**OUTDATED** writeData when subcycling is used by BenjaminRodenberg · Pull Request #1414 · precice/precice

BenjaminRodenberg · 2022-08-11T12:03:16Z

Main changes of this PR

Allows to write data to preCICE when subcycling is used. This either results in a piecewise constant or linear interpolation. Possibly also in higher-order interpolation.

Motivation and additional information

See #1171

Author's checklist

Reviewers' checklist

Does the changelog entry make sense? Is it formatted correctly?
Do you understand the code changes?

* Remove mapReadDataTo and mapWriteDataFrom. * Remove tests using removed API.

* Merge initialize and initializeData in SolverInterfaceImpl. * Merge initialize and initializeData in CouplingScheme. * Update tests correspondingly * Allow to check actions before initialize, especially checking whether initial data has to be written * Check for sending and receiving of initial data during coupling scheme configuration * Code simplification w.r.t waveform. * Use dt=0.0 in actions in initialize. Co-authored-by: Benjamin Uekermann <[email protected]> Co-authored-by: Frédéric Simonis <[email protected]>

…recice#1352) * Get tests and configs for serial implicit waveform interpolation from precice#1029. * Differentiate between received initial data and data (from advance). * Split data communication for SerialCouplingScheme and store initial data in waveform. * Move common functionality of serial and parallel coupling scheme into BiCouplingScheme. Co-authored-by: Benjamin Uekermann <[email protected]>

* Add test Integration/Serial/InitializeData/ImplicitBoth where both participants initialize. * Allow initialize=true for both participants in serial-implicit coupling scheme. * Raise warning, if first participant in serial coupling scheme initializes data with waveform order zero. * Remove check for write data, because participant does not know about waveform order. Co-authored-by: Benjamin Uekermann <[email protected]>

precice#1349)

Co-authored-by: Benjamin Uekermann <[email protected]>

* Make read mapping conditional to avoid unnecessary mappings. * Update corresponding tests. Co-authored-by: Benjamin Uekermann <[email protected]>

…sts for piecewise interpolation.

…olation when subcycling.

… for subcycling.

…Spline interpolation.

…very often.

BenjaminRodenberg · 2022-08-19T19:24:25Z

Summary of important changes of today:

I'm using Eigen's BSpline interpolation for higher order. This needs a sufficient number of samples on the window. If there is an insufficient number of sample, we will fall back to lower order interpolation on the window.
I removed the feature to use older samples (only properly implemented for quadratic interpolation and not accessible through API), because it made the implementation much more complicated and has no use-case.

* Remove mapReadDataTo and mapWriteDataFrom. * Remove tests using removed API.

* Increase timeouts for precice.cplscheme and precice.integration.Serial, because they will take longer with new tests introduced in #1414 Co-authored-by: Frédéric Simonis <[email protected]>

…-used

uekerman · 2022-11-21T10:43:43Z

src/cplscheme/BaseCouplingScheme.cpp

+  } else {
+    // work-around for explicit coupling, because it does not support waveform relaxation.
+    if (reachedEndOfTimeWindow()) {                     // only necessary to trigger at end of time window.
+      storeTimeStepSendData(time::Storage::WINDOW_END); // only write data at end of window
+    }
+  }


We could probably remove this if-else block if we always give the correct data to time and let time decide what to do with it.

uekerman · 2022-11-21T10:45:34Z

src/precice/impl/SolverInterfaceImpl.cpp

+  if (_couplingScheme->isTimeWindowComplete() && _couplingScheme->moveWindowBeforeMapping()) { /*
+     * move to next window before mapping, if second in serial coupling.
+     * Reason:
+     * Window is complete, but first participant has already performed an iteration in the new window and the data just received is already from the new window. So we want to use it, but before doing this we have to use the data from the old window to initialize the waveform at 0.0.
+     */


Discussed together: let's try to move the "before-mapping" branch up till the advance call from the previous iteration. Should always be possible when hacking isTimeWindowComplete.
But maybe we find a proper solution when seeing this refactoring?

…ultOfFirstAdvance() * Use time::Storage as receive data buffer * CplSchemeTests, TimeTests and some integration tests for time interpolation are failing

…. Remove support for second order extrapolation.

* Partially tested for serial coupling. * Integration/Serial/Time/Implicit/SerialCoupling/ReadWriteScalarDataWithWaveformSamplingZero works. * Initializing with zero is an open todo. * Integration/Serial/Time/Implicit/SerialCoupling/ReadWriteScalarDataWithWaveformSamplingFirstNoInit breaks.

