Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fixes errors in ghosting update for some overset mesh problems. #1214

Merged
merged 4 commits into from
Dec 23, 2023
Merged

Fixes errors in ghosting update for some overset mesh problems. #1214

merged 4 commits into from
Dec 23, 2023

Conversation

ddement
Copy link
Contributor

@ddement ddement commented Sep 19, 2023

Instead of trying to modify the ghosting for overset problems with relative motion, we now just rebuild the ghosting from scratch every time update_ghosting is called. This appears to fix errors such as those reported in #936

@ddement
Fixes errors in ghosting update for some overset mesh problems.
4c5ae63 
Instead of trying to modify the ghosting for overset problems with
relative motion, we now just rebuild the ghosting from scratch.
This appears to fix errors such as those reported in #936
@ashesh2512
Copy link
Contributor

ashesh2512 commented Sep 19, 2023 •
edited
Loading

@ddement So we are essentially recreating the ghosting object every time step. Is that accurate? Did something with compute_precise_ghosting_lists change?

@ddement
Copy link
Contributor Author

ddement commented Sep 19, 2023

@ashesh2512 Exactly. I'm sure there's a more efficient way to incrementally update it like we were attempting to do previously, but this approach should always produce a correct ghosting. If this turns out to be a significant cost, we can readdress optimizing it at a later point.

@alanw0
Copy link
Contributor

alanw0 commented Sep 19, 2023

That's right. At each time step there is a list of elements that need to be ghosted. The list for the next time step includes a lot of the same elements, but some new elements get added, and some get dropped. We were trying to just incrementally update the list, rather than completely dropping the old list and adding the new list. (I worked on this years ago with Stefan.) Apparently the incremental update wasn't correct, so dropping all of the old ones and then ghosting the new list is easier to get correct. Hopefully the performance difference shouldn't be too great. Like David says, we can revisit this if there is a big performance difference.

@ddement ddement enabled auto-merge (squash) September 20, 2023 15:51
@ddement ddement disabled auto-merge September 20, 2023 15:55
@ashesh2512
Copy link
Contributor

@gantech Could you please try the changes in this PR, for your simulations?

@alanw0
Copy link
Contributor

alanw0 commented Oct 3, 2023

We think this PR is ready to merge. I'm only hesitating because I don't want to breathe on anything until the challenge problem is finished.

@ashesh2512
Copy link
Contributor

We think this PR is ready to merge. I'm only hesitating because I don't want to breathe on anything until the challenge problem is finished.

I think that is a good life choice considering recent events

psakievich
psakievich approved these changes Oct 3, 2023
View reviewed changes
Copy link
Contributor

@psakievich psakievich left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've stalled merging everything until we get it done as well. There are several pending PR's that are ready to merge.

@psakievich psakievich merged commit 70e576b into master Dec 23, 2023
3 checks passed
@psakievich psakievich deleted the fix_ghosting_update branch December 23, 2023 05:16
psakievich added a commit that referenced this pull request Apr 30, 2024
@wjhorne @psakievich
@neilmatula @gantech @saitoasukakawaii @rcknaus @djglaze @ddement @jrood-nrel @itopcuoglu @mbkuhn
Rebase Multiphase_Dev and add VOF overset dpdx improvements (#1239)
003056d 
* Add Timers for FSI (#1221)

* Add Timers for FSI

* Style

* Add FSI section to Input File docs. (#1220)

* Add FSI section to Input File docs.

* Updating the FSI docs to incorporate PR comments.

* Relax projection for FSI mapping to account for curvature (#1223)

* Relax projection for FSI mapping to account for curvature

* Style

* Format again

* Add option to dump mesh on failed jacobian check (#1226)

* Add option to dump mesh on failed jacobian check

* Style

* Cleaner, parallel consistent impl

* Correctly set current_coordinates for restart (#1227)

* Actually set current_coordinates for restart

* Remove redundant call updating displacements

* Revert "Remove redundant call updating displacements"

This reverts commit dca015c.

* Fix restart issue for FSI simulations (#1228)

---------

Co-authored-by: Ganesh Vijayakumar <[email protected]>

* Update FieldRegistry.h (#1229)

missing std::.

* FSI: Set ramping defaults to false (#1231)

* FSI: Set ramping defaults to false

Since we are moving to split meshes as the current strategy setting
ramping defaults to `False`.

I have been investigating why the temporal ramping is causing
simulations to fail since the bug fixes regarding hub motion, and I
can't find any issues with it at the moment. So it seems best to keep
that turned off as well.

