[[Imported from SVN: r1474]]

parameter class. Now the constructors of AMG are less cumbersome and
error prone due to lesser parameters.

[[Imported from SVN: r1468]]

[[Imported from SVN: r1422]]

[[Imported from SVN: r1421]]

[[Imported from SVN: r1388]]

[[Imported from SVN: r1353]]

[[Imported from SVN: r1330]]

BTW it is nonesense to set update differently for a 1-level hierarchy.


[[Imported from SVN: r1326]]

- made kamg more configurable.

[[Imported from SVN: r1317]]

nonoverlapping grids.

[[Imported from SVN: r1314]]

- Added AMG with krylov cylcle. Up to now it seems to be more scalable
int terms of number of iterations needed for convergence. Albeit the
time to solution does not improve that much because of intensive
copying.

[[Imported from SVN: r1309]]

use same name for template parameters in the definition as in the declaration. This matters for xlc in template members.Ã

[[Imported from SVN: r1293]]

Really postsmooth.

[[Imported from SVN: r1228]]

This prepares for changes in superlu.m4 that make usage of SuperLU optional,
even if SuperLU was found by configure.

[[Imported from SVN: r1180]]

Distribute matrixmarket.hh
AMG: make x consistent even if there is only one level
galerkin.hh: Initialize Matrixblock correctly
splvers.hh: Always call pre mthod of the preconditioner _before_ using
the operator. Thus x is consistent and the computed defect will be,
too.
matrixmarket.hh: use int instead of size_t as template parameter as
this is the case for the FieldMatrix, too.

[[Imported from SVN: r1179]]

* update configure and Makefiles

[[Imported from SVN: r1100]]

Really aggregate isolated nodes.
Correctly use all aggregates (isolated and connect) to compute
coarsening rate!

[[Imported from SVN: r1088]]

Made parallel SOR work with amg.

[[Imported from SVN: r1079]]

Parallel AMG works on helics with repartitioning!

matrixredistribute.hh : Ensure that overlap rows have the right entries
matrixutils.hh : Added method to print a distributed sparse matrix
amg.hh: initialize pointer with zero to prevent deallocation of nonallocated storage.
hierarchy.hh: print operator complexity in verbose mode.
transfer.hh: removed wrong assertions (yes, Index sets can have holes!)
repartition.hh: Bugfix: Use the correct MPI datatype to unpack. (Unpacking a std::size_t using MPI_INT on 64 bit is not a good idea!)

[[Imported from SVN: r1076]]

ParMETIS. (The additive version is not yet debugged!!!)
Graph::maxVertex does what it says now
Made index sets accessible in RemoteIndices class.
Fixed some typos.

[[Imported from SVN: r1072]]

[[Imported from SVN: r1056]]

Randomize coarsest aggregate numbers to get better contrast between
aggregates.

[[Imported from SVN: r1043]]

[[Imported from SVN: r1025]]

(Using the static streams of DUNE is just a pain as we need to
recompile to activate debugging.

[[Imported from SVN: r1014]]

DO not use sequential superlu in parallel mode!

[[Imported from SVN: r924]]

[[Imported from SVN: r916]]

[[Imported from SVN: r911]]

[[Imported from SVN: r905]]

[[Imported from SVN: r865]]

[[Imported from SVN: r828]]

[[Imported from SVN: r798]]

[[Imported from SVN: r751]]

Still we need a direct solver here *sigh*

[[Imported from SVN: r708]]

Prevent error accumulation in smoothing.

[[Imported from SVN: r703]]

[[Imported from SVN: r677]]

[[Imported from SVN: r601]]

[[Imported from SVN: r600]]

[[Imported from SVN: r582]]

Fixed premature checkin.

[[Imported from SVN: r557]]

Further tests will follow

[[Imported from SVN: r556]]