feat: SympyStepper

[andiwand]

An attempt to use generated code instead of Eigen for stepping, jacobians and covariance transport.

[andiwand]

The covariance transport seems to be twice as fast which results in a ~7% speed up of the trackfinding.

The stepping code is not as fast as it could be yet looking at the AtlasStepper performance. This needs a bit more tweaking of the codegen. It looks like the stepping with and without cov can twice as fast as well.

[github-actions]

📊: Physics performance monitoring for 94b2a2f

Full contents

physmon summary

• ✅ CKF truth_smeared
• ✅ IVF truth_smeared
• ✅ AMVF truth_smeared
• ✅ Track Summary CKF truth_smeared
• ✅ Seeding truth_estimated
• ✅ CKF truth_estimated
• ✅ IVF truth_estimated
• ✅ AMVF truth_estimated
• ✅ Track Summary CKF truth_estimated
• ✅ Seeding seeded
• ✅ CKF seeded
• ✅ IVF seeded
• ✅ AMVF seeded
• ✅ AMVF (+grid seeder) seeded
• ✅ Track Summary CKF seeded
• ✅ Seeding orthogonal
• ✅ CKF orthogonal
• ✅ IVF orthogonal
• ✅ AMVF orthogonal
• ✅ Track Summary CKF orthogonal
• ✅ Ambisolver seeded
• ✅ Ambisolver orthogonal
• ✅ Seeding ttbar
• ✅ CKF ttbar
• ✅ Ambisolver
• ✅ Track Summary CKF ttbar
• ✅ AMVF ttbar
• ✅ AMVF (+grid seeder) ttbar
• ✅ Truth tracking (GSF)
• ✅ Truth tracking
• ✅ Truth tracking (GX2F)
• ✅ Particles fatras
• ✅ Particles geant4

[andiwand]

I reverted to using the EigenStepper in the Examples after testing the SympyStepper in the CI. I think for now it makes sense to consistently rely on the EigenStepper.

At some point we should invest some time to properly monitor the different steppers CPU and physics wise. That way we can guarantee that they all work as they should.

The user should be able to transparently swap the EigenStepper out and use the SympyStepper after this PR which should come with some CPU time improvements while physics should stay (mostly up to FP magic) the same.

[asalzburger]

Let's go for it.

[codecov]

Codecov Report

Attention: Patch coverage is 53.67232% with 410 lines in your changes missing coverage. Please review.

Project coverage is 47.61%. Comparing base (57fd232) to head (94b2a2f).
Report is 250 commits behind head on main.

Files	Patch %	Lines
...e/src/Propagator/detail/codegen/sympy_jac_math.hpp	0.00%	242 Missing ⚠️
Core/src/Propagator/codegen/sympy_stepper_math.hpp	80.63%	0 Missing and 61 partials ⚠️
...re/src/Propagator/detail/SympyCovarianceEngine.cpp	20.00%	13 Missing and 31 partials ⚠️
Core/src/Propagator/SympyStepper.cpp	56.84%	5 Missing and 36 partials ⚠️
Core/include/Acts/Propagator/SympyStepper.hpp	73.80%	0 Missing and 11 partials ⚠️
Core/src/Propagator/detail/SympyJacobianEngine.cpp	0.00%	11 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3150      +/-   ##
==========================================
+ Coverage   47.42%   47.61%   +0.18%     
==========================================
  Files         499      507       +8     
  Lines       28278    29163     +885     
  Branches    13831    13996     +165     
==========================================
+ Hits        13410    13885     +475     
- Misses       4989     5260     +271     
- Partials     9879    10018     +139     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[acts-project-service]

✅ Athena integration test results [d7aa303]

✅ All tests successful

status	job	report
🟢	run_unit_tests
🟢	test_ActsCheckObjectCountsWorkflow
🟢	test_ActsEFTrackFit
🟢	test_ActsPersistifySeeds
🟢	test_ActsBenchmarkWithSpot
🟢	test_ActsAnalogueClustering
🟢	test_ActsWorkflowHeavyIons
🟢	test_ActsWorkflowFastTracking
🟢	test_ActsWorkflowCached
🟢	test_ActsWorkflow
🟢	test_ActsValidateAmbiguityResolution
🟢	test_ActsValidateResolvedTracks
🟢	test_ActsValidateTracks
🟢	test_ActsValidateActsCoreSpacePoints
🟢	test_ActsValidateActsSpacePoints
🟢	test_ActsValidateSeeds
🟢	test_ActsValidateOrthogonalSeeds
🟢	test_ActsValidateClusters
🟢	test_ActsPersistifyEDM
🟢	test_ActsGSFRefitting
🟢	test_ActsKfRefitting
🟢	test_ActsExtrapolationAlgTest
🟢	test_ActsITkTest
🟢	run_workflow_tests_run4_mc
🟢	run_workflow_tests_run2_mc
🟢	run_workflow_tests_run2_data
🟢	run_workflow_tests_run3_mc
🟢	run_workflow_tests_run3_data
🟢	run_art_test: test_data18_13TeV_1000evt
🟢	run_art_test: test_ttbarPU40_reco

[paulgessinger]

As discussed, if we want to rely on this more going forward, we will want to improve the documentation of the generating code.

I would also suggest we try to avoid committing the generated code to the repository.

[andiwand]

As discussed, if we want to rely on this more going forward, we will want to improve the documentation of the generating code.

I would also suggest we try to avoid committing the generated code to the repository.

Agreed - Generating the code in a CMake step or pulling it in if this is not possible would be much better. Since it is not a huge amount of code we decided to put it in temporarily. I added a small CI check to see if the generated code is what we expect to avoid this from getting out of sync.

I guess we have a similar problem here https://github.com/acts-project/acts/blob/a48fa8ac4d74f6647c64bb97e14d6d21276636f1/Core/src/Definitions/ParticleDataTable.hpp. Maybe we can find a common solution for this.

Documentations is lacking far behind on this I agree. I will try to put this in as soon as possible