Dune Core Modules (2.7.1)

poweriteration.hh
1// -*- tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*-
2// vi: set et ts=4 sw=2 sts=2:
3#ifndef DUNE_ISTL_EIGENVALUE_POWERITERATION_HH
4#define DUNE_ISTL_EIGENVALUE_POWERITERATION_HH
5
6#include <cstddef> // provides std::size_t
7#include <cmath> // provides std::sqrt, std::abs
8
9#include <type_traits> // provides std::is_same
10#include <iostream> // provides std::cout, std::endl
11#include <limits> // provides std::numeric_limits
12#include <ios> // provides std::left, std::ios::left
13#include <iomanip> // provides std::setw, std::resetiosflags
14#include <memory> // provides std::unique_ptr
15#include <string> // provides std::string
16
17#include <dune/common/exceptions.hh> // provides DUNE_THROW(...)
18
19#include <dune/istl/operators.hh> // provides Dune::LinearOperator
20#include <dune/istl/solvercategory.hh> // provides Dune::SolverCategory::sequential
21#include <dune/istl/solvertype.hh> // provides Dune::IsDirectSolver
22#include <dune/istl/operators.hh> // provides Dune::MatrixAdapter
23#include <dune/istl/istlexception.hh> // provides Dune::ISTLError
24#include <dune/istl/io.hh> // provides Dune::printvector(...)
25#include <dune/istl/solvers.hh> // provides Dune::InverseOperatorResult
26
27namespace Dune
28{
29
34 namespace Impl {
42 template <class X, class Y = X>
43 class ScalingLinearOperator : public Dune::LinearOperator<X,Y>
44 {
45 public:
46 typedef X domain_type;
47 typedef Y range_type;
48 typedef typename X::field_type field_type;
49
50 ScalingLinearOperator (field_type immutable_scaling,
51 const field_type& mutable_scaling)
52 : immutable_scaling_(immutable_scaling),
53 mutable_scaling_(mutable_scaling)
54 {}
55
56 virtual void apply (const X& x, Y& y) const
57 {
58 y = x;
59 y *= immutable_scaling_*mutable_scaling_;
60 }
61
62 virtual void applyscaleadd (field_type alpha, const X& x, Y& y) const
63 {
64 X temp(x);
65 temp *= immutable_scaling_*mutable_scaling_;
66 y.axpy(alpha,temp);
67 }
68
71 {
73 }
74
75 protected:
76 const field_type immutable_scaling_;
77 const field_type& mutable_scaling_;
78 };
79
80
89 template <class OP1, class OP2>
90 class LinearOperatorSum
91 : public Dune::LinearOperator<typename OP1::domain_type,
92 typename OP1::range_type>
93 {
94 public:
95 typedef typename OP1::domain_type domain_type;
96 typedef typename OP1::range_type range_type;
97 typedef typename domain_type::field_type field_type;
98
99 LinearOperatorSum (const OP1& op1, const OP2& op2)
100 : op1_(op1), op2_(op2)
101 {
102 static_assert(std::is_same<typename OP2::domain_type,domain_type>::value,
103 "Domain type of both operators doesn't match!");
104 static_assert(std::is_same<typename OP2::range_type,range_type>::value,
105 "Range type of both operators doesn't match!");
106 }
107
108 virtual void apply (const domain_type& x, range_type& y) const
109 {
110 op1_.apply(x,y);
111 op2_.applyscaleadd(1.0,x,y);
112 }
113
114 virtual void applyscaleadd (field_type alpha,
115 const domain_type& x, range_type& y) const
116 {
117 range_type temp(y);
118 op1_.apply(x,temp);
119 op2_.applyscaleadd(1.0,x,temp);
120 y.axpy(alpha,temp);
121 }
122
124 virtual SolverCategory::Category category() const
125 {
127 }
128
129 protected:
130 const OP1& op1_;
131 const OP2& op2_;
132 };
133 } // end namespace Impl
134
171 template <typename BCRSMatrix, typename BlockVector>
173 {
174 protected:
175 // Type definitions for type of iteration operator (m_ - mu_*I)
178 typedef Impl::ScalingLinearOperator<BlockVector> ScalingOperator;
