DUNE PDELab (2.8)

fastamg.hh
Go to the documentation of this file.
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_FASTAMG_HH
4#define DUNE_ISTL_FASTAMG_HH
5
6#include <memory>
12#include <dune/istl/solvers.hh>
14#include <dune/istl/superlu.hh>
15#include <dune/istl/umfpack.hh>
17#include <dune/istl/io.hh>
19
20#include "fastamgsmoother.hh"
21
30namespace Dune
31{
32 namespace Amg
33 {
56 template<class M, class X, class PI=SequentialInformation, class A=std::allocator<X> >
57 class FastAMG : public Preconditioner<X,X>
58 {
59 public:
61 typedef M Operator;
73
75 typedef X Domain;
77 typedef X Range;
80
88 FastAMG(OperatorHierarchy& matrices, CoarseSolver& coarseSolver,
89 const Parameters& parms,
90 bool symmetric=true);
91
103 template<class C>
104 FastAMG(const Operator& fineOperator, const C& criterion,
105 const Parameters& parms=Parameters(),
106 bool symmetric=true,
108
112 FastAMG(const FastAMG& amg);
113
115 void pre(Domain& x, Range& b);
116
118 void apply(Domain& v, const Range& d);
119
122 {
124 }
125
127 void post(Domain& x);
128
133 template<class A1>
134 void getCoarsestAggregateNumbers(std::vector<std::size_t,A1>& cont);
135
136 std::size_t levels();
137
138 std::size_t maxlevels();
139
149 {
150 matrices_->recalculateGalerkin(NegateSet<typename PI::OwnerSet>());
151 }
152
157 bool usesDirectCoarseLevelSolver() const;
158
159 private:
166 template<class C>
167 void createHierarchies(C& criterion,
168 const std::shared_ptr<const Operator>& matrixptr,
169 const PI& pinfo);
170
177 struct LevelContext
178 {
190 typename OperatorHierarchy::RedistributeInfoList::const_iterator redist;
194 typename OperatorHierarchy::AggregatesMapList::const_iterator aggregates;
210 std::size_t level;
211 };
212
214 void mgc(LevelContext& levelContext, Domain& x, const Range& b);
215
222 void presmooth(LevelContext& levelContext, Domain& x, const Range& b);
223
230 void postsmooth(LevelContext& levelContext, Domain& x, const Range& b);
231
238 void moveToFineLevel(LevelContext& levelContext, bool processedFineLevel,
239 Domain& fineX);
240
245 bool moveToCoarseLevel(LevelContext& levelContext);
246
251 void initIteratorsWithFineLevel(LevelContext& levelContext);
252
254 std::shared_ptr<OperatorHierarchy> matrices_;
256 std::shared_ptr<CoarseSolver> solver_;
258 std::shared_ptr<Hierarchy<Range,A>> rhs_;
260 std::shared_ptr<Hierarchy<Domain,A>> lhs_;
262 std::shared_ptr<Hierarchy<Domain,A>> residual_;
263
267 std::shared_ptr<ScalarProduct> scalarProduct_;
269 std::size_t gamma_;
271 std::size_t preSteps_;
273 std::size_t postSteps_;
274 std::size_t level;
275 bool buildHierarchy_;
276 bool symmetric;
277 bool coarsesolverconverged;
279 typedef std::shared_ptr<Smoother> SmootherPointer;
280 SmootherPointer coarseSmoother_;
282 std::size_t verbosity_;
283 };
284
285 template<class M, class X, class PI, class A>
287 : matrices_(amg.matrices_), solver_(amg.solver_),
288 rhs_(), lhs_(), residual_(), scalarProduct_(amg.scalarProduct_),
289 gamma_(amg.gamma_), preSteps_(amg.preSteps_), postSteps_(amg.postSteps_),
290 symmetric(amg.symmetric), coarsesolverconverged(amg.coarsesolverconverged),
291 coarseSmoother_(amg.coarseSmoother_), verbosity_(amg.verbosity_)
292 {}
293
294 template<class M, class X, class PI, class A>
296 const Parameters& parms, bool symmetric_)
297 : matrices_(stackobject_to_shared_ptr(matrices)), solver_(&coarseSolver),
298 rhs_(), lhs_(), residual_(), scalarProduct_(),