* Update real defaults

* add a string-function temperature IC/dirichlet bc (#1198)

* add a string-function temperature IC/dirichlet bc

* undo stk deprecation fixes

---------

Co-authored-by: psakievich <[email protected]>

* Converted to the new STK simple_fields workflow (#1233)

STK is migrating to a new strategy for registering and managing
Fields, where sizing information is purely specified at run-time
instead of the previous technique of specifying it in a confusing
blend of both compile-time and run-time information.  The compile-time
specification was just a suggestion, as it could be overridden
(possibly inconsistently) at run-time to support variable-length
Fields.  This made it unclear what the true size of a Field was
and where it should be specified.

As an example, registering a vector field on the entire mesh
previously looked like this:

  using VectorField = stk::mesh::Field<double, stk::mesh::Cartesian3d>;
  VectorField & field = meta.declare_field<VectorField>(stk::topology::NODE_RANK, "velocity");
  stk::mesh::put_field_on_mesh(field, meta.universal_part(), 3, nullptr);

and now, it looks like this:

  using VectorField = stk::mesh::Field<double>;
  VectorField & field = meta.declare_field<double>(stk::topology::NODE_RANK, "velocity");
  stk::mesh::put_field_on_mesh(field, meta.universal_part(), 3, nullptr);

  stk::io::set_field_output_type(field, stk::io::FieldOutputType::VECTOR_3D); // Optional

The only template parameter for a Field is now the datatype parameter.
Sizing information now exclusively comes from put_field_on_mesh() calls.
The optional set_field_output_type() function call registers with the
IO sub-system how a multi-component Field should be subscripted in
Exodus files.  If this call is left off, you will get the default
[_1, _2, _3] subscripting.  With the above call, you will instead get
[_x, _y, _z] subscripting.

The MetaData::use_simple_fields() flag is set everywhere possible in
the code to prevent accidental regressions before the old behavior
is formally deprecated and removed.  This will yield a run-time error
if the old-style extra template parameters are used anywhere.  These
calls to use_simple_fields() can be removed in the future once the
STK Mesh back-end has removed support for the old behavior.

This wasn't a completely straightforward conversion due to nalu-wind
making heavy use of various algorithm selections based on the
templated Field type.  The ScalarFieldType, VectorFieldType,
TensorFieldType, and GenericFieldType types are now all identical,
so different techniques had to be used to switch behaviors.

* Fixes errors in ghosting update for some overset mesh problems. (#1214)

Instead of trying to modify the ghosting for overset problems with
relative motion, we now just rebuild the ghosting from scratch.
This appears to fix errors such as those reported in #936

Co-authored-by: dcdemen <[email protected]>
Co-authored-by: psakievich <[email protected]>

* Revert "Converted to the new STK simple_fields workflow (#1233)" (#1234)

This reverts commit 8e8f4d5.

* Fixed the iblank inconsistency at shared nodes. The iblank field is returned to Tioga after the stk::mesh::copy_owned_to_shared operation, prior to field interpolation (#1238)

* Multiphase milestone (#1222)

* Divide through by density to get velocity form

* Fix VOF velocity flux to finalize

* Fix missing pressure gradient density norm and sharpen interfaces more

* Add diffusion term to VOF

* center droplet and provide velocity

* first version of sloshing tank

* sloshing tank case, can change parameters

* Mass-momentum consistency with initial vel scale and diffusion values

* Remove unused forced mass flux from pressure equation

* clean-up and safe settings for VOF advection

* Density face definition for minimized buoyancy noise

* rho_ref == initial density

* Change initial density to a specified function and introduce wall bcs for VOF

* sloshing tank pressure profile

* user function for generic flat water level in z

* Cleaning

* Balanced Buoyancy Forcing

* Complete milestone implementation

* Full implementation with fixed testing

* Fix allocation of fields

* Formatting

* Documentation of VOF advection scheme

---------

Co-authored-by: whorne <[email protected]>
Co-authored-by: Michael Kuhn <[email protected]>

* Rebase and pressure gradient improvements for VOF

* Fixed interface width work

* Allow slip of VOF at wall bcs

---------

Co-authored-by: psakievich <[email protected]>
Co-authored-by: neilmatula <[email protected]>
Co-authored-by: Ganesh Vijayakumar <[email protected]>
Co-authored-by: 四月是你的谎言 <[email protected]>
Co-authored-by: rcknaus <[email protected]>
Co-authored-by: djglaze <[email protected]>
Co-authored-by: ddement <[email protected]>
Co-authored-by: dcdemen <[email protected]>
Co-authored-by: Jon Rood <[email protected]>
Co-authored-by: itopcuoglu <[email protected]>
Co-authored-by: whorne <[email protected]>
Co-authored-by: Michael Kuhn <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ashesh2512 ashesh2512 ashesh2512 approved these changes

@alanw0 alanw0 alanw0 approved these changes

@psakievich psakievich psakievich approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@ddement @ashesh2512 @alanw0 @psakievich