fluids: Implement bc_slip

This is a weak version of slip boundary condition based on a Riemann
solve

fluids: Change -bc_slip_x -> -bc_symmetry_x

Merge pull request #1361 from CEED/jrwrigh/external_sgs

fluids: Add external SGS DD inference capability

Fixing SmartRedis meta data arrays to be of type Int64 (#1501)

* fluids: fixed SmartRedis meta data arrays to be of type Int64

* fluids: style fix

---------

Co-authored-by: Riccardo Balin <

Fixing SmartRedis meta data arrays to be of type Int64 (#1501)

* fluids: fixed SmartRedis meta data arrays to be of type Int64

* fluids: style fix

---------

Co-authored-by: Riccardo Balin <balin@aurora-uan-0010.hostmgmt.cm.aurora.alcf.anl.gov>

show more ...

Merge pull request #1496 from CEED/jrwrigh/ksp_mass

Fluids: Fix mass operator for explicit timestepping, move to KSP for mass operator

fixup! fluids: Use KSP for explicit mass matrix

fluids: Fix mass matrix for explicit time stepping

The local vector should be initialized to zero so that all the essential
nodes (ie. strong boundary condition nodes) are not factored in for the
ma

fluids: Fix mass matrix for explicit time stepping

The local vector should be initialized to zero so that all the essential
nodes (ie. strong boundary condition nodes) are not factored in for the
matrix multiply. The PR associated with this commit should have a
document attached detailing why this is the case.

Previous tests used this incorrect mass matrix and so now fail. The
results for these tests are corrected as well.

show more ...

fluids: Destroy stray Mats

fluids: Use KSP for explicit mass matrix

fluids: Remove convergence test from velocity_gradient_projection

Testing with GDB, the `Mass_N` QFunction is only called when creating
the rowsum preconditioner and isn't called afterwards. This is

fluids: Remove convergence test from velocity_gradient_projection

Testing with GDB, the `Mass_N` QFunction is only called when creating
the rowsum preconditioner and isn't called afterwards. This is true with
or without the convergence test. So I've removed the convergence test as
it's not necessary

show more ...

fluids: Refactor function names

- PascalCase for acronyms longer than 2 letters
- Remove the "Anisotropic" qualifier from function names, as we only
have one data-driven SGS model. If we add an is

fluids: Refactor function names

- PascalCase for acronyms longer than 2 letters
- Remove the "Anisotropic" qualifier from function names, as we only
have one data-driven SGS model. If we add an isotropic model later, we
can give that the "Isotropic" qualifier on it's function names
- Remove the "Model" qualifier from function names. "SgsDD" is
communicative enough (Subgrid-stress inherently means a model)

show more ...

fluids: Add LogEvent for VeloGradProj and DiffFilter

Merge balin/smartsim branch into main with timers for SmartRedis API calls (#1480)


---------

Co-authored-by: James Wright <james@jameswright.xyz>
Co-authored-by: Riccardo Balin <balin@aurora-u

Merge balin/smartsim branch into main with timers for SmartRedis API calls (#1480)


---------

Co-authored-by: James Wright <james@jameswright.xyz>
Co-authored-by: Riccardo Balin <balin@aurora-uan-0012.hostmgmt.cm.aurora.alcf.anl.gov>

show more ...

fluids: Replace if...SETERRQ with PetscCheck

fluids: Add missing OperatorApplyContextDestroy's

fluids: Report both amat and pmat types

fluids: Merge 2D and 3D advection setup

Instead of `-problem advection` and `-problem advection2d`, there's just
`-problem advection` and the dimensionality is taken from the dm.

fluids: Stats don't overwrite solution on first-step failure

- When you have an initial condition from a previous problem, it might
fail on the first step. This will cause `reason != TS_CONVERGED_

fluids: Stats don't overwrite solution on first-step failure

- When you have an initial condition from a previous problem, it might
fail on the first step. This will cause `reason != TS_CONVERGED_ITERATING`
and thus it will continue on with the rest of the TSMonitor routine.
If you have a stats file from that initial condition, it will be
overwritten.

Regardless of the overwriting behavior, it *never* makes sense to
write out a stats file on the first step; Using left-rectangle rule
for the time averaging integration, the initial step should *never* be
used, thus there is nothing to collect or write.

show more ...

interface - remove legacy CEED_BASIS_COLLOCATED

Merge pull request #1423 from CEED/jrwrigh/fluids_high_order_geom

fluids: Support higher-order geometry

fluids: Support higher-order geometry

fluids: Add uses_newtonian flag

fluids: Change SmartRedisCall -> PetscSmartRedisCall

fluids: Add SmartSimDataDestroy

fluids: Add -sgs_train_enable flag