Plex: fixes to recent merge

Merge branch 'tisaac/feature-fe-geom-dim' into tisaac/plex-flexible-projection

Use the new PetscFECellGeom

Conflicts:
include/petscdmplex.h
include/petscfe.h
src/dm/impls/plex/plexfem.c

PetscFE: added dim, dimEmbed to PetscFECellGeom

This is to allow PetscDualSpaceApply() to be used in situations where
a cell is embedded in a higher spatial dimension.

Merge branch 'tisaac/thplex'

This merge brings in related work in the following areas:
- Added a local constraint matrix to DM (see DMSetDefaultConstraints())
- Added point-to-anchor-point constrain

Merge branch 'tisaac/thplex'

This merge brings in related work in the following areas:
- Added a local constraint matrix to DM (see DMSetDefaultConstraints())
- Added point-to-anchor-point constraints in DMPlex (see
DMPlexSetAnchors()), which build on the DM constraints
- Added hierarchically-constrained points to DMPlex (see
DMPlexSetReferenceTree(), DMPlexSetTree()), which builds on the anchor
constraints
- Tests for these constraints in src/dm/impls/plex/examples/ex3.c (try
runex3_* examples in that directory)
- Added quadrature order (see PetscQuadratureGet/SetOrder()) (work by
Matt Knepley)
- Added default return values for DMGetCoordinateDim() (defaults to
DMGetDimension())

Conflicts:
include/petscdmplex.h
src/dm/dt/interface/dt.c
src/dm/impls/plex/examples/tests/ex3.c
src/dm/impls/plex/makefile
src/dm/impls/plex/plexfem.c

show more ...

Merge branch 'knepley/feature-fe-fv-hybrid'

* knepley/feature-fe-fv-hybrid: (43 commits)
PetscFE: Small fixes
Plex+SNES: Volumes are PetscReal
Plex+SNES: Missing declaration
TS ex11: Remove

Merge branch 'knepley/feature-fe-fv-hybrid'

* knepley/feature-fe-fv-hybrid: (43 commits)
PetscFE: Small fixes
Plex+SNES: Volumes are PetscReal
Plex+SNES: Missing declaration
TS ex11: Remove unused var
Plex+SNES+TS: Complete reorganization of residual evaluation - Now both FE and FV examples use the same method, but there are a bunch of switches which should be eliminated - There is a confusion about IFunction and RHS, but I think they can both be in one function - Added DMPlexSNESGetGeometryFVM() and DMPlexSNESGetGradientFVM() - Added DMPlexGet/RestoreCellFields() and DMPlexGet/RestoreFaceFields() and DMPlexGet/RestoreFaceGeometry() - I think that FV geometry caching needs to be redone - Added DMPlexReconstructGradients_Internal() - Added DMPlexComputeBdResidual_Internal() which does not work for FV - Replaced DMPlexComputeResidualFEM_Internal() with DMPlexComputeResidual_Internal()
Plex: Added DMPlexInsertBoundaryValues() - Unified FE and FV boundary values
Plex: DMPlexProjectFunctionLableLocal() gets discretizations from the DM now
Plex: Added DMPlexComputeGradientFVM()
PetscDS: Fix up some functions to work for both FV and FE
PetscFV: Change PetscFVIntegrateRHSFunction() to get info from PetscDS
TS ex11: There is only one field in our current paradigm - We would need to split Riemann solvers for multiple fields right now
Plex+TS: Remove use of computeGradients in favor of check for dmGrad
TS ex11: Updated interface calls
Plex+TS: Change DMPlexTSGetGradientDM() to automatically construct the gradient info, DMPlexTSSetRHSFunctionLocal() can now be replaced by the DMTS version - Remove unnecessary declarations
Plex+SNES: Add missing declaration
Plex+FE: Removed all use of PetscCellGeometry, Changed FE integration to use PetscFECellGeom
DMSNES+Plex: Added DMPlexSNESGetGeometryFEM() to cache geometry
PetscDualSpace: Remove use of PetscCellGeometry in PetscDualSpaceApply() - Fixed up calls to it for projection into spaces
DMDA: Remove use of PetscCellGeometry in DMDAProjectFunctionLocal()
Plex: Added DMPlexComputeGeometryFEM() and PetscFECellGeom structure
...

show more ...

Plex: skip empty loops for empty dual spaces

Plex: protect against field projection with height > 0.

The function DMPlexProjectFieldLocal() relies on PetscDS more than
PetscDualSpace, so using PetscDualSpaceGetHeightSubspace() won't work.

DMP

Plex: protect against field projection with height > 0.

The function DMPlexProjectFieldLocal() relies on PetscDS more than
PetscDualSpace, so using PetscDualSpaceGetHeightSubspace() won't work.

DMPlexProjectField() isn't used anywhere, and it isn't clear to me why
it looks its implementation is so different from
DMPlexProjectFunction(). It has added complexity because it uses field
jets instead of field values to evaluation the functions, but it tries
to project the resulting functions back into the original discretization
space. If I'm using P_1 Lagrange finite elements, and I want to
calculate a function involving its spatial derivatives, would I want to
project that function into a P_0 discontinuous approximation space?

Anyway, no support for now.

Plex: fixes to recently added (squash this)

show more ...

Plex: start implementing projection with height > 0.

