[[Imported from SVN: r734]]

[[Imported from SVN: r731]]

Replaced it with #if HAVE_MPI

[[Imported from SVN: r726]]

[[Imported from SVN: r723]]

[[Imported from SVN: r722]]

[[Imported from SVN: r721]]

[[Imported from SVN: r720]]

[[Imported from SVN: r719]]

[[Imported from SVN: r717]]

[[Imported from SVN: r715]]

Bugfix: Number of rows has to be copied. Otherwise we get segfaults if each row was allocated independently

[[Imported from SVN: r714]]

[[Imported from SVN: r712]]

[[Imported from SVN: r711]]

[[Imported from SVN: r710]]

graph.
This is proposed for elasticity problems by several people.
(Works better for Olivers problems).

[[Imported from SVN: r709]]

Still we need a direct solver here *sigh*

[[Imported from SVN: r708]]

If this is the case find will return the end iterator.

Fixes flyspray #222

[[Imported from SVN: r707]]

[[Imported from SVN: r706]]

[[Imported from SVN: r705]]

Prevent using too much memory.

[[Imported from SVN: r704]]

Prevent error accumulation in smoothing.

[[Imported from SVN: r703]]

 c, int n, int m>A>

Only the nonzeros blocks are printed here (without any dots for zero entries).

[[Imported from SVN: r702]]

[[Imported from SVN: r700]]

In operator= for assigning a BCRSMatrix that was constructed without
specifying the number of nonzero one has to calculate the number of
nonzeros before allocating memory.

[[Imported from SVN: r699]]

[[Imported from SVN: r698]]

[[Imported from SVN: r697]]

[[Imported from SVN: r696]]

[[Imported from SVN: r695]]

[[Imported from SVN: r694]]

[[Imported from SVN: r693]]

writeMatrixToMatlab now write arbitrary ISTL matrices of arbitrary nesting depth.  this fixes FlySpray issue 201

[[Imported from SVN: r692]]

[[Imported from SVN: r691]]

[[Imported from SVN: r690]]

[[Imported from SVN: r689]]

const qualifier makes no sense.

[[Imported from SVN: r688]]

Do not overwrite index. Is this is a bug if lower_bound returns the
end iterator.

[[Imported from SVN: r687]]

added check whether row has nonzero size.

Fixes flyspray 202.

[[Imported from SVN: r686]]

Rely on MPI daemons being started.

[[Imported from SVN: r685]]

Rely on MPI daemons being started.

[[Imported from SVN: r684]]

The problem was with row pointers of empty rows being initialized to the same value
as the previous nonempty row (or to the first data element if no such rows existed).

Fixed this.

Reduced duplicate code.


[[Imported from SVN: r683]]