[[Imported from SVN: r3568]]

[[Imported from SVN: r3567]]

[[Imported from SVN: r3566]]

subsize vectors.

[[Imported from SVN: r3565]]

[[Imported from SVN: r3564]]

[[Imported from SVN: r3563]]

[[Imported from SVN: r3562]]

[[Imported from SVN: r3561]]

[[Imported from SVN: r3560]]

[[Imported from SVN: r3559]]

[[Imported from SVN: r3558]]

[[Imported from SVN: r3557]]

Now it can be used if the remote index lists where setup by hand.

[[Imported from SVN: r3556]]

[[Imported from SVN: r3555]]

[[Imported from SVN: r3554]]

[[Imported from SVN: r3553]]

of O(N^2).  This makes a noticeable speed difference when reading grids with more
than a few hundred elements.  And using the STL the code actually got
simpler :-)

This closes FlySpray-Task 5

[[Imported from SVN: r3552]]

allow the boundary extraction algorithm to store the segments in a set
instead of a vector.  That will improve the complexity of the extraction
algorithm from quadratic to nlogn.

[[Imported from SVN: r3551]]

[[Imported from SVN: r3550]]

[[Imported from SVN: r3549]]

from the sequential version.

It is now possible to use a purely sequential version of AMG by using
the MatrixAdapter as Operator and SequentialInformation instead of
ParallelInformation. In this case now IndexSets, RemoteIndices and
other stuff needed for communication are created and used. Thus there
is now memory overhead.

TODO: MPI is still needed to compile AMG. For the stable version it
should be possible to compile AMG even if MPI is not present.

[[Imported from SVN: r3548]]

[[Imported from SVN: r3547]]

not always possible to access a typedef inside "const Foo" class, even
if it is possible to access the same typedef inside the mutable "Foo"
class. A proper fix would introduce bigger changes in the interface,
especially in the template parameters. The quick fix are a bunch of
RemoveConsts to access the typedefs via mutable class.

[[Imported from SVN: r3546]]

[[Imported from SVN: r3545]]

[[Imported from SVN: r3544]]

Now the DiscreteFunctionType is determined when addData is called.
This enables visualisation of different DiscreteFunctionType (i.e.
different polynomial order).

[[Imported from SVN: r3543]]

[[Imported from SVN: r3542]]

[[Imported from SVN: r3541]]

different polynomial order).

Therefore changed func_real call to evalDof and evalCoord call which know
the right DiscreteFunctionsType inside.

[[Imported from SVN: r3540]]

[[Imported from SVN: r3539]]

[[Imported from SVN: r3538]]

[[Imported from SVN: r3537]]

[[Imported from SVN: r3536]]

[[Imported from SVN: r3535]]

(Well, slight changes had to be made. I discussed this with Robert
and we will soon disclose the details ...)

[[Imported from SVN: r3534]]

[[Imported from SVN: r3533]]

bug: simplexquadraturerules for ID tries to return rule for 2D :(  fixed and i think this may also fix the ID bug that oliver reported..??

[[Imported from SVN: r3532]]

[[Imported from SVN: r3531]]

  Dune::MatrixAdapter<BCRSMat,Vector,Vector> fop(mat);

  typedef Dune::Amg::MatrixHierarchy<BCRSMat,ParallelIndexSet> MHierarchy;
  typedef Dune::Amg::CoarsenCriterion<Dune::Amg::SymmetricCriterion<BCRSMat,Dune::Amg::FirstDiagonal> >
    Criterion;
  typedef Dune::SeqSSOR<BCRSMat,Vector,Vector> Smoother;
  typedef Dune::Amg::SmootherTraits<Smoother>::Arguments SmootherArgs;

  SmootherArgs smootherArgs;

  smootherArgs.iterations = 2;


  Criterion criterion(15,coarsenTarget);
  criterion.setMaxDistance(2);

  Dune::SeqScalarProduct<Vector> sp;
  typedef Dune::Amg::AMG<MHierarchy,Vector,Smoother> AMG;

  AMG amg(fop.getmat(), criterion, smootherArgs, 1, 1);

[[Imported from SVN: r3530]]

[[Imported from SVN: r3529]]