Plex: Added DMPlexInsertBoundaryValues()
- Unified FE and FV boundary values

Plex: DMPlexProjectFunctionLableLocal() gets discretizations from the DM now

Plex: added DMPlexGet/SetMaxProjectionHeight()

The idea is to accomodate meshes (such as non-conforming meshes), where
a global basis function restricted to a cell may not correspond to a
single bas

Plex: added DMPlexGet/SetMaxProjectionHeight()

The idea is to accomodate meshes (such as non-conforming meshes), where
a global basis function restricted to a cell may not correspond to a
single basis function on the cell, but to a linear combination. The
most general case would be to do an L2 projection, but we would have to
find a way to invert control to use a KSPSolve(). This approach at
least expands the available set of local projection operations.

show more ...

Move DMPlexProject*() functions to src/ksp to remove ksp dependency in dm

Added DMGlobalToLocalSolve()

DMPlex: added DMPlexProjectConstraints_Internal()

DMPlex: early exits from empty partitions in plexfem.c

This avoids trying to get the closure of cell 0 in an empty process,
which was causing errors.

Conflict was DMPlexGetDimension()/DMGetDimensio

DMPlex: early exits from empty partitions in plexfem.c

This avoids trying to get the closure of cell 0 in an empty process,
which was causing errors.

Conflict was DMPlexGetDimension()/DMGetDimension()

Conflicts:
src/dm/impls/plex/plexfem.c

Conflicts:
src/dm/impls/plex/plexfem.c

show more ...

Plex+FE: Removed all use of PetscCellGeometry, Changed FE integration to use PetscFECellGeom

PetscDualSpace: Remove use of PetscCellGeometry in PetscDualSpaceApply()
- Fixed up calls to it for projection into spaces

Plex+SNES+TS: Moved residual computation to solver utilities

Plex: Moved BC stuff to plexbc.c

Plex+FE+FV: Fixed projection and L2 diff for FV
- Make a fake, simple PetscDualSpace for FV when projecting
- Added PetscFE/FVInterpolate_Static() for synthesis of field from coefficients

Merge branch 'knepley/feature-dm-dimension-handling'

* knepley/feature-dm-dimension-handling:
DM: Fixed DMDASetDim() in examples
Plex: Fix for isoparametric geometry
Plex: Complex fix
TS ex1

Merge branch 'knepley/feature-dm-dimension-handling'

* knepley/feature-dm-dimension-handling:
DM: Fixed DMDASetDim() in examples
Plex: Fix for isoparametric geometry
Plex: Complex fix
TS ex11: Fixed calls to DMSetCoordinateSection()
DM: FIxed DMClone() to copy dimension
Plex+DA: Now cell geometry routine can take a PetscFE in order to evaluate at quadrature points - We evaluate isoparametrically at the quadrature points - Changed DMDAComputeCellGeometry() --> DMDAComputeCellGeometryFEM() - Changed DMPlexComputeCellGeometry() --> DMPlexComputeCellGeometryFEM() - Added DMPlexComputeCellGeometryAffineFEM()
DM: Added a embedding dimension, different from the topological dimension - This can now be set in DMSetCoordinateSection(), or it can be automatically determined
Plex: Fix output for ex4
DM: Added DMGetDimPoints() - This replicates DMPlex/DAGetDepthStratum(), but uses the dimension which people have said is more intuitive
DM: Move dimension up into DM - Both DA and Plex had this and I think it makes sense

Conflicts:
src/dm/impls/plex/plexcreate.c

show more ...

Plex+DA: Now cell geometry routine can take a PetscFE in order to evaluate at quadrature points
- We evaluate isoparametrically at the quadrature points
- Changed DMDAComputeCellGeometry() --> DMDACo

Plex+DA: Now cell geometry routine can take a PetscFE in order to evaluate at quadrature points
- We evaluate isoparametrically at the quadrature points
- Changed DMDAComputeCellGeometry() --> DMDAComputeCellGeometryFEM()
- Changed DMPlexComputeCellGeometry() --> DMPlexComputeCellGeometryFEM()
- Added DMPlexComputeCellGeometryAffineFEM()

show more ...

DM: Move dimension up into DM
- Both DA and Plex had this and I think it makes sense

Merge branch 'knepley/feature-quad-fas'

* knepley/feature-quad-fas:
DMPlex: Proper preallocation for interpolators
DMPlex: Added events for interpolator and injector construction - Fixed event n

Merge branch 'knepley/feature-quad-fas'

* knepley/feature-quad-fas:
DMPlex: Proper preallocation for interpolators
DMPlex: Added events for interpolator and injector construction - Fixed event names
DM: Added sequence value for each number - Added DMOutputSequenceLoad() - Changed DMGet/SetOutputSequenceNumber() signature
Sys: Made a simple two-int hash - Someone should remove that other IJ crap
PetscDS: Renamed from PetscProblem
DMPlex: Fixed allocation
DMPlex: Support for injection from more complicated discretizations - Added DMPlexMatGetClosureIndicesRefined() - Added DMPlexComputeInjectorFEM() - Fixed check
PetscFE: Use in-cell test specific to cell shape
DMPlex: Added CellRefinerInCellTest_Internal()

show more ...

DMPlex: Proper preallocation for interpolators