299 gamma_(parms.getGamma()), preSteps_(parms.getNoPreSmoothSteps()),
300 postSteps_(parms.getNoPostSmoothSteps()), buildHierarchy_(false),
301 symmetric(symmetric_), coarsesolverconverged(true),
302 coarseSmoother_(), verbosity_(parms.debugLevel())
303 {
304 if(preSteps_>1||postSteps_>1)
305 {
306 std::cerr<<"WARNING only one step of smoothing is supported!"<<std::endl;
307 preSteps_=postSteps_=0;
308 }
309 assert(matrices_->isBuilt());
310 static_assert(std::is_same<PI,SequentialInformation>::value,
311 "Currently only sequential runs are supported");
312 }
313 template<class M, class X, class PI, class A>
314 template<class C>
316 const C& criterion,
317 const Parameters& parms,
318 bool symmetric_,
319 const PI& pinfo)
320 : solver_(), rhs_(), lhs_(), residual_(), scalarProduct_(), gamma_(parms.getGamma()),
321 preSteps_(parms.getNoPreSmoothSteps()), postSteps_(parms.getNoPostSmoothSteps()),
322 buildHierarchy_(true),
323 symmetric(symmetric_), coarsesolverconverged(true),
324 coarseSmoother_(), verbosity_(criterion.debugLevel())
325 {
326 if(preSteps_>1||postSteps_>1)
327 {
328 std::cerr<<"WARNING only one step of smoothing is supported!"<<std::endl;
329 preSteps_=postSteps_=1;
330 }
331 static_assert(std::is_same<PI,SequentialInformation>::value,
332 "Currently only sequential runs are supported");
333 // TODO: reestablish compile time checks.
334 //static_assert(static_cast<int>(PI::category)==static_cast<int>(S::category),
335 // "Matrix and Solver must match in terms of category!");
336 auto matrixptr = stackobject_to_shared_ptr(matrix);
337 createHierarchies(criterion, matrixptr, pinfo);
338 }
339
340 template<class M, class X, class PI, class A>
341 template<class C>
342 void FastAMG<M,X,PI,A>::createHierarchies(C& criterion,
343 const std::shared_ptr<const Operator>& matrixptr,
344 const PI& pinfo)
345 {
346 Timer watch;
347 matrices_ = std::make_shared<OperatorHierarchy>(
348 std::const_pointer_cast<Operator>(matrixptr),
349 stackobject_to_shared_ptr(const_cast<PI&>(pinfo)));
350
351 matrices_->template build<NegateSet<typename PI::OwnerSet> >(criterion);
352
353 if(verbosity_>0 && matrices_->parallelInformation().finest()->communicator().rank()==0)
354 std::cout<<"Building Hierarchy of "<<matrices_->maxlevels()<<" levels took "<<watch.elapsed()<<" seconds."<<std::endl;
355
356 if(buildHierarchy_ && matrices_->levels()==matrices_->maxlevels()) {
357 // We have the carsest level. Create the coarse Solver
358 typedef typename SmootherTraits<Smoother>::Arguments SmootherArgs;
359 SmootherArgs sargs;
360 sargs.iterations = 1;
361
363 cargs.setArgs(sargs);
364 if(matrices_->redistributeInformation().back().isSetup()) {
365 // Solve on the redistributed partitioning
366 cargs.setMatrix(matrices_->matrices().coarsest().getRedistributed().getmat());
367 cargs.setComm(matrices_->parallelInformation().coarsest().getRedistributed());
368 }else{
369 cargs.setMatrix(matrices_->matrices().coarsest()->getmat());
370 cargs.setComm(*matrices_->parallelInformation().coarsest());
371 }
372
373 coarseSmoother_ = ConstructionTraits<Smoother>::construct(cargs);
374 scalarProduct_ = createScalarProduct<X>(cargs.getComm(),category());
375
376#if HAVE_SUPERLU|| HAVE_SUITESPARSE_UMFPACK
377#if HAVE_SUITESPARSE_UMFPACK
378#define DIRECTSOLVER UMFPack
379#else
380#define DIRECTSOLVER SuperLU
381#endif
382 // Use superlu if we are purely sequential or with only one processor on the coarsest level.