179 typedef Impl::LinearOperatorSum<MatrixOperator,ScalingOperator> OperatorSum;
180
181 public:
184
186 typedef OperatorSum IterationOperator;
187
188 public:
204 const unsigned int nIterationsMax = 1000,
205 const unsigned int verbosity_level = 0)
206 : m_(m), nIterationsMax_(nIterationsMax),
207 verbosity_level_(verbosity_level),
208 mu_(0.0),
209 matrixOperator_(m_),
210 scalingOperator_(-1.0,mu_),
211 itOperator_(matrixOperator_,scalingOperator_),
212 nIterations_(0),
213 title_(" PowerIteration_Algorithms: "),
214 blank_(title_.length(),' ')
215 {
216 // assert that BCRSMatrix type has blocklevel 2
217 static_assert
219 "Only BCRSMatrices with blocklevel 2 are supported.");
220
221 // assert that BCRSMatrix type has square blocks
222 static_assert
223 (BCRSMatrix::block_type::rows == BCRSMatrix::block_type::cols,
224 "Only BCRSMatrices with square blocks are supported.");
225
226 // assert that m_ is square
227 const int nrows = m_.M() * BCRSMatrix::block_type::rows;
228 const int ncols = m_.N() * BCRSMatrix::block_type::cols;
229 if (nrows != ncols)
230 DUNE_THROW(Dune::ISTLError,"Matrix is not square ("
231 << nrows << "x" << ncols << ").");
232 }
233
238
244
257 inline void applyPowerIteration (const Real& epsilon,
258 BlockVector& x, Real& lambda) const
259 {
260 // print verbosity information
261 if (verbosity_level_ > 0)
262 std::cout << title_
263 << "Performing power iteration approximating "
264 << "the dominant eigenvalue." << std::endl;
265
266 // allocate memory for auxiliary variables
267 BlockVector y(x);
268 BlockVector temp(x);
269
270 // perform power iteration
271 x *= (1.0 / x.two_norm());
272 m_.mv(x,y);
274 nIterations_ = 0;
275 while (r_norm > epsilon)
276 {
277 // update and check number of iterations
278 if (++nIterations_ > nIterationsMax_)
279 DUNE_THROW(Dune::ISTLError,"Power iteration did not converge "
280 << "in " << nIterationsMax_ << " iterations "
281 << "(║residual║_2 = " << r_norm << ", epsilon = "
282 << epsilon << ").");
283
284 // do one iteration of the power iteration algorithm
285 // (use that y = m_ * x)
286 x = y;
287 x *= (1.0 / y.two_norm());
288
289 // get approximated eigenvalue lambda via the Rayleigh quotient
290 m_.mv(x,y);
291 lambda = x * y;
292
293 // get norm of residual (use that y = m_ * x)
294 temp = y;
295 temp.axpy(-lambda,x);
296 r_norm = temp.two_norm();
297
298 // print verbosity information
299 if (verbosity_level_ > 1)
300 std::cout << blank_ << std::left
301 << "iteration " << std::setw(3) << nIterations_
302 << " (║residual║_2 = " << std::setw(11) << r_norm
303 << "): λ = " << lambda << std::endl
304 << std::resetiosflags(std::ios::left);
305 }
306
307 // print verbosity information
308 if (verbosity_level_ > 0)
309 {
310 std::cout << blank_ << "Result ("
311 << "#iterations = " << nIterations_ << ", "
312 << "║residual║_2 = " << r_norm << "): "
313 << "λ = " << lambda << std::endl;
314 if (verbosity_level_ > 2)
315 {
316 // print approximated eigenvector via DUNE-ISTL I/O methods
317 Dune::printvector(std::cout,x,blank_+"x",blank_+"row");
318 }
319 }
320 }
321
350 template <typename ISTLLinearSolver,
351 bool avoidLinSolverCrime = false>
352 inline void applyInverseIteration (const Real& epsilon,
353 ISTLLinearSolver& solver,
354 BlockVector& x, Real& lambda) const
355 {
356 constexpr Real gamma = 0.0;
357 applyInverseIteration(gamma,epsilon,solver,x,lambda);
358 }
359
389 template <typename ISTLLinearSolver,
390 bool avoidLinSolverCrime = false>
391 inline void applyInverseIteration (const Real& gamma,