* Integration/Serial/Time/Implicit/SerialCoupling works * Integration/Serial/Time/Implicit/ParallelCoupling breaks

* Integration/Serial/Time/Implicit/SerialCoupling works * Integration/Serial/Time/Implicit/ParallelCoupling works * Integration/Serial/Time/Implicit/MultiCoupling breaks

BenjaminRodenberg · 2022-12-02T10:13:44Z

Current state:

Most of the Integration tests for time interpolation are working again
I did not check the unit tests thoroughly, but there are still some issues, for example in CplSchemeTests. Probably minor.
Code structure is in my opinion much more understandable now, because we require less going back and forth between CouplingScheme and SolverInterfaceImpl. (I managed to remove CouplingScheme::receiveResultOfFirstAdvance, everything happens now directly in CouplingScheme::initialize())

Next steps:

Get the remaining tests working again (some tests in MultiCoupling are optional, because this is basically a new feature).
Try to also remove CouplingScheme::moveWindowBeforeMapping
Implement Extrapolation again, but inside Waveform.
Try to factor out changes (refactoring SolverInterfaceImpl/CouplingScheme and Extrapolation) into a smaller PRs.

BenjaminRodenberg · 2022-12-16T22:45:33Z

I tried to merge #1504 into this PR in BenjaminRodenberg@3f35b8c, but there are a lot of conflicts and I think it's probably easier to reimplement the changes from #1414 on top of #1504 instead of trying to fix the merge.

BenjaminRodenberg · 2022-12-22T12:56:19Z

This PR is outdated and won't be merged. I will port all relevant work from here to #1523 and then close this PR.

* Copied functions from precice#1414 * Requires still some debugging to get tests working

BenjaminRodenberg · 2023-02-23T10:50:23Z

Closing this PR, development will continue in #1523

BenjaminRodenberg and others added 10 commits August 11, 2022 13:02

Remove mapWriteDataFrom and mapReadDataTo (precice#1222)

118abaa

* Remove mapReadDataTo and mapWriteDataFrom. * Remove tests using removed API.

Remove isReadDataAvailable and isWriteDataRequired (precice#1362)

e9ab7dc

Add integration test to reproduce first-participant + initData problem (

5e67a8c

precice#1349)

Simplify signature of actions and remove ScaleByDtAction (precice#1403)

d99966c

Co-authored-by: Benjamin Uekermann <[email protected]>

Make read mapping conditional. (precice#1404)

e314f68

* Make read mapping conditional to avoid unnecessary mappings. * Update corresponding tests. Co-authored-by: Benjamin Uekermann <[email protected]>

Simplify test.

f674733

Add failing test for piecewise subcycling.

acd6eb1

BenjaminRodenberg added the enhancement A new feature, a new functionality of preCICE (from user perspective) label Aug 11, 2022

BenjaminRodenberg added this to the Version 3.0.0 milestone Aug 11, 2022

BenjaminRodenberg self-assigned this Aug 11, 2022

BenjaminRodenberg added 12 commits August 15, 2022 20:39

Remove unused.

987476b

WIP, broken: Use map for storing samples in waveform. Provide some te…

fabf77f

…sts for piecewise interpolation.

Fix wrong update of map.

d36b780

Fix some existing tests.

27ab861

Implement piecewise constant interpolation.

3581ea7

Add unit test for zeroth and first order piecewise interpolation.

6def768

Implement piecewise linear interpolation.

59950d4

Add some stubs for implementation of quadratic and third order interp…

45950f1

…olation when subcycling.

Introduce helper for quadratic interpolation. Add fake implementation…

1e5e18b

… for subcycling.

Introduce BSpline interpolation. Merge quadratic interpolation into B…

48363bb

…Spline interpolation.

Do testing and check some parametrizations. Bug: Wrong order is used …

7172524

…very often.

Don't allow using old windows. Cleanup.

548a485

BenjaminRodenberg added 4 commits August 19, 2022 23:25

More simplification and another test.

1744141

Simplify class.

25bc85b

Remove mapWriteDataFrom and mapReadDataTo (precice#1222)

5beb7b3

* Remove mapReadDataTo and mapWriteDataFrom. * Remove tests using removed API.

Remove isReadDataAvailable and isWriteDataRequired (precice#1362)

9fb5d8a

BenjaminRodenberg added 4 commits November 4, 2022 17:46

Merge branch 'develop-v3.0.0' into 1171-write-data-when-subcycling-is…

d0f93b8

…-used

Merge branch 'develop-v3.0.0' into 1171-write-data-when-subcycling-is…

71fbc92

…-used

Add a warning, if requested order is not reached.