383 if(std::is_same<ParallelInformation,SequentialInformation>::value // sequential mode
384 || matrices_->parallelInformation().coarsest()->communicator().size()==1 //parallel mode and only one processor
385 || (matrices_->parallelInformation().coarsest().isRedistributed()
386 && matrices_->parallelInformation().coarsest().getRedistributed().communicator().size()==1
387 && matrices_->parallelInformation().coarsest().getRedistributed().communicator().size()>0)) { // redistribute and 1 proc
388 if(verbosity_>0 && matrices_->parallelInformation().coarsest()->communicator().rank()==0)
389 std::cout<<"Using superlu"<<std::endl;
390 if(matrices_->parallelInformation().coarsest().isRedistributed())
391 {
392 if(matrices_->matrices().coarsest().getRedistributed().getmat().N()>0)
393 // We are still participating on this level
394 solver_.reset(new DIRECTSOLVER<typename M::matrix_type>(matrices_->matrices().coarsest().getRedistributed().getmat(), false, false));
395 else
396 solver_.reset();
397 }else
398 solver_.reset(new DIRECTSOLVER<typename M::matrix_type>(matrices_->matrices().coarsest()->getmat(), false, false));
399 }else
400#undef DIRECTSOLVER
401#endif // HAVE_SUPERLU|| HAVE_SUITESPARSE_UMFPACK
402 {
403 if(matrices_->parallelInformation().coarsest().isRedistributed())
404 {
405 if(matrices_->matrices().coarsest().getRedistributed().getmat().N()>0)
406 // We are still participating on this level
407 solver_.reset(new BiCGSTABSolver<X>(const_cast<M&>(matrices_->matrices().coarsest().getRedistributed()),
408 *scalarProduct_,
409 *coarseSmoother_, 1E-2, 1000, 0));
410 else
411 solver_.reset();
412 }else
413 solver_.reset(new BiCGSTABSolver<X>(const_cast<M&>(*matrices_->matrices().coarsest()),
414 *scalarProduct_,
415 *coarseSmoother_, 1E-2, 1000, 0));
416 }
417 }
418
419 if(verbosity_>0 && matrices_->parallelInformation().finest()->communicator().rank()==0)
420 std::cout<<"Building Hierarchy of "<<matrices_->maxlevels()<<" levels took "<<watch.elapsed()<<" seconds."<<std::endl;
421 }
422
423
424 template<class M, class X, class PI, class A>
426 {
427 Timer watch, watch1;
428 // Detect Matrix rows where all offdiagonal entries are
429 // zero and set x such that A_dd*x_d=b_d
430 // Thus users can be more careless when setting up their linear
431 // systems.
432 typedef typename M::matrix_type Matrix;
433 typedef typename Matrix::ConstRowIterator RowIter;
434 typedef typename Matrix::ConstColIterator ColIter;
435 typedef typename Matrix::block_type Block;
436 Block zero;
437 zero=typename Matrix::field_type();
438
439 const Matrix& mat=matrices_->matrices().finest()->getmat();
440 for(RowIter row=mat.begin(); row!=mat.end(); ++row) {
441 bool isDirichlet = true;
442 bool hasDiagonal = false;
443 ColIter diag;
444 for(ColIter col=row->begin(); col!=row->end(); ++col) {
445 if(row.index()==col.index()) {
446 diag = col;
447 hasDiagonal = (*col != zero);
448 }else{
449 if(*col!=zero)
450 isDirichlet = false;
451 }
452 }
453 if(isDirichlet && hasDiagonal)
454 diag->solve(x[row.index()], b[row.index()]);
455 }
456 if (verbosity_>0)
457 std::cout<<" Preprocessing Dirichlet took "<<watch1.elapsed()<<std::endl;
458 watch1.reset();
459 // No smoother to make x consistent! Do it by hand
460 matrices_->parallelInformation().coarsest()->copyOwnerToAll(x,x);
461 rhs_ = std::make_shared<Hierarchy<Range,A>>(std::make_shared<Range>(b));
462 lhs_ = std::make_shared<Hierarchy<Domain,A>>(std::make_shared<Domain>(x));
463 residual_ = std::make_shared<Hierarchy<Domain,A>>(std::make_shared<Domain>(x));
464 matrices_->coarsenVector(*rhs_);
465 matrices_->coarsenVector(*lhs_);
466 matrices_->coarsenVector(*residual_);
467
468 // The preconditioner might change x and b. So we have to
469 // copy the changes to the original vectors.