392 const Real& epsilon,
393 ISTLLinearSolver& solver,
394 BlockVector& x, Real& lambda) const
395 {
396 // print verbosity information
397 if (verbosity_level_ > 0)
398 {
399 std::cout << title_;
400 if (gamma == 0.0)
401 std::cout << "Performing inverse iteration approximating "
402 << "the least dominant eigenvalue." << std::endl;
403 else
404 std::cout << "Performing inverse iteration with shift "
405 << "gamma = " << gamma << " approximating the "
406 << "eigenvalue closest to gamma." << std::endl;
407 }
408
409 // initialize iteration operator,
410 // initialize iteration matrix when needed
411 updateShiftMu(gamma,solver);
412
413 // allocate memory for linear solver statistics
414 Dune::InverseOperatorResult solver_statistics;
415
416 // allocate memory for auxiliary variables
417 BlockVector y(x);
418 Real y_norm;
419 BlockVector temp(x);
420
421 // perform inverse iteration with shift
422 x *= (1.0 / x.two_norm());
424 nIterations_ = 0;
425 while (r_norm > epsilon)
426 {
427 // update and check number of iterations
428 if (++nIterations_ > nIterationsMax_)
429 DUNE_THROW(Dune::ISTLError,"Inverse iteration "
430 << (gamma != 0.0 ? "with shift " : "") << "did not "
431 << "converge in " << nIterationsMax_ << " iterations "
432 << "(║residual║_2 = " << r_norm << ", epsilon = "
433 << epsilon << ").");
434
435 // do one iteration of the inverse iteration with shift algorithm,
436 // part 1: solve (m_ - gamma*I) * y = x for y
437 // (protect x from being changed)
438 temp = x;
439 solver.apply(y,temp,solver_statistics);
440
441 // get norm of y
442 y_norm = y.two_norm();
443
444 // compile time switch between accuracy and efficiency
445 if (avoidLinSolverCrime)
446 {
447 // get approximated eigenvalue lambda via the Rayleigh quotient
448 // (use that x_new = y / y_norm)
449 m_.mv(y,temp);
450 lambda = (y * temp) / (y_norm * y_norm);
451
452 // get norm of residual
453 // (use that x_new = y / y_norm, additionally use that temp = m_ * y)
454 temp.axpy(-lambda,y);
455 r_norm = temp.two_norm() / y_norm;
456 }
457 else
458 {
459 // get approximated eigenvalue lambda via the Rayleigh quotient
460 // (use that x_new = y / y_norm and use that (m_ - gamma*I) * y = x)
461 lambda = gamma + (y * x) / (y_norm * y_norm);
462
463 // get norm of residual
464 // (use that x_new = y / y_norm and use that (m_ - gamma*I) * y = x)
465 temp = x; temp.axpy(gamma-lambda,y);
466 r_norm = temp.two_norm() / y_norm;
467 }
468
469 // do one iteration of the inverse iteration with shift algorithm,
470 // part 2: update x
471 x = y;
472 x *= (1.0 / y_norm);
473
474 // print verbosity information
475 if (verbosity_level_ > 1)
476 std::cout << blank_ << std::left
477 << "iteration " << std::setw(3) << nIterations_
478 << " (║residual║_2 = " << std::setw(11) << r_norm
479 << "): λ = " << lambda << std::endl
480 << std::resetiosflags(std::ios::left);
481 }
482
483 // print verbosity information
484 if (verbosity_level_ > 0)
485 {
486 std::cout << blank_ << "Result ("
487 << "#iterations = " << nIterations_ << ", "
488 << "║residual║_2 = " << r_norm << "): "
489 << "λ = " << lambda << std::endl;
490 if (verbosity_level_ > 2)
491 {
492 // print approximated eigenvector via DUNE-ISTL I/O methods
493 Dune::printvector(std::cout,x,blank_+"x",blank_+"row");
494 }
495 }
496 }
497
528 template <typename ISTLLinearSolver,
529 bool avoidLinSolverCrime = false>
530 inline void applyRayleighQuotientIteration (const Real& epsilon,
531 ISTLLinearSolver& solver,
532 BlockVector& x, Real& lambda) const
533 {
