Anasazi preconditioner #195
Conversation
|
Anasazi is merged. You can rebase. |
| @@ -22,6 +22,19 @@ enum class OperatorMode | |||
| TRANS | |||
| }; | |||
|
|
|||
| template <typename VectorType> | |||
| class OperatorBase | |||
There was a problem hiding this comment.
I don't really think that we need this new class but if you really want to add it at least make Operator derive from OperatorBase.
| size_type m() const { return _dof_handler.n_dofs(); } | ||
| double diag_element(const size_type i) const | ||
| { | ||
| auto x = _mesh_evaluator.matrix_free_get_agglomerate_diagonal(_dof_handler, |
There was a problem hiding this comment.
This is not a cheap operation. Can't you just get the whole vector at once.
| @@ -120,7 +126,7 @@ struct MatrixFreeAgglomerateOperator | |||
| * get_diag_elements() to the correct subspace. | |||
| */ | |||
| dealii::AffineConstraints<double> &_constraints; | |||
| }; | |||
| }; // namespace mfmg | |||
There was a problem hiding this comment.
Looks like there is a problem in the number of parenthesis.
| @@ -24,6 +25,71 @@ | |||
| namespace mfmg | |||
| { | |||
|
|
|||
| template <typename VectorType> | |||
| class DiagonalOperator : public OperatorBase<VectorType> | |||
There was a problem hiding this comment.
deal.II already has a class DiagonalMatrix
| { | ||
| _diag.resize(diag.size()); | ||
| std::transform(diag.begin(), diag.end(), _diag.begin(), | ||
| [](auto v) { return (v ? 1. / v : 0.); }); |
There was a problem hiding this comment.
You could have added a function for the inverse of the diagonal directly similar to https://github.com/ORNL-CEES/mfmg/blob/master/tests/test_hierarchy_helpers.hpp#L345 This way you also get a DiagonalMatrix right away.
Any explanation what this pull request does at all? |
Will rebase on
masteronce Anasazi is merged. All relevant changes are in the last commit.