470 x = *lhs_->finest();
471 b = *rhs_->finest();
472 }
473 template<class M, class X, class PI, class A>
474 std::size_t FastAMG<M,X,PI,A>::levels()
475 {
476 return matrices_->levels();
477 }
478 template<class M, class X, class PI, class A>
479 std::size_t FastAMG<M,X,PI,A>::maxlevels()
480 {
481 return matrices_->maxlevels();
482 }
483
485 template<class M, class X, class PI, class A>
487 {
488 LevelContext levelContext;
489 // Init all iterators for the current level
490 initIteratorsWithFineLevel(levelContext);
491
492 assert(v.two_norm()==0);
493
494 level=0;
495 if(matrices_->maxlevels()==1){
496 // The coarse solver might modify the d!
497 Range b(d);
498 mgc(levelContext, v, b);
499 }else
500 mgc(levelContext, v, d);
501 if(postSteps_==0||matrices_->maxlevels()==1)
502 levelContext.pinfo->copyOwnerToAll(v, v);
503 }
504
505 template<class M, class X, class PI, class A>
506 void FastAMG<M,X,PI,A>::initIteratorsWithFineLevel(LevelContext& levelContext)
507 {
508 levelContext.matrix = matrices_->matrices().finest();
509 levelContext.pinfo = matrices_->parallelInformation().finest();
510 levelContext.redist =
511 matrices_->redistributeInformation().begin();
512 levelContext.aggregates = matrices_->aggregatesMaps().begin();
513 levelContext.lhs = lhs_->finest();
514 levelContext.residual = residual_->finest();
515 levelContext.rhs = rhs_->finest();
516 levelContext.level=0;
517 }
518
519 template<class M, class X, class PI, class A>
520 bool FastAMG<M,X,PI,A>
521 ::moveToCoarseLevel(LevelContext& levelContext)
522 {
523 bool processNextLevel=true;
524
525 if(levelContext.redist->isSetup()) {
526 throw "bla";
527 levelContext.redist->redistribute(static_cast<const Range&>(*levelContext.residual),
528 levelContext.residual.getRedistributed());
529 processNextLevel = levelContext.residual.getRedistributed().size()>0;
530 if(processNextLevel) {
531 //restrict defect to coarse level right hand side.
532 ++levelContext.pinfo;
533 Transfer<typename OperatorHierarchy::AggregatesMap::AggregateDescriptor,Range,ParallelInformation>
534 ::restrictVector(*(*levelContext.aggregates), *levelContext.rhs,
535 static_cast<const Range&>(levelContext.residual.getRedistributed()),
536 *levelContext.pinfo);
537 }
538 }else{
539 //restrict defect to coarse level right hand side.