534 // print verbosity information
535 if (verbosity_level_ > 0)
536 std::cout << title_
537 << "Performing Rayleigh quotient iteration for "
538 << "estimated eigenvalue " << lambda << "." << std::endl;
539
540 // allocate memory for linear solver statistics
541 Dune::InverseOperatorResult solver_statistics;
542
543 // allocate memory for auxiliary variables
544 BlockVector y(x);
545 Real y_norm;
546 Real lambda_update;
547 BlockVector temp(x);
548
549 // perform Rayleigh quotient iteration
550 x *= (1.0 / x.two_norm());
552 nIterations_ = 0;
553 while (r_norm > epsilon)
554 {
555 // update and check number of iterations
556 if (++nIterations_ > nIterationsMax_)
557 DUNE_THROW(Dune::ISTLError,"Rayleigh quotient iteration did not "
558 << "converge in " << nIterationsMax_ << " iterations "
559 << "(║residual║_2 = " << r_norm << ", epsilon = "
560 << epsilon << ").");
561
562 // update iteration operator,
563 // update iteration matrix when needed
564 updateShiftMu(lambda,solver);
565
566 // do one iteration of the Rayleigh quotient iteration algorithm,
567 // part 1: solve (m_ - lambda*I) * y = x for y
568 // (protect x from being changed)
569 temp = x;
570 solver.apply(y,temp,solver_statistics);
571
572 // get norm of y
573 y_norm = y.two_norm();
574
575 // compile time switch between accuracy and efficiency
576 if (avoidLinSolverCrime)
577 {
578 // get approximated eigenvalue lambda via the Rayleigh quotient
579 // (use that x_new = y / y_norm)
580 m_.mv(y,temp);
581 lambda = (y * temp) / (y_norm * y_norm);
582
583 // get norm of residual
584 // (use that x_new = y / y_norm, additionally use that temp = m_ * y)
585 temp.axpy(-lambda,y);
586 r_norm = temp.two_norm() / y_norm;
587 }
588 else
589 {
590 // get approximated eigenvalue lambda via the Rayleigh quotient
591 // (use that x_new = y / y_norm and use that (m_ - lambda_old*I) * y = x)
592 lambda_update = (y * x) / (y_norm * y_norm);
593 lambda += lambda_update;
594
595 // get norm of residual
596 // (use that x_new = y / y_norm and use that (m_ - lambda_old*I) * y = x)
597 temp = x; temp.axpy(-lambda_update,y);
598 r_norm = temp.two_norm() / y_norm;
599 }
600
601 // do one iteration of the Rayleigh quotient iteration algorithm,
602 // part 2: update x
603 x = y;
604 x *= (1.0 / y_norm);
605
606 // print verbosity information
607 if (verbosity_level_ > 1)
608 std::cout << blank_ << std::left
609 << "iteration " << std::setw(3) << nIterations_
610 << " (║residual║_2 = " << std::setw(11) << r_norm
611 << "): λ = " << lambda << std::endl
612 << std::resetiosflags(std::ios::left);
613 }
614
615 // print verbosity information
616 if (verbosity_level_ > 0)
617 {
618 std::cout << blank_ << "Result ("
619 << "#iterations = " << nIterations_ << ", "
620 << "║residual║_2 = " << r_norm << "): "
621 << "λ = " << lambda << std::endl;
622 if (verbosity_level_ > 2)
623 {
624 // print approximated eigenvector via DUNE-ISTL I/O methods
625 Dune::printvector(std::cout,x,blank_+"x",blank_+"row");
626 }
627 }
628 }
629
686 template <typename ISTLLinearSolver,
687 bool avoidLinSolverCrime = false>
688 inline void applyTLIMEIteration (const Real& gamma, const Real& eta,
689 const Real& epsilon,
690 ISTLLinearSolver& solver,
691 const Real& delta, const std::size_t& m,
692 bool& extrnl,
693 BlockVector& x, Real& lambda) const
694 {
695 // use same variable names as in [Szyld, 1988]
696 BlockVector& x_s = x;
697 Real& mu_s = lambda;
698
699 // print verbosity information
700 if (verbosity_level_ > 0)
701 std::cout << title_
702 << "Performing TLIME iteration for "
