Fix merge of master into implicit-fp-collisions - #376
Merged
mrhardman merged 153 commits intoJun 17, 2025
Merged
Conversation
Remove PyPlot from plots_post_processing to permit development withou…
Although other tests segfault with NCDatasets, HDF5 fails to precompile when using OpenMPI_jll unless NCDatasets is also installed. NCDatasets was originally removed from the 'debug checks' CI test only because it didn't work for some reason, so makes sense to add it back now that it works again on the CI server, and for some strange reason helps HDF5 to precompile.
Install NCDatasets in 'debug checks' CI test
Actions that use julia-actions/cache should have some permissions set, see https://github.com/julia-actions/cache#cache-retention.
Even after clearing the cache manually, the documentation workflow still fails unless we remove the julia-actions/cache action entirely. Don't understand why!
Disable cache in documentation workflow
…en adding multi sources)
This allows a quicker debug check for consistency of synchronization calls as macros can capture the file/line where they were called.
Convert _block_synchronize() and begin_*_region() to macros, more efficient debug check for synchronization consistency
For use in the implicit electron solve,
electron_kinetic_equation_euler_update!() is parallelised over
{z,vperp,vpa} and is expected to be called inside a non-parallelised
loop over r. Therefore in the explicit advance, the loop over r must not
use the @loop_r macro.
Otherwise runs likely fail because timesteps are inconsistent on different processes in the shared-memory block. Also simplifies slightly the broadcasting of code_time and pdf_electron_converged.
Use a shorter timestep to ensure the simulation does not crash when started without a restart from a Boltzmann-electron converged, steady-state solution.
…sation Fix parallelisation and typo in explicit electron solve
Change is physically equivalent, although discretization changes slightly. Updated version is more consistent with original notes deriving moment-kinetics model, and with version that was implemented for kinetic electrons. Some expected values for tests that use very tight tolerances are updated.
Increase z-resolution, decrease Lvpa (except in the split3 case), decrease timestep. Reduce order of vpa elements to make test more robust. Switch the test to use `elementwise_isapprox()` instead of `isapprox()` because `isapprox()` uses `LinearAlgebra.norm()`, which does not divide out the number of grid points, so that the tolerances would have to be increased when increasing the grid size.
…e at midpoint (electron steps not supported yet), which can be useful when gauging effectiveness of PI controllers - you can essentially 'see' the graphs of density and temperature motion without having to use makie_post_processing
…aming-term Use d(g_s)/dz instead of d(f_s)/dz in free-streaming term
This will allow the time derivatives of moments to be used directly in the kinetic equation.
Simplifies code, making fewer places that need to be updated if new terms are added (now only RHS of kinetic equation and moment equations, not zdot, wpadot, Fdot coefficients). Updates expected values in some tests: discretization errors are changed by the rearrangement of the moment kinetic advection coefficients, and the tests are generally poorly resolved (in order to be fast), so are sensitive to this.
We don't currently use a time advance scheme where the electron density (equal to the ion density) is updated during the same step as the electron pressure (because explicit and implicit steps are separate even if the electrons are updated directly on the ion timestep), but if we did, and also used the temperature equation option for the electrons (instead of the default pressure equation), then the implementation would be wrong because it did not distinguish the density at the beginning and the end of the timestep (compare the implementation of the parallel velocity update using a momentum equation in force_balance.jl). This commit fixes this potential error, by passing in separate electron_density_out and electron_density in to electron_energy_equation!().
Not useful for substituting the time derivatives of electron moments directly into the electron kinetic equation, because mass ratio simplifications are done after substituting the moment equations, but useful for diagnostics.
atol was set too small before, causing intermittent test failures on the CI servers.
…ils.jl Needs updating because the `get_electron_Jacobian_matrix()` function has all of the flags default to true, so need to set them all to false.
The changes in PR#322 increase the size of the dimensionless qpar by sqrt(2), so the previous value of atol may now be too small. The new version suffers from occasional faliures in `braginskii_electrons_imex_tests.jl`, so increase the atol slightly to avoid these.
Consistent 1D2V and 1D1V moment-kinetic equations
…ed results" This reverts commit b1344a2 and updates expected results of tests.
…#322 code" This reverts commit 24957ac.
These should not have affect simulation results as the advection speeds would be re-calculated before being used, but might possibly have affected the initial application of velocity space boundary conditions and also trigger the debug checks for use of uninitialized values.
* Use `elementwise_isapprox()` instead of `isapprox()`. * Decrease the solver tolerances to very small values to make the test results more consistent when there are rounding errors, different iteration counts on different machines or numbers of processes, etc. * Use the higher-order IMEX time-solver for the default case, as this is faster when using very tight time-solver tolerances.
Remove hacks needed to exactly match pre-PR#322 code
Speeds up compilation, I think by allowing the compiler to mostly optimise out the large if-elseif-... chain.
New test script `test_scripts/check-makie_post_processing.jl` runs a low-resolution case (similar to `moment_kinetics/tests/kinetic_electron_tests.jl`, but with neutrals and for very short output times) and then makes plots for its output, with 'all' plots enabled (post-processing input generated by `generate_maximal_input_file()`). Checks that `makie_post_process()` runs without errors.
Previously when animating a variable and the minimum and maximum of a variable were equal, tried to pass them to `ylims!()` anyway, which caused an error. This commit adds a check that the minimum and maximum are not equal before calling `ylims!()`, leaving the y-limits at the default values if they are equal.
Optimize compilation of `get_variable()`; CI check for `makie_post_process()`
Add missing factor of 1/2 in 'Reduction to 1D1V' docs
…-fp-collisions-really-merge-master
Collaborator
Collaborator
|
I just merged this hoping that I could then simply merge |
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
3 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.
Not sure what happened in #363. It looks like it resolved the differences with master, but didn't actually merge master. I think that's why #353 is still showing so many files changed. This PR merges master (from May 5th) into implicit-fp-collisions. When it's merged, the differences with master will look sensible.
@mrhardman making this as a PR into implicit-fp-collisions in case you were working on the #367 bug in another branch.