540 ++levelContext.rhs;
541 ++levelContext.pinfo;
542 Transfer<typename OperatorHierarchy::AggregatesMap::AggregateDescriptor,Range,ParallelInformation>
543 ::restrictVector(*(*levelContext.aggregates), *levelContext.rhs,
544 static_cast<const Range&>(*levelContext.residual), *levelContext.pinfo);
545 }
546
547 if(processNextLevel) {
548 // prepare coarse system
549 ++levelContext.residual;
550 ++levelContext.lhs;
551 ++levelContext.matrix;
552 ++levelContext.level;
553 ++levelContext.redist;
554
555 if(levelContext.matrix != matrices_->matrices().coarsest() || matrices_->levels()<matrices_->maxlevels()) {
556 // next level is not the globally coarsest one
557 ++levelContext.aggregates;
558 }
559 // prepare the lhs on the next level
560 *levelContext.lhs=0;
561 *levelContext.residual=0;
562 }
563 return processNextLevel;
564 }
565
566 template<class M, class X, class PI, class A>
567 void FastAMG<M,X,PI,A>
568 ::moveToFineLevel(LevelContext& levelContext, bool processNextLevel, Domain& x)
569 {
570 if(processNextLevel) {
571 if(levelContext.matrix != matrices_->matrices().coarsest() || matrices_->levels()<matrices_->maxlevels()) {
572 // previous level is not the globally coarsest one
573 --levelContext.aggregates;
574 }
575 --levelContext.redist;
576 --levelContext.level;
577 //prolongate and add the correction (update is in coarse left hand side)
578 --levelContext.matrix;
579 --levelContext.residual;
580
581 }
582
583 typename Hierarchy<Domain,A>::Iterator coarseLhs = levelContext.lhs--;
584 if(levelContext.redist->isSetup()) {
585
586 // Need to redistribute during prolongate
587 Transfer<typename OperatorHierarchy::AggregatesMap::AggregateDescriptor,Range,ParallelInformation>
588 ::prolongateVector(*(*levelContext.aggregates), *coarseLhs, x,
589 levelContext.lhs.getRedistributed(),
590 matrices_->getProlongationDampingFactor(),
591 *levelContext.pinfo, *levelContext.redist);
592 }else{
593 Transfer<typename OperatorHierarchy::AggregatesMap::AggregateDescriptor,Range,ParallelInformation>
594 ::prolongateVector(*(*levelContext.aggregates), *coarseLhs, x,
595 matrices_->getProlongationDampingFactor(), *levelContext.pinfo);
596
597 // printvector(std::cout, *lhs, "prolongated coarse grid correction", "lhs", 10, 10, 10);
598 }
599
600
601 if(processNextLevel) {
602 --levelContext.rhs;
603 }
604
605 }
606
607
608 template<class M, class X, class PI, class A>
609 void FastAMG<M,X,PI,A>
610 ::presmooth(LevelContext& levelContext, Domain& x, const Range& b)
611 {
612 constexpr auto bl = blockLevel<typename M::matrix_type>();
613 GaussSeidelPresmoothDefect<bl>::apply(levelContext.matrix->getmat(),
614 x,
615 *levelContext.residual,
616 b);
617 }
618
619 template<class M, class X, class PI, class A>
620 void FastAMG<M,X,PI,A>
621 ::postsmooth(LevelContext& levelContext, Domain& x, const Range& b)
622 {
623 constexpr auto bl = blockLevel<typename M::matrix_type>();
624 GaussSeidelPostsmoothDefect<bl>
625 ::apply(levelContext.matrix->getmat(), x, *levelContext.residual, b);
626 }
627
628
629 template<class M, class X, class PI, class A>
631 {
632 return IsDirectSolver< CoarseSolver>::value;
633 }
634
635 template<class M, class X, class PI, class A>
636 void FastAMG<M,X,PI,A>::mgc(LevelContext& levelContext, Domain& v, const Range& b){
637
638 if(levelContext.matrix == matrices_->matrices().coarsest() && levels()==maxlevels()) {
639 // Solve directly
641 res.converged=true; // If we do not compute this flag will not get updated
642 if(levelContext.redist->isSetup()) {
643 levelContext.redist->redistribute(b, levelContext.rhs.getRedistributed());
644 if(levelContext.rhs.getRedistributed().size()>0) {
645 // We are still participating in the computation
646 levelContext.pinfo.getRedistributed().copyOwnerToAll(levelContext.rhs.getRedistributed(),
647 levelContext.rhs.getRedistributed());
648 solver_->apply(levelContext.lhs.getRedistributed(), levelContext.rhs.getRedistributed(), res);
649 }
