GitLab

Simplify setup of the transfer operators

See merge request !53

The setup was needlessly complicated.  I think that's because
it predates the time where std::shared_ptr could be used.

I has virtual methods, and therefore also needs a virtual destructor.

Set the FE-order in finite-strain-elasticity back to: order = 1

See merge request !51

Fixes towards allowing stateful bases in TrustRegionSolver

See merge request !52

Currently the TrustRegionSolver class default-constructs a few bases
objects of its own.  That only works if the actual bases used for
the simulation does not contain state, because the state doesn't
make it into the new bases.

Fix this by avoid the default constructor.

There are remaining issues for the HAVE_DUNE_PARMG case, which
I indent to ignore for now.

Johnny has left science.  Thanks for all his work!

In the default setting, we want finite-strain-elasticity to compile with dune-parmg,
and dune-parmg can only handle FE of order=1.

MooneyRivlinDensity: Correct comment

See merge request !50

Small suggestions

See merge request !46

Trustregionsolver: re-enable parmg with powerBases

See merge request !48

With a small hack in dune-parmg it is now possible to run parallel code
in the TrustRegionSolver again. To to this, a preprocessor flag is defined.

Add wriggers-l-shape-dirichlet-values.py in dune-elasticity as well - we always took it from the module dune-gfe

Mooney-Rivlin energy: parameter are always `double`

See merge request !44

Make mooneyrivlindensity.hh beautiful again

See merge request !47

This improves the automatic differentiation a bit.

Simplify the calculation in mooneyrivlindensity.hh

See merge request !45

Simplify the calculation in mooneyrivlindensity.hh: Do not calulate the eigenvalues of F^TF and replace it by a function of trace(F^TF)

Add missing include (grid/common/partitionset.hh) in feassembler.hh, as the FEAssembler uses Dune::Partitions::interiorBorder

Improved calculation of hyperdual sqrt

See merge request !43

Fix: Implement std::numeric_limits and Dune::MathematicalConstants for hyperdual numbers

See merge request !42

This is simply done by forwarding to the `double` implementation.
This is safe since `hyperdual` numbers can be constructed from `double` numbers.

Implementation of LocalHyperDualStiffness

See merge request !41

 - implementation of hyper-dual numbers is added
 - LocalHyperDualStiffness is a new alternative to LocalADOLCStiffness that uses hyper-dual numbers instead of AD by ADOL-C
 - localhyperdualstiffnesstest verifies the results of LocalHyperDualStiffness

Fix for compiling mooneyrivlindensity: Do not wrap return statements in if-statements

See merge request !40

LocalADOLCStiffness: use "hessian2" instead of "hess_vec"

See merge request !39

With this commit, we do two things:

-  replace "hess_vec" with a simpler call of  "hessian" which
   calls internally "hess_vec".

- use "hessian2" by default instead of "hessian". "hessian2" calls
  internally "hess_mat" which is faster than "hessian/hess_vec".
  If the user encounters probems with the new "hessian2" driver, a
  boolean "useHessian2" can be set to false in the constructor.

Avoid direct calls of std:: namespaces

See merge request !38

Fix/minor improvements

See merge request !35

Whitespaces only: Adjust indentation for commit 403dc032 (Add try{}catch{} around the energy calculation and the solve() call inside the trustregion solver)