palace::Mesh and libCEED quadrature data refactor
#166
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
sebastiangrimberg
added
enhancement
sebastiangrimberg
force-pushed
the
sjg/mat-coeff-revisions
branch
2 times, most recently
from
37672e3 to
f17ab7d
Compare
sebastiangrimberg
force-pushed
the
sjg/mat-coeff-revisions
branch
from
74e0b26 to
1905d0b
Compare
sebastiangrimberg
force-pushed
the
sjg/mat-coeff-revisions
branch
from
1905d0b to
7ac1abc
Compare
… handle material property coefficients on the mesh
…aredFaceTransformations and the need for passing around a local to shared face map
…tance for palace::FiniteElementSpace
Includes MUMPS fix for wave ports.
…or wave ports (no external dependencies, resolves some singular matrix issues)
…ore eigenvalue solver is destroyed)
hughcars
reviewed
Jan 16, 2024
There was a problem hiding this comment.
Mostly just minor things, typos, range loops etc. There's the possibility of some a slightly different design on the restricted coefficients, and some slightly expanded or simplified interface on integrators. Suggestions on hughcars/mat-coeff-revisions use whatever you like. Almost certainly missed some stuff as this was a lot of files to read.
Comment on lines
+127
to
+131
| for (auto it = bdr_attr_map.begin(); it != bdr_attr_map.end(); ++it) | ||
| { |
There was a problem hiding this comment.
for (const auto &[k,v] : bdr_attr_map)
{
loc_attr_list.Append(v);
}
There was a problem hiding this comment.
Sure, but I don't think this is substantially better than the iterator-based for loop.
| // Use default | ||
| case config::LinearSolverData::SymFactType::METIS: | ||
| this->SetReorderingStrategy(strumpack::ReorderingStrategy::METIS); | ||
| // this->SetReorderingStrategy(strumpack::ReorderingStrategy::AND); |
| break; | ||
| case config::LinearSolverData::SymFactType::AMD: | ||
| this->SetReorderingStrategy(strumpack::ReorderingStrategy::AMD); | ||
| // this->SetReorderingStrategy(strumpack::ReorderingStrategy::MMD); |
sebastiangrimberg
force-pushed
the
sjg/mat-coeff-revisions
branch
from
7ac1abc to
776ce39
Compare
hughcars
approved these changes
Jan 18, 2024
sebastiangrimberg
added a commit
that referenced
this pull request
Jan 18, 2024
Minor fixes following #166
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR aims to refactor the quadrature data construction for libCEED operators by precomputing the geometric factors (element Jacobian determinants, inverse, etc.) and using these in QFunctions rather than assembling and storing quadrature data for every operator. This also necessitates changing how material properties are handled as coefficients. The end result is improved memory usage and operator setup time (especially on GPUs, coming soon).
Summary of changes:
MaterialPropertyCoefficientclass which contains a mapping from attribute to material property and a list of material properties. This also introduces the concept of different domain and boundary element attributes: The MFEM attributes (global, starting at 1, do not need to be contiguous) are associated with the mesh. There are now libCEED attributes which are local to a process and contiguous, and conversions between the two.MaterialPropertyOperatorandMaterialPropertyCoefficientobjects work with the libCEED attributes. For certain models with hundreds or even thousands of attributes, this improves scalability. This set of changes simplifies a lot of the models and their interaction with material properties (for example, boundary conditions with dependence on neighboring element material properties, see wave ports). It is fully functional with boundarySubMeshobjects.palace::Meshclass which composes anmfem::ParMeshwith data for Palace (similar topalace::FiniteElementSpacewhich composes anmfem::ParFiniteElementSpaceand libCEED objects). Thepalace::Meshincludes functionality for precomputing libCEED geometry factor quadrature data like the libCEED attributes and Jacobian data, to be used in QFunctions.CeedBasisandCeedElemRestrictionobjects directly to their associatedFiniteElementSpace. This makes up a lot of the large number of file changes, and the changes are the same across all files. The unit tests are comprehensive here and ensure correctness across all integrators. Performance against the previous implementation has been validated.Resolves #122