650 levelContext.redist->redistributeBackward(v, levelContext.lhs.getRedistributed());
651 levelContext.pinfo->copyOwnerToAll(v, v);
652 }else{
653 levelContext.pinfo->copyOwnerToAll(b, b);
654 solver_->apply(v, const_cast<Range&>(b), res);
655 }
656
657 // printvector(std::cout, *lhs, "coarse level update", "u", 10, 10, 10);
658 // printvector(std::cout, *rhs, "coarse level rhs", "rhs", 10, 10, 10);
659 if (!res.converged)
660 coarsesolverconverged = false;
661 }else{
662 // presmoothing
663 presmooth(levelContext, v, b);
664 // printvector(std::cout, *lhs, "update", "u", 10, 10, 10);
665 // printvector(std::cout, *residual, "post presmooth residual", "r", 10);
666#ifndef DUNE_AMG_NO_COARSEGRIDCORRECTION
667 bool processNextLevel = moveToCoarseLevel(levelContext);
668
669 if(processNextLevel) {
670 // next level
671 for(std::size_t i=0; i<gamma_; i++)
672 mgc(levelContext, *levelContext.lhs, *levelContext.rhs);
673 }
674
675 moveToFineLevel(levelContext, processNextLevel, v);
676#else
677 *lhs=0;
678#endif
679
680 if(levelContext.matrix == matrices_->matrices().finest()) {
681 coarsesolverconverged = matrices_->parallelInformation().finest()->communicator().prod(coarsesolverconverged);
682 if(!coarsesolverconverged)
683 DUNE_THROW(MathError, "Coarse solver did not converge");
684 }
685
686 postsmooth(levelContext, v, b);
687 }
688 }
689
690
692 template<class M, class X, class PI, class A>
693 void FastAMG<M,X,PI,A>::post([[maybe_unused]] Domain& x)
694 {
695 lhs_=nullptr;
696 rhs_=nullptr;
697 residual_=nullptr;
698 }
699
700 template<class M, class X, class PI, class A>
701 template<class A1>
702 void FastAMG<M,X,PI,A>::getCoarsestAggregateNumbers(std::vector<std::size_t,A1>& cont)
703 {
704 matrices_->getCoarsestAggregatesOnFinest(cont);
705 }
706
707 } // end namespace Amg
708} // end namespace Dune
709
710#endif
A fast (sequential) algebraic multigrid based on agglomeration that saves memory bandwidth.
Definition: fastamg.hh:58
LevelIterator< Hierarchy< ParallelInformation, Allocator >, ParallelInformation > Iterator
Type of the mutable iterator.
Definition: hierarchy.hh:214
LevelIterator< const Hierarchy< MatrixOperator, Allocator >, const MatrixOperator > ConstIterator
Type of the const iterator.
Definition: hierarchy.hh:217
The hierarchies build by the coarsening process.
Definition: matrixhierarchy.hh:59
All parameters for AMG.
Definition: parameters.hh:391
ConstIterator class for sequential access.
Definition: matrix.hh:402
A generic dynamic dense matrix.
Definition: matrix.hh:559
typename Imp::BlockTraits< T >::field_type field_type
Export the type representing the underlying field.
Definition: matrix.hh:563
row_type::const_iterator ConstColIterator
Const iterator for the entries of each row.
Definition: matrix.hh:587
T block_type
Export the type representing the components.
Definition: matrix.hh:566
The negation of a set. An item is contained in the set if and only if it is not contained in the nega...
Definition: enumset.hh:94
Base class for matrix free definition of preconditioners.
Definition: preconditioner.hh:30
Base class for scalar product and norm computation.
Definition: scalarproducts.hh:50
Sequential SSOR preconditioner.
Definition: preconditioners.hh:139
A simple stop watch.
Definition: timer.hh:41
void reset() noexcept
Reset timer while keeping the running/stopped state.
Definition: timer.hh:55
double elapsed() const noexcept
Get elapsed user-time from last reset until now/last stop in seconds.
Definition: timer.hh:75
A few common exception classes.
Traits for type conversions and type information.
Some generic functions for pretty printing vectors and matrices.
Implementations of the inverse operator interface.
#define DUNE_THROW(E, m)
Definition: exceptions.hh:216
OperatorHierarchy::ParallelMatrixHierarchy::ConstIterator matrix
The iterator over the matrices.
Definition: fastamg.hh:182
void getCoarsestAggregateNumbers(std::vector< std::size_t, A1 > &cont)
Get the aggregate number of each unknown on the coarsest level.
