332  /// MatrixVectorProduct operator for F if BFBt is not to be formed.
334 
335  /// MatrixVectorProduct operator for QBt if BFBt is not to be formed.
337 
338  /// MatrixVectorProduct operator for Bt;
340 
341  /// MatrixVectorProduct operator for E (BFBt) if BFBt is to be formed.
343 
344  /// indicates whether BFBt should be formed or the component matrices
345  /// should be retained.
346  /// If true then:
347  /// in setup(...) : BFBt is computed.
348  /// in preconditioner_solve(...) : a single matrix vector product with
349  /// BFBt is performed.
350  /// if false then:
351  /// in setup(...) : the matrices B, F are assembled and stored.
352  /// in preconditioner_solve(...) : a sequence of matrix vector products
353  /// with B, F, and Bt is performed.
354  /// (Note: in this discussion no scaling was considered but B and Bt
355  /// are replaced with BQ and QBt with scaling)
357 
358  /// the pointer to the mesh of block preconditionable solid
359  /// elements.
361  };
362 
363 
364  /// ////////////////////////////////////////////////////////////////////////////
365  /// ////////////////////////////////////////////////////////////////////////////
366  /// ////////////////////////////////////////////////////////////////////////////
367 
368 
369  //============================================================================
370  /// The exact solid preconditioner. This extracts 2x2 blocks
371  /// (corresponding to the displacement/position and pressure unknowns)
372  /// and uses these to
373  /// build a single preconditioner matrix for testing purposes.
374  /// Iterative solvers should converge in a single step if this is used.
375  /// If it doesn't something is wrong in the setup of the block matrices.
376  //=============================================================================
377  template<typename MATRIX>
379  : public BlockPreconditioner<MATRIX>
380  {
381  public:
382  /// Constructor - do nothing
384 
385 
386  /// Destructor - do nothing
388 
389 
390  /// Broken copy constructor
392  const PressureBasedSolidExactPreconditioner&) = delete;
393 
394 
395  /// Broken assignment operator
396  /*void operator=(const PressureBasedSolidExactPreconditioner&) = delete;*/
397 
398 
399  /// Setup the preconditioner
400  void setup();
401 
402  /// Apply preconditioner to r
404 
405  protected:
406  /// Preconditioner matrix
407  MATRIX P_matrix;
408  };
409 
410 } // namespace oomph
411 #endif
Block Preconditioner base class. The block structure of the overall problem is determined from the Me...
const Mesh * mesh_pt(const unsigned &i) const
Access to i-th mesh (of the various meshes that contain block preconditionable elements of the same n...
void set_nmesh(const unsigned &n)
Specify the number of meshes required by this block preconditioner. Note: elements in different meshe...
A class for compressed row matrices. This is a distributable object.
Definition: matrices.h:888
A vector in the mathematical sense, initially developed for linear algebra type applications....
Definition: double_vector.h:58
Matrix vector product helper class - primarily a wrapper to Trilinos's Epetra matrix vector product m...
A general mesh class.
Definition: mesh.h:67
Preconditioner base class. Gives an interface to call all other preconditioners through and stores th...
//////////////////////////////////////////////////////////////////////////// ////////////////////////...
PressureBasedSolidExactPreconditioner()
Constructor - do nothing.
~PressureBasedSolidExactPreconditioner()
Destructor - do nothing.
PressureBasedSolidExactPreconditioner(const PressureBasedSolidExactPreconditioner &)=delete
Broken copy constructor.
void setup()
Broken assignment operator.
void preconditioner_solve(const Vector< double > &r, Vector< double > &z)
Apply preconditioner to r.
The least-squares commutator (LSC; formerly BFBT) preconditioner. It uses blocks corresponding to the...
MatrixVectorProduct * QBt_mat_vec_pt
MatrixVectorProduct operator for QBt if BFBt is not to be formed.
void enable_doc_time()
Enable documentation of time.
bool Form_BFBt_product
indicates whether BFBt should be formed or the component matrices should be retained....
CRDoubleMatrix * assemble_mass_matrix_diagonal()
Helper function to assemble the diagonal of the mass matrix from the elemental contributions defined ...
void enable_form_BFBt_product()
If this function is called then: in setup(...) : BFBt is computed. in preconditioner_solve(....
PressureBasedSolidLSCPreconditioner()
Constructor - sets defaults for control flags.
bool Preconditioner_has_been_setup
Control flag is true if the preconditioner has been setup (used so we can wipe the data when the prec...
Preconditioner * F_preconditioner_pt
Pointer to the 'preconditioner' for the F matrix.
MatrixVectorProduct * E_mat_vec_pt
MatrixVectorProduct operator for E (BFBt) if BFBt is to be formed.
void disable_doc_time()
Disable documentation of time.
bool Doc_time
Set Doc_time to true for outputting results of timings.
PressureBasedSolidLSCPreconditioner(const PressureBasedSolidLSCPreconditioner &)=delete
Broken copy constructor.
void set_p_preconditioner(Preconditioner *new_p_preconditioner_pt)
Function to set a new pressure matrix preconditioner (inexact solver)
Mesh * Solid_mesh_pt
the pointer to the mesh of block preconditionable solid elements.
void disable_p_matrix_scaling()
Enable mass matrix diagonal scaling in the Schur complement approximation.
bool F_preconditioner_is_block_preconditioner
Boolean indicating whether the momentum system preconditioner is a block preconditioner.
void preconditioner_solve(const DoubleVector &r, DoubleVector &z)
Apply preconditioner to Vector r.
void setup()
Broken assignment operator.
bool is_p_matrix_using_scaling() const
Return whether the mass matrix is using diagonal scaling or not.
void set_f_preconditioner(Preconditioner *new_f_preconditioner_pt)
Function to set a new momentum matrix preconditioner (inexact solver)
MatrixVectorProduct * Bt_mat_vec_pt
MatrixVectorProduct operator for Bt;.
bool Using_default_p_preconditioner
flag indicating whether the default P preconditioner is used
MatrixVectorProduct * F_mat_vec_pt
MatrixVectorProduct operator for F if BFBt is not to be formed.
void set_solid_mesh(Mesh *mesh_pt)
specify the mesh containing the mesh containing the block-preconditionable solid elements....
void clean_up_memory()
Helper function to delete preconditioner data.
bool Using_default_f_preconditioner
flag indicating whether the default F preconditioner is used
void set_p_superlu_preconditioner()
Function to (re-)set pressure matrix preconditioner (inexact solver) to SuperLU.
void disable_form_BFBt_product()
if this function is called then: in setup(...) : the matrices B, F are assembled and stored (the defa...
void enable_p_matrix_scaling()
Enable mass matrix diagonal scaling in the Schur complement approximation.
Preconditioner * P_preconditioner_pt
Pointer to the 'preconditioner' for the pressure matrix.
bool P_matrix_using_scaling
Control flag is true if mass matrix diagonal scaling is used in the Schur complement approximation.
void set_f_superlu_preconditioner()
Function to (re-)set momentum matrix preconditioner (inexact solver) to SuperLU.
An interface to allow SuperLU to be used as an (exact) Preconditioner.
//////////////////////////////////////////////////////////////////// ////////////////////////////////...