703 << "estimated eigenvalue in the "
704 << "interval (" << gamma - eta << ","
705 << gamma + eta << ")." << std::endl;
706
707 // allocate memory for linear solver statistics
708 Dune::InverseOperatorResult solver_statistics;
709
710 // allocate memory for auxiliary variables
711 bool doRQI;
712 Real mu;
713 BlockVector y(x_s);
714 Real omega;
715 Real mu_s_old;
716 Real mu_s_update;
717 BlockVector temp(x_s);
718 Real q_norm, r_norm;
719
720 // perform TLIME iteration
721 x_s *= (1.0 / x_s.two_norm());
722 extrnl = true;
723 doRQI = false;
725 nIterations_ = 0;
726 while (r_norm > epsilon)
727 {
728 // update and check number of iterations
729 if (++nIterations_ > nIterationsMax_)
730 DUNE_THROW(Dune::ISTLError,"TLIME iteration did not "
731 << "converge in " << nIterationsMax_
732 << " iterations (║residual║_2 = " << r_norm
733 << ", epsilon = " << epsilon << ").");
734
735 // set shift for next iteration according to inverse iteration
736 // with shift (II) resp. Rayleigh quotient iteration (RQI)
737 if (doRQI)
738 mu = mu_s;
739 else
740 mu = gamma;
741
742 // update II/RQI iteration operator,
743 // update II/RQI iteration matrix when needed
744 updateShiftMu(mu,solver);
745
746 // do one iteration of the II/RQI algorithm,
747 // part 1: solve (m_ - mu*I) * y = x for y
748 temp = x_s;
749 solver.apply(y,temp,solver_statistics);
750
751 // do one iteration of the II/RQI algorithm,
752 // part 2: compute omega
753 omega = (1.0 / y.two_norm());
754
755 // backup the old Rayleigh quotient
756 mu_s_old = mu_s;
757
758 // compile time switch between accuracy and efficiency
759 if (avoidLinSolverCrime)
760 {
761 // update the Rayleigh quotient mu_s, i.e. the approximated eigenvalue
762 // (use that x_new = y * omega)
763 m_.mv(y,temp);
764 mu_s = (y * temp) * (omega * omega);
765
766 // get norm of "the residual with respect to the shift used by II",
767 // use normal representation of q
768 // (use that x_new = y * omega, use that temp = m_ * y)
769 temp.axpy(-gamma,y);
770 q_norm = temp.two_norm() * omega;
771
772 // get norm of "the residual with respect to the Rayleigh quotient"
773 r_norm = q_norm*q_norm - (gamma-mu_s)*(gamma-mu_s);
774 // prevent that truncation errors invalidate the norm
775 // (we don't want to calculate sqrt of a negative number)
776 if (r_norm >= 0)
777 {
778 // use relation between the norms of r and q for efficiency
779 r_norm = std::sqrt(r_norm);
780 }
781 else
782 {
783 // use relation between r and q
784 // (use that x_new = y * omega, use that temp = (m_ - gamma*I) * y = q / omega)
785 temp.axpy(gamma-mu_s,y);
786 r_norm = temp.two_norm() * omega;
787 }
788 }
789 else
790 {
791 // update the Rayleigh quotient mu_s, i.e. the approximated eigenvalue
792 if (!doRQI)
793 {
794 // (use that x_new = y * omega, additionally use that (m_ - gamma*I) * y = x_s)
795 mu_s = gamma + (y * x_s) * (omega * omega);
796 }
797 else
798 {
799 // (use that x_new = y * omega, additionally use that (m_ - mu_s_old*I) * y = x_s)
800 mu_s_update = (y * x_s) * (omega * omega);
801 mu_s += mu_s_update;
802 }
803
804 // get norm of "the residual with respect to the shift used by II"
805 if (!doRQI)
806 {
807 // use special representation of q in the II case
808 // (use that x_new = y * omega, additionally use that (m_ - gamma*I) * y = x_s)
809 q_norm = omega;
810 }
811 else
812 {
813 // use special representation of q in the RQI case
814 // (use that x_new = y * omega, additionally use that (m_ - mu_s_old*I) * y = x_s)
815 temp = x_s; temp.axpy(mu_s-gamma,y);