7668d29

Merge branch 'develop' into 1171-write-data-when-subcycling-is-used

f886c75

uekerman reviewed Nov 21, 2022

View reviewed changes

BenjaminRodenberg added 5 commits November 25, 2022 10:28

WIP: Try to remove need to explicitly call CouplingScheme::receiveRes…

1a1d82b

…ultOfFirstAdvance() * Use time::Storage as receive data buffer * CplSchemeTests, TimeTests and some integration tests for time interpolation are failing

Remove implementation of extrapolation, will be moved inside Waveform…

0098a45

…. Remove support for second order extrapolation.

WIP: Take care of zero initialization

ba32bb0

* Integration/Serial/Time/Implicit/SerialCoupling works * Integration/Serial/Time/Implicit/ParallelCoupling breaks

WIP: Fix ParallelCoupling for converged.

4b0eb86

* Integration/Serial/Time/Implicit/SerialCoupling works * Integration/Serial/Time/Implicit/ParallelCoupling works * Integration/Serial/Time/Implicit/MultiCoupling breaks

BenjaminRodenberg mentioned this pull request Dec 3, 2022

Simplify compositional couplingscheme #1462

Merged

9 tasks

BenjaminRodenberg added 4 commits December 4, 2022 17:23

Remove tests with second order extrapolation.

248a29e

Remove unnecessary default values. Simplify Waveform class.

8b71ea5

Remove more second order extrapolation.

0605480

Fix some more tests.

ba03dc5

This was referenced Dec 4, 2022

Use time::Storage for sending and receiving data #1504

Closed

Add more parallel coupling scheme unit tests and QN integration tests #1507

Merged

BenjaminRodenberg changed the base branch from develop-v3.0.0 to develop December 12, 2022 12:11

BenjaminRodenberg added a commit to BenjaminRodenberg/precice that referenced this pull request Dec 16, 2022

Broken merge of precice#1504 into precice#1414

3f35b8c

BenjaminRodenberg mentioned this pull request Dec 20, 2022

Use data structure similar to time::Storage in mesh::Data::_values #1522

Closed

BenjaminRodenberg added a commit to BenjaminRodenberg/precice that referenced this pull request Dec 22, 2022

Get Integration Tests from precice#1414 and apply API updates.

e35265a

BenjaminRodenberg changed the title ~~WIP: writeData when subcycling is used~~ **OUTDATED** writeData when subcycling is used Dec 22, 2022

BenjaminRodenberg mentioned this pull request Dec 22, 2022

WIP: write data when subcycling is used (outdated) #1523

Closed

25 tasks

BenjaminRodenberg added a commit to BenjaminRodenberg/precice that referenced this pull request Dec 28, 2022

Reimplement most part of functionality

7ef1a6d

* Copied functions from precice#1414 * Requires still some debugging to get tests working

BenjaminRodenberg closed this Feb 23, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

OUTDATED writeData when subcycling is used#1414

OUTDATED writeData when subcycling is used#1414
BenjaminRodenberg wants to merge 163 commits intoprecice:developfrom
BenjaminRodenberg:1171-write-data-when-subcycling-is-used

BenjaminRodenberg commented Aug 11, 2022 •

edited

Loading

Uh oh!

BenjaminRodenberg commented Aug 19, 2022

Uh oh!

uekerman Nov 21, 2022

Uh oh!

uekerman Nov 21, 2022

Uh oh!

BenjaminRodenberg commented Dec 2, 2022

Uh oh!

BenjaminRodenberg commented Dec 16, 2022

Uh oh!

BenjaminRodenberg commented Dec 22, 2022

Uh oh!

BenjaminRodenberg commented Feb 23, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

BenjaminRodenberg commented Aug 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Main changes of this PR

Motivation and additional information

Author's checklist

Reviewers' checklist

Uh oh!

BenjaminRodenberg commented Aug 19, 2022

Uh oh!

uekerman Nov 21, 2022

Choose a reason for hiding this comment

Uh oh!

uekerman Nov 21, 2022

Choose a reason for hiding this comment

Uh oh!

BenjaminRodenberg commented Dec 2, 2022

Uh oh!

BenjaminRodenberg commented Dec 16, 2022

Uh oh!

BenjaminRodenberg commented Dec 22, 2022

Uh oh!

BenjaminRodenberg commented Feb 23, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

BenjaminRodenberg commented Aug 11, 2022 •

edited

Loading