Definition: fastamg.hh:702
ParallelInformationHierarchy::Iterator pinfo
The iterator over the parallel information.
Definition: fastamg.hh:186
void recalculateHierarchy()
Recalculate the matrix hierarchy.
Definition: fastamg.hh:148
void post(Domain &x)
Clean up.
Definition: fastamg.hh:693
X Domain
The domain type.
Definition: fastamg.hh:75
Hierarchy< Domain, A >::Iterator residual
The iterator over the residuals.
Definition: fastamg.hh:202
virtual SolverCategory::Category category() const
Category of the preconditioner (see SolverCategory::Category)
Definition: fastamg.hh:121
MatrixHierarchy< M, ParallelInformation, A > OperatorHierarchy
The operator hierarchy type.
Definition: fastamg.hh:70
OperatorHierarchy::RedistributeInfoList::const_iterator redist
The iterator over the redistribution information.
Definition: fastamg.hh:190
X Range
The range type.
Definition: fastamg.hh:77
PI ParallelInformation
The type of the parallel information. Either OwnerOverlapCommunication or another type describing the...
Definition: fastamg.hh:68
M Operator
The matrix operator type.
Definition: fastamg.hh:61
InverseOperator< X, X > CoarseSolver
the type of the coarse solver.
Definition: fastamg.hh:79
bool usesDirectCoarseLevelSolver() const
Check whether the coarse solver used is a direct solver.
Definition: fastamg.hh:630
Hierarchy< Domain, A >::Iterator lhs
The iterator over the left hand side.
Definition: fastamg.hh:198
Hierarchy< Range, A >::Iterator rhs
The iterator over the right hand sided.
Definition: fastamg.hh:206
std::size_t level
The level index.
Definition: fastamg.hh:210
void apply(Domain &v, const Range &d)
Apply one step of the preconditioner to the system A(v)=d.
Definition: fastamg.hh:486
OperatorHierarchy::ParallelInformationHierarchy ParallelInformationHierarchy
The parallal data distribution hierarchy type.
Definition: fastamg.hh:72
void pre(Domain &x, Range &b)
Prepare the preconditioner.
Definition: fastamg.hh:425
FastAMG(OperatorHierarchy &matrices, CoarseSolver &coarseSolver, const Parameters &parms, bool symmetric=true)
Construct a new amg with a specific coarse solver.
Definition: fastamg.hh:295
OperatorHierarchy::AggregatesMapList::const_iterator aggregates
The iterator over the aggregates maps.
Definition: fastamg.hh:194
void presmooth(LevelContext &levelContext, size_t steps)
Apply pre smoothing on the current level.
Definition: smoother.hh:404
const void * Arguments
A type holding all the arguments needed to call the constructor.
Definition: construction.hh:42
static std::shared_ptr< T > construct(Arguments &args)
Construct an object with the specified arguments.
Definition: construction.hh:50
void postsmooth(LevelContext &levelContext, size_t steps)
Apply post smoothing on the current level.
Definition: smoother.hh:426
Provides a classes representing the hierarchies in AMG.
Dune namespace.
Definition: alignedallocator.hh:11
std::shared_ptr< T > stackobject_to_shared_ptr(T &t)
Create a shared_ptr for a stack-allocated object.
Definition: shared_ptr.hh:70
Define general preconditioner interface.
Define base class for scalar product and norm.
Classes for the generic construction and application of the smoothers.
Templates characterizing the type of a solver.
Traits class for getting the attribute class of a smoother.
Definition: smoother.hh:64
Statistics about the application of an inverse operator.
Definition: solver.hh:46
bool converged
True if convergence criterion has been met.
Definition: solver.hh:71
Category
Definition: solvercategory.hh:21
@ sequential
Category for sequential solvers.
Definition: solvercategory.hh:23
Classes for using SuperLU with ISTL matrices.
Prolongation and restriction for amg.
Classes for using UMFPack with ISTL matrices.
Creative Commons License   |  Legal Statements / Impressum  |  Hosted by TU Dresden  |  generated with Hugo v0.111.3 (Jul 27, 22:29, 2024)