816 q_norm = temp.two_norm() * omega;
817 }
818
819 // get norm of "the residual with respect to the Rayleigh quotient"
820 // don't use efficient relation between the norms of r and q, as
821 // this relation seems to yield a less accurate r_norm in the case
822 // where linear solver crime is admitted
823 if (!doRQI)
824 {
825 // (use that x_new = y * omega and use that (m_ - gamma*I) * y = x_s)
826 temp = x_s; temp.axpy(gamma-lambda,y);
827 r_norm = temp.two_norm() * omega;
828 }
829 else
830 {
831 // (use that x_new = y * omega and use that (m_ - mu_s_old*I) * y = x_s)
832 temp = x_s; temp.axpy(-mu_s_update,y);
833 r_norm = temp.two_norm() * omega;
834 }
835 }
836
837 // do one iteration of the II/RQI algorithm,
838 // part 3: update x
839 x_s = y; x_s *= omega;
840
841 // // for relative residual norm mode, scale with mu_s^{-1}
842 // r_norm /= std::abs(mu_s);
843
844 // print verbosity information
845 if (verbosity_level_ > 1)
846 std::cout << blank_ << "iteration "
847 << std::left << std::setw(3) << nIterations_
848 << " (" << (doRQI ? "RQI," : "II, ")
849 << " " << (doRQI ? "—>" : " ") << " "
850 << "║r║_2 = " << std::setw(11) << r_norm
851 << ", " << (doRQI ? " " : "—>") << " "
852 << "║q║_2 = " << std::setw(11) << q_norm
853 << "): λ = " << lambda << std::endl
854 << std::resetiosflags(std::ios::left);
855
856 // check if the eigenvalue closest to gamma lies in J
857 if (!doRQI && q_norm < eta)
858 {
859 // J is not free of eigenvalues
860 extrnl = false;
861
862 // by theory we know now that mu_s also lies in J
863 assert(std::abs(mu_s-gamma) < eta);
864
865 // switch to RQI
866 doRQI = true;
867 }
868
869 // revert to II if J is not free of eigenvalues but
870 // at some point mu_s falls back again outside J
871 if (!extrnl && doRQI && std::abs(mu_s-gamma) >= eta)
872 doRQI = false;
873
874 // if eigenvalue closest to gamma does not lie in J use RQI
875 // solely to accelerate the convergence to this eigenvalue
876 // when II has become stationary
877 if (extrnl && !doRQI)
878 {
879 // switch to RQI if the relative change of the Rayleigh
880 // quotient indicates that II has become stationary
881 if (nIterations_ >= m &&
882 std::abs(mu_s - mu_s_old) / std::abs(mu_s) < delta)
883 doRQI = true;
884 }
885 }
886
887 // // compute final residual and lambda again (paranoia....)
888 // m_.mv(x_s,temp);
889 // mu_s = x_s * temp;
890 // temp.axpy(-mu_s,x_s);
891 // r_norm = temp.two_norm();
892 // // r_norm /= std::abs(mu_s);
893
894 // print verbosity information
895 if (verbosity_level_ > 0)
896 {
897 if (extrnl)
898 std::cout << blank_ << "Interval "
899 << "(" << gamma - eta << "," << gamma + eta
900 << ") is free of eigenvalues, approximating "
901 << "the closest eigenvalue." << std::endl;
902 std::cout << blank_ << "Result ("
903 << "#iterations = " << nIterations_ << ", "
904 << "║residual║_2 = " << r_norm << "): "
905 << "λ = " << lambda << std::endl;
906 if (verbosity_level_ > 2)
907 {
908 // print approximated eigenvector via DUNE-ISTL I/O methods
909 Dune::printvector(std::cout,x,blank_+"x",blank_+"row");
910 }
911 }
912 }
913
923 {
924 // return iteration operator
925 return itOperator_;
926 }
927
942 inline const BCRSMatrix& getIterationMatrix () const
943 {
944 // create iteration matrix on demand
945 if (!itMatrix_)
946 itMatrix_ = std::make_unique<BCRSMatrix>(m_);
947
948 // return iteration matrix
949 return *itMatrix_;
950 }
951
956 inline unsigned int getIterationCount () const
957 {
958 if (nIterations_ == 0)
959 DUNE_THROW(Dune::ISTLError,"No algorithm applied, yet.");
