Platypus to MOOSE migration #29633
Platypus to MOOSE migration #29633
Conversation
…ake ProblemData a struct
…w EquationSystem is owned by ProblemData
…uring destructor calls
…hod names to match MOOSE style
…er inclusions in MFEMProblem
…lair/hephaestus-cleanup Remove ProblemBuilders and clean up MFEMProblem
refs #0
refs #0
loganharbour
force-pushed
the
mfem-conf
branch
from
f53dacd to
23d47ba
Compare
Yes - the tests that check that the output |
Yup, this should be straightforward by reducing tolerances/default no. of iterations (or just switching to smaller meshes...) |
| -DLAPACK_INCLUDE_DIRS="$PETSC_DIR/include" \ | ||
| -DCMAKE_INSTALL_PREFIX="$MFEM_DIR" \ | ||
| -DMFEM_USE_PETSC=YES \ | ||
| -DPETSC_DIR="$PETSC_DIR" \ |
There was a problem hiding this comment.
In addition to this -DPETSC_DIR is set twice in this set of cmake flags
|
|
||
| // clang-format off | ||
| //* in weak form | ||
| //* (\rho\nabla\times H, \nanbla\times v) + (\mu dH/dt, v) + (dB^e/dt, v) - <(\rho\nabla\times H)\times n, v> = 0 |
| # Build and install mfem | ||
| mkdir -p "$MFEM_BUILD_DIR" | ||
| cd "$MFEM_BUILD_DIR" | ||
| cmake .. \ |
There was a problem hiding this comment.
This is currently for an MFEM build for CPU only - we'll probably want a script to support users who wish to build for use on GPU too.
Probably out of scope for this PR, but a more generic Platypus build script is under development in Platypus PR #67
There was a problem hiding this comment.
Yeah, I have something wrapped up for a GPU environment but didn't wanna do it all at once here. There's already enough to review as is.
Due to the way that we to distribution, MFEM (and conduit) need to be built separately, ahead of time within the moose-dev container layer (which all CI and app distribution use at a later time). We could rely on a build of MFEM from platypus, but it would need to use a MFEM submodule in moose and also support an install
| mfem::Array<int> GetEssentialBdrMarkers(const std::string & name_, mfem::Mesh * mesh_); | ||
|
|
||
| void ApplyEssentialBCs(const std::string & name_, | ||
| mfem::Array<int> & ess_tdof_list, | ||
| mfem::GridFunction & gridfunc, | ||
| mfem::Mesh * mesh_); | ||
|
|
||
| void ApplyIntegratedBCs(const std::string & name_, mfem::BilinearForm & blf, mfem::Mesh * mesh_); | ||
|
|
||
| void ApplyIntegratedBCs(const std::string & name_, mfem::LinearForm & lf, mfem::Mesh * mesh_); |
There was a problem hiding this comment.
Agree with the above - should also note that we plan to update BoundaryConditions soon to make integrated BCs / boundary integrators handled more similarly to kernels/domain integrators - i.e. with boundary integrators stored in the EquationSystem, to facilitate easier mesh/FESpace updates.
|
|
||
| Preferentially, users should create an `MFEMVariable` with respect to an `MFEMFESpace`: | ||
|
|
||
| !listing test/tests/kernels/diffusion.i block=Problem FESpaces Variables |
There was a problem hiding this comment.
!listing paths to the test input files in documentation files are incorrect/outdated - these should point within test/tests/mfem now
| `MFEMProblem` methods are called by `Actions` during parsing of the user's input file, which add | ||
| and/or initialize members of the owned [MFEMProblemData](source/problem/MFEMProblemData.md) struct. | ||
| The order in which these actions are executed respects the dependencies declared in | ||
| [PlatypusApp](source/base/PlatypusApp.md). |
There was a problem hiding this comment.
All documentation paths for docs from Platypus including source should update to source/mfem - documentation is currently failing to create these links during build
refs #29632