Implement Symbolic Recursive Integration for Dommaschk #1496
+302
−185
Conversation
This file contains 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
| benchmark_name | dt(%) | dt(s) | t_new(s) | t_old(s) |
| -------------------------------------- | ---------------------- | ---------------------- | ---------------------- | ---------------------- |
test_build_transform_fft_lowres | +0.15 +/- 4.29 | +7.86e-04 +/- 2.25e-02 | 5.25e-01 +/- 1.7e-02 | 5.24e-01 +/- 1.4e-02 |
test_equilibrium_init_medres | +1.23 +/- 1.04 | +5.09e-02 +/- 4.31e-02 | 4.18e+00 +/- 3.3e-02 | 4.13e+00 +/- 2.7e-02 |
test_equilibrium_init_highres | +0.40 +/- 2.62 | +2.16e-02 +/- 1.42e-01 | 5.44e+00 +/- 1.3e-01 | 5.42e+00 +/- 6.7e-02 |
test_objective_compile_dshape_current | -0.46 +/- 4.45 | -1.89e-02 +/- 1.82e-01 | 4.08e+00 +/- 1.3e-01 | 4.09e+00 +/- 1.3e-01 |
test_objective_compute_dshape_current | -1.07 +/- 1.78 | -5.53e-05 +/- 9.21e-05 | 5.12e-03 +/- 6.2e-05 | 5.18e-03 +/- 6.8e-05 |
test_objective_jac_dshape_current | +0.54 +/- 7.61 | +2.29e-04 +/- 3.23e-03 | 4.27e-02 +/- 2.4e-03 | 4.24e-02 +/- 2.2e-03 |
test_perturb_2 | +0.64 +/- 2.09 | +1.26e-01 +/- 4.10e-01 | 1.97e+01 +/- 8.1e-02 | 1.96e+01 +/- 4.0e-01 |
test_proximal_freeb_jac | +0.47 +/- 1.01 | +3.44e-02 +/- 7.47e-02 | 7.43e+00 +/- 3.5e-02 | 7.40e+00 +/- 6.6e-02 |
test_solve_fixed_iter | -0.26 +/- 2.08 | -8.19e-02 +/- 6.60e-01 | 3.17e+01 +/- 5.0e-01 | 3.18e+01 +/- 4.4e-01 |
test_LinearConstraintProjection_build | -1.51 +/- 2.93 | -1.57e-01 +/- 3.05e-01 | 1.02e+01 +/- 7.9e-02 | 1.04e+01 +/- 2.9e-01 |
test_build_transform_fft_midres | +0.48 +/- 2.54 | +2.90e-03 +/- 1.52e-02 | 6.01e-01 +/- 1.3e-02 | 5.99e-01 +/- 8.6e-03 |
test_build_transform_fft_highres | +0.12 +/- 2.35 | +1.14e-03 +/- 2.25e-02 | 9.61e-01 +/- 2.0e-02 | 9.60e-01 +/- 1.1e-02 |
test_equilibrium_init_lowres | +0.82 +/- 2.07 | +3.08e-02 +/- 7.80e-02 | 3.80e+00 +/- 7.3e-02 | 3.77e+00 +/- 2.7e-02 |
test_objective_compile_atf | +1.24 +/- 2.50 | +1.01e-01 +/- 2.04e-01 | 8.26e+00 +/- 9.4e-02 | 8.16e+00 +/- 1.8e-01 |
test_objective_compute_atf | +0.49 +/- 1.26 | +7.66e-05 +/- 1.97e-04 | 1.57e-02 +/- 1.5e-04 | 1.57e-02 +/- 1.2e-04 |
test_objective_jac_atf | +1.21 +/- 2.15 | +2.32e-02 +/- 4.11e-02 | 1.94e+00 +/- 3.2e-02 | 1.91e+00 +/- 2.6e-02 |
test_perturb_1 | +0.08 +/- 1.00 | +1.13e-02 +/- 1.44e-01 | 1.44e+01 +/- 1.1e-01 | 1.44e+01 +/- 8.8e-02 |
test_proximal_jac_atf | +0.11 +/- 1.13 | +8.76e-03 +/- 9.26e-02 | 8.22e+00 +/- 5.3e-02 | 8.21e+00 +/- 7.6e-02 |
test_proximal_freeb_compute | +0.25 +/- 1.47 | +4.98e-04 +/- 2.92e-03 | 2.00e-01 +/- 2.3e-03 | 1.99e-01 +/- 1.8e-03 |
test_solve_fixed_iter_compiled | +0.25 +/- 0.51 | +5.04e-02 +/- 1.02e-01 | 2.03e+01 +/- 7.8e-02 | 2.03e+01 +/- 6.7e-02 | |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #1496 +/- ##
=======================================
Coverage 95.68% 95.69%
=======================================
Files 101 101
Lines 25604 25606 +2
=======================================
+ Hits 24500 24503 +3
+ Misses 1104 1103 -1
|
dpanici
requested review from
a team,
rahulgaur104,
f0uriest,
ddudt,
kianorr,
sinaatalay,
unalmis and
YigitElma
and removed request for
a team
dpanici
commented
Jan 8, 2025
f0uriest
requested changes
Jan 21, 2025
dpanici
requested review from
a team and
f0uriest
and removed request for
a team
f0uriest
approved these changes
Jan 22, 2025
|
Moving notes to separate comment so it is not in PR commit comment NotesI've found that using sympy and symbolic integration to use the recursive form of the Dommaschk potentials to be a much more accurate way to use them (accurate as in, recreating poincare plots from the IPP report linked here, which previously I failed to find any flux surface using the nonrecursive algorithm) Previous implementation was non-recursive and suffered from issues for certain more complex potentials, which were either due to numerical issues in the algorithm (precision issues) or due to an incorrect implementation on my part (which I never was able to track down). Either way, this symbolic method seems to work much better. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Resolves #1467
Resolves #1278
Changes implementation of Dommaschk potentials to use
sympyand symbolic integration to recursively compute the potentials with symbolic algebra. I then use thesympy.lambdifyfunction to turn this into a jax-differentiable expression usable with ourScalarPotentialFieldclass.