960
961 return nIterations_;
962 }
963
964 protected:
979 template <typename ISTLLinearSolver>
980 inline void updateShiftMu (const Real& mu,
981 ISTLLinearSolver& solver) const
982 {
983 // do nothing if new shift equals the old one
984 if (mu == mu_) return;
985
986 // update shift mu_, i.e. update iteration operator
987 mu_ = mu;
988
989 // update iteration matrix when needed
990 if (itMatrix_)
991 {
992 // iterate over entries in iteration matrix diagonal
993 constexpr int rowBlockSize = BCRSMatrix::block_type::rows;
994 constexpr int colBlockSize = BCRSMatrix::block_type::cols;
995 for (typename BCRSMatrix::size_type i = 0;
996 i < itMatrix_->M()*rowBlockSize; ++i)
997 {
998 // access m_[i,i] where i is the flat index of a row/column
999 const Real& m_entry = m_
1000 [i/rowBlockSize][i/colBlockSize][i%rowBlockSize][i%colBlockSize];
1001 // access *itMatrix[i,i] where i is the flat index of a row/column
1002 Real& entry = (*itMatrix_)
1003 [i/rowBlockSize][i/colBlockSize][i%rowBlockSize][i%colBlockSize];
1004 // change current entry in iteration matrix diagonal
1005 entry = m_entry - mu_;
1006 }
1007 // notify linear solver about change of the iteration matrix object
1009 (solver,*itMatrix_);
1010 }
1011 }
1012
1013 protected:
1014 // parameters related to iterative eigenvalue algorithms
1015 const BCRSMatrix& m_;
1016 const unsigned int nIterationsMax_;
1017
1018 // verbosity setting
1019 const unsigned int verbosity_level_;
1020
1021 // shift mu_ used by iteration operator/matrix (m_ - mu_*I)
1022 mutable Real mu_;
1023
1024 // iteration operator (m_ - mu_*I), passing shift mu_ by reference
1025 const MatrixOperator matrixOperator_;
1026 const ScalingOperator scalingOperator_;
1027 OperatorSum itOperator_;
1028
1029 // iteration matrix (m_ - mu_*I), provided on demand when needed
1030 // (e.g. for preconditioning)
1031 mutable std::unique_ptr<BCRSMatrix> itMatrix_;
1032
1033 // memory for storing temporary variables (mutable as they shall
1034 // just be effectless auxiliary variables of the const apply*(...)
1035 // methods)
1036 mutable unsigned int nIterations_;
1037
1038 // constants for printing verbosity information
1039 const std::string title_;
1040 const std::string blank_;
1041 };
1042
1045} // namespace Dune
1046
1047#endif // DUNE_ISTL_EIGENVALUE_POWERITERATION_HH
A sparse block matrix with compressed row storage.
Definition: bcrsmatrix.hh:426
static constexpr unsigned int blocklevel
increment block level counter
Definition: bcrsmatrix.hh:466
A::size_type size_type
The type for the index access and the size.
Definition: bcrsmatrix.hh:460
size_type M() const
number of columns (counted in blocks)
Definition: bcrsmatrix.hh:1937
void mv(const X &x, Y &y) const
y = A x
Definition: bcrsmatrix.hh:1571
size_type N() const
number of rows (counted in blocks)
Definition: bcrsmatrix.hh:1931
A vector of blocks with memory management.
Definition: bvector.hh:403
typename Imp::BlockTraits< B >::field_type field_type
export the type representing the field
Definition: bvector.hh:409
derive error class from the base class in common
Definition: istlexception.hh:16
A linear operator.
Definition: operators.hh:65
X::field_type field_type
The field type of the operator.
Definition: operators.hh:72
virtual void applyscaleadd(field_type alpha, const X &x, X &y) const=0
apply operator to x, scale and add:
virtual SolverCategory::Category category() const=0
Category of the linear operator (see SolverCategory::Category)
X range_type
The type of the range of the operator.
