Topic linearsolvers#585
Draft
ax3l wants to merge 9 commits into
Draft
Conversation
ax3l
commented
Jun 5, 2026
ax3l
commented
Member
- rebase after MultiFabUtil & IO, AmrCore app support, ParmParse & ParallelDescriptor #584
Reintroduce `AMRCore` with trampolines for virtual members. Also add `AMRMesh` and GDB classes.
Add the Boundary module with bindings for:
- BCType and PhysBCType enums (AMReX_BC_TYPES.H)
- BCRec and Vector<BCRec>, incl. the setBC helper functions
- fill_domain_boundary (AMReX_BCUtil.H)
- PhysBCFunctNoOp, CpuBndryFuncFab,
PhysBCFunct<CpuBndryFuncFab> ("PhysBCFunct_CpuBndryFuncFab")
- PhysBCFunctUser: a physical boundary condition functor calling
back into Python at the MultiFab level, e.g., for external
Dirichlet (BCType.ext_dir) fills
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
pybind11 renders parameter/return type names in docstrings at binding definition time. ParticleContainer (ParGDBBase args), AmrParGDB (AmrCore ctor arg) and AmrCore.get_par_gdb (AmrParGDB return) were defined before the respective types were registered, leaving raw C++ type names in docstrings and failing pybind11-stubgen. Declare the AmrCore class before ParGDB and the particle containers, and add its member functions after ParGDB is registered. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add bindings for inter-level and inter-centering averaging operations (AMReX_MultiFabUtil.H): - average_down (cell-centered w/ volume weighting; centering-agnostic without), average_down_faces, average_down_edges, average_down_nodal - average_node_to_cellcenter, average_edge_to_cellcenter, average_face_to_cellcenter, average_cellcenter_to_face - sum_fine_to_coarse Per-direction MultiFab lists and destination component counts are validated: AMReX release builds otherwise corrupt memory on, e.g., a cell-centered destination with too few components. Add multi-level IO bindings: - write_multi_level_plotfile, plotfile path helpers (level_path, multifab_header_path, level_full_path, multifab_file_full_prefix), pre_build_director_hierarchy - util_create_directory, util_create_clean_directory, util_create_directory_destructive, file_exists Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Complete the AmrCore module so that full AMR applications can be written in Python, following the C++ AmrCore pattern (e.g., the Advection_AmrCore tutorial): - TagBoxArray & TagVal: cell tagging for refinement, with zero-copy Array4 (of char) views for error_est overrides - Array4<char> instantiations (Array4_char, Array4_char_const) - Interpolater & MFInterpolater class hierarchy and the global interpolater singleton instances (cell_cons_interp, ...) - fill_patch_single_level, fill_patch_two_levels, interp_from_coarse_level: MultiFab instantiations, overloaded for the PhysBCFunctNoOp / PhysBCFunct_CpuBndryFuncFab / PhysBCFunctUser boundary functors and dispatching on Interpolater vs MFInterpolater - FluxRegister: conservative flux corrections (refluxing) at coarse-fine interfaces - AmrMesh: per-level box_array, dist_map, set_box_array, set_dist_map, clear_box_array, clear_dist_map, blocking_factor, max_grid_size, n_error_buf, max_ref_ratio, count_cells, level_defined - AmrCore.error_est trampoline: pass the (non-copyable) TagBoxArray by reference so that tags set in Python propagate back to AMReX - BoxArray: == and != operators The new test_amrcore_fillpatch.py runs a full two-level AmrCore lifecycle in Python: init from scratch, threshold tagging, new level from coarse (verifying conservation of cell_cons_interp against average_down), level remake, and level removal. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Gaps found while translating the Advection_AmrCore tutorial to Python:
- ParmParse: complete the scalar get/query family (get_string,
query_bool, query_real, query_string) and add typed array accessors
(get_<T>_arr, query_<T>_arr, countval)
- ParmParse.to_dict: fix a crash on prefixed instances (the table
names are fully qualified; lookups now use an unprefixed instance)
- ParallelDescriptor: Barrier and Reduce{Real,Int,Long}{Min,Max,Sum},
ReduceBool{And,Or}; reduced values are returned (e.g., needed for
the C++ ComputeDt pattern of reducing per-level dt estimates)
- AmrMesh.set_finest_level: needed when restarting from checkpoints
- average_down_faces: add the periodicity-aware per-direction-list
overload (used by the no-subcycling advance to average down fluxes)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add the LinearSolvers module with bindings of the MultiFab
instantiations of the multi-level multigrid solver stack:
- enums: LinOpBCType, BottomSolver, Location; LPInfo
- the operator class hierarchy MLLinOp -> MLCellLinOp ->
MLCellABecLap -> {MLPoisson, MLABecLaplacian} and
MLLinOp -> MLNodeLinOp -> MLNodeLaplacian, with the operator
surface used by the AMReX tutorials: set_domain_bc, set_level_bc,
set_coarse_fine_bc (keep_alive: AMReX stores the raw pointer),
set_max_order, set_scalars, set_a_coeffs, set_b_coeffs, set_sigma
- MLMG: solve, get_grad_solution, get_fluxes (face and cell-centered),
comp_residual, and the iteration/bottom-solver controls
Tests solve the Poisson (single-level and two-level composite),
ABecLaplacian (MultiFab and constant coefficients) and nodal Poisson
problems against analytic solutions with second-order accuracy
checks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
|
|
||
| def test_singletons_identical(): | ||
| """Repeated access returns the same global instance""" | ||
| assert amr.cell_cons_interp is amr.cell_cons_interp |
|
|
||
| def test_physbcfunct_noop(std_box): | ||
| """PhysBCFunctNoOp leaves ghost cells untouched""" | ||
| geom = make_geometry(std_box) # noqa: F841 |
|
|
||
| def test_physbcfunct_user(std_box): | ||
| """PhysBCFunctUser calls back into Python with reference semantics""" | ||
| geom = make_geometry(std_box) # noqa: F841 |
| #include <sstream> | ||
|
|
||
|
|
||
| void init_FluxRegister(py::module& m) |
| } | ||
|
|
||
|
|
||
| void init_MultiFabUtil (py::module& m) |
| #include <vector> | ||
|
|
||
|
|
||
| void init_BCRec(py::module& m) |
| s << "<amrex.BCRec ("; | ||
| for (int dir = 0; dir < AMREX_SPACEDIM; ++dir) { | ||
| s << bcr.lo(dir); | ||
| s << (dir < AMREX_SPACEDIM-1 ? "," : ") ("); |
| } | ||
| for (int dir = 0; dir < AMREX_SPACEDIM; ++dir) { | ||
| s << bcr.hi(dir); | ||
| s << (dir < AMREX_SPACEDIM-1 ? "," : ")>"); |
| * instantiation. | ||
| */ | ||
| template< typename F > | ||
| void dispatch_mapper (InterpBase* mapper, F&& f) |
| } | ||
|
|
||
| template< typename BC > | ||
| void register_fill_patch (py::module& m) |
|
|
||
| /** per-level lists of per-direction (face) MultiFabs */ | ||
| Vector<Array<MultiFab*, AMREX_SPACEDIM>> | ||
| to_face_vector (std::vector<std::vector<MultiFab*>> const & in) |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.