[[Imported from SVN: r7961]]

[[Imported from SVN: r7959]]

Add an operator() with three arguments to grant access
to the individual entries.

[[Imported from SVN: r7934]]

[[Imported from SVN: r7925]]

[[Imported from SVN: r7913]]

[[Imported from SVN: r7906]]

[[Imported from SVN: r7905]]

[[Imported from SVN: r7898]]

[[Imported from SVN: r7886]]

simplify the code by making the AverageDistanceAssembler accept the interpolation weights in a std::vector of FieldVectors

[[Imported from SVN: r7885]]

[[Imported from SVN: r7848]]

[[Imported from SVN: r7847]]

[[Imported from SVN: r7842]]

[[Imported from SVN: r7838]]

[[Imported from SVN: r7837]]

[[Imported from SVN: r7617]]

[[Imported from SVN: r7611]]

[[Imported from SVN: r7553]]

That space is a product space, and its LocalGeodesicFEFunction can be
implemented much more efficiently than the generic method.

[[Imported from SVN: r7551]]

and barycentric coordinates.  This is easy because we know the
number of element corners at compile time (I currently don't
see support for non-simplex elements at all).  This saves 
time, but not much: about 1%.

[[Imported from SVN: r7550]]

[[Imported from SVN: r7548]]

[[Imported from SVN: r7547]]

[[Imported from SVN: r7412]]

Solve the minimization problems for the evaluation of gfe functions really down to machine precision.  This appears to make the overall algorithm much more stable

[[Imported from SVN: r7355]]

[[Imported from SVN: r7346]]

[[Imported from SVN: r7345]]

[[Imported from SVN: r7344]]

instead of a singular value decomposition.  This makes computing
a Hesse matrix roughly 20% faster.

[[Imported from SVN: r7232]]

Also: fix more bugs in the implementation of the derivative of gfe
function gradients wrt the coefficients.  This appears to work now,
but inexact evaluations remain an issue.

[[Imported from SVN: r7180]]

[[Imported from SVN: r7179]]

[[Imported from SVN: r7177]]

rename method assembleHessian of AverageDistanceAssembler to assembleEmbeddedHessian, to make clearer what it does

[[Imported from SVN: r7175]]

[[Imported from SVN: r7133]]

add a method with an fd approximation of the derivative of the function gradient.  only for testing purposes

[[Imported from SVN: r7130]]

[[Imported from SVN: r7128]]

[[Imported from SVN: r7127]]

[[Imported from SVN: r7126]]

[[Imported from SVN: r7125]]

transpose one gradient matrix.  This makes the code run, but it still doesn't work even for RealTuple<1>

[[Imported from SVN: r6547]]

[[Imported from SVN: r6546]]