Definition: operators.hh:70
X domain_type
The type of the domain of the operator.
Definition: operators.hh:68
Adapter to turn a matrix into a linear operator.
Definition: operators.hh:135
Iterative eigenvalue algorithms based on power iteration.
Definition: poweriteration.hh:173
PowerIteration_Algorithms(const PowerIteration_Algorithms &)=delete
void applyInverseIteration(const Real &epsilon, ISTLLinearSolver &solver, BlockVector &x, Real &lambda) const
Perform the inverse iteration algorithm to compute an approximation lambda of the least dominant (i....
Definition: poweriteration.hh:352
void applyTLIMEIteration(const Real &gamma, const Real &eta, const Real &epsilon, ISTLLinearSolver &solver, const Real &delta, const std::size_t &m, bool &extrnl, BlockVector &x, Real &lambda) const
Perform the "two-level iterative method for eigenvalue calculations (TLIME)" iteration algorit...
Definition: poweriteration.hh:688
IterationOperator & getIterationOperator()
Return the iteration operator (m_ - mu_*I).
Definition: poweriteration.hh:922
PowerIteration_Algorithms(const BCRSMatrix &m, const unsigned int nIterationsMax=1000, const unsigned int verbosity_level=0)
Construct from required parameters.
Definition: poweriteration.hh:203
void applyPowerIteration(const Real &epsilon, BlockVector &x, Real &lambda) const
Perform the power iteration algorithm to compute an approximation lambda of the dominant (i....
Definition: poweriteration.hh:257
OperatorSum IterationOperator
Type of iteration operator (m_ - mu_*I)
Definition: poweriteration.hh:186
void applyRayleighQuotientIteration(const Real &epsilon, ISTLLinearSolver &solver, BlockVector &x, Real &lambda) const
Perform the Rayleigh quotient iteration algorithm to compute an approximation lambda of an eigenvalue...
Definition: poweriteration.hh:530
void applyInverseIteration(const Real &gamma, const Real &epsilon, ISTLLinearSolver &solver, BlockVector &x, Real &lambda) const
Perform the inverse iteration with shift algorithm to compute an approximation lambda of the eigenval...
Definition: poweriteration.hh:391
void updateShiftMu(const Real &mu, ISTLLinearSolver &solver) const
Update shift mu_, i.e. update iteration operator/matrix (m_ - mu_*I).
Definition: poweriteration.hh:980
PowerIteration_Algorithms & operator=(const PowerIteration_Algorithms &)=delete
unsigned int getIterationCount() const
Return the number of iterations in last application of an algorithm.
Definition: poweriteration.hh:956
const BCRSMatrix & getIterationMatrix() const
Return the iteration matrix (m_ - mu_*I), provided on demand when needed (e.g. for direct solvers or ...
Definition: poweriteration.hh:942
BlockVector::field_type Real
Type of underlying field.
Definition: poweriteration.hh:183
Helper class for notifying a DUNE-ISTL linear solver about a change of the iteration matrix object in...
Definition: solver.hh:520
A few common exception classes.
Some generic functions for pretty printing vectors and matrices.
decltype(auto) apply(F &&f, ArgTuple &&args)
Apply function with arguments given as tuple.
Definition: apply.hh:46
#define DUNE_THROW(E, m)
Definition: exceptions.hh:216
void printvector(std::ostream &s, const V &v, std::string title, std::string rowtext, int columns=1, int width=10, int precision=2)
Print an ISTL vector.
Definition: io.hh:86
auto max(ADLTag< 0 >, const V &v1, const V &v2)
implements binary Simd::max()
Definition: defaults.hh:79
Dune namespace.
Definition: alignedallocator.hh:14
Define general, extensible interface for operators. The available implementation wraps a matrix.
Implementations of the inverse operator interface.
Templates characterizing the type of a solver.
Statistics about the application of an inverse operator.
Definition: solver.hh:46
Category
Definition: solvercategory.hh:21
@ sequential
Category for sequential solvers.
Definition: solvercategory.hh:23
Creative Commons License   |  Legal Statements / Impressum  |  Hosted by TU Dresden  |  generated with Hugo v0.111.3 (Nov 12, 23:30, 2024)