Level set: composition #14
Conversation
|
|
||
| /** | ||
| * TODO | ||
| */ | ||
| TimeStepping::Scheme time_step_scheme; |
source/level_set_okz.cc
Outdated
| @@ -262,6 +262,7 @@ LevelSetOKZSolver<dim>::initialize_data_structures() | |||
| params.convection_stabilization = this->parameters.convection_stabilization; | |||
| params.do_iteration = this->parameters.do_iteration; | |||
| params.tol_nl_iteration = this->parameters.tol_nl_iteration; | |||
| params.time_step_scheme = this->parameters.time_step_scheme; | |||
| @@ -113,9 +113,8 @@ class LevelSetOKZSolverAdvanceConcentration | |||
| const LevelSetOKZSolverAdvanceConcentrationBoundaryDescriptor<dim> &boundary, | |||
| const MatrixFree<dim> & matrix_free, | |||
| const LevelSetOKZSolverAdvanceConcentrationParameter & parameters, | |||
| double & global_max_velocity, | |||
| const DiagonalPreconditioner<double> & preconditioner, | |||
| AlignedVector<Tensor<1, dim, VectorizedArray<double>>> &evaluated_convection); | |||
There was a problem hiding this comment.
@mschreter Please not the updated interface.
| AssertThrow(degree_u >= 2 && degree_u <= 5, ExcNotImplemented()); \ | ||
| AssertThrow(degree_u >= 1 && degree_u <= 5, ExcNotImplemented()); \ | ||
| AssertThrow(ls_degree >= 1 && ls_degree <= 4, ExcNotImplemented()); \ | ||
| if (ls_degree == 1) \ | ||
| { \ | ||
| if (degree_u == 2) \ | ||
| if (degree_u == 1) \ | ||
| OPERATION(1, 1); \ | ||
| else if (degree_u == 2) \ |
| /** | ||
| * TODO: needed? this is equivalent to `fe.tensor_degree()+1`? | ||
| */ | ||
| unsigned int concentration_subdivisions; |
|
@kronbichler This PR contains the hard work of two very intensive days by @mschreter and myself, trying to to use more modules of adaflo in MeltPoolDG. The work here is definitely no finished yet: 1) there are many TODOs, which is mainly related to the fact that we have not had the time to document everything to a level as we wanted and 2) we have only had time to integrate the advection operator till now, so that the interfaces, as proposed here, might still change. Nevertheless, I think it would be a good idea to merge so that I can open the follow-up PR for the simplex support and other follow-up PRs have a more human size. What do you think? I have run all tests locally: all are still passing. |
| // [PM]: The following code has been removed since it was not used in any test | ||
| // and it introduced annoying cyclic dependencies. |
There was a problem hiding this comment.
This is experimental code anyway, so it can remain in this state until there appears a need for it again with some additional experiments.
This PR restructures the level-set module so that
reinitialization,advection,curvature_computation, andnormal_vector_computationbecome operators on their own and thelevel_setoperator uses them via composition.FYI @mschreter The dependencies between the modules still need to be made more clear but this should allow us to use them from extern.