feat!: Add IVertexFinder interface, use in vertexing

[paulgessinger]

This PR introduces a new IVertexFinder with an interface that can support the existing vertex finder implementations, with the exception of the seed vertex finder (which depends on specific spacepoint triplets).

The interface introduces a type-erased state holder that the client of the finder initializes via the makeState interface function, and which the finder implementation downcasts to manipulate the state. This enables the concrete interface without templating.

Since some vertex finders use other vertex finders as seeders, this also allows to have the seed finder property use this IVertexFinder interface.

I did not observe CPU performance degradation as documented in #2842.

Part of:

• refactor: Vertex implementation #2842

Blocked by:

• refactor!: Use Delegate for track linearizers #2946

[codecov]

Codecov Report

Attention: 34 lines in your changes are missing coverage. Please review.

Comparison is base (0c753b1) 48.85% compared to head (ff46291) 48.82%.

Files	Patch %	Lines
...e/include/Acts/Vertexing/IterativeVertexFinder.ipp	47.36%	3 Missing and 7 partials ⚠️
...e/include/Acts/Vertexing/IterativeVertexFinder.hpp	20.00%	5 Missing and 3 partials ⚠️
...clude/Acts/Vertexing/AdaptiveMultiVertexFinder.hpp	42.85%	2 Missing and 2 partials ⚠️
...clude/Acts/Vertexing/AdaptiveMultiVertexFinder.ipp	77.77%	0 Missing and 4 partials ⚠️
Core/include/Acts/Vertexing/ZScanVertexFinder.hpp	33.33%	1 Missing and 1 partial ⚠️
...Acts/Vertexing/AdaptiveGridDensityVertexFinder.hpp	83.33%	0 Missing and 1 partial ⚠️
...Acts/Vertexing/AdaptiveGridDensityVertexFinder.ipp	66.66%	0 Missing and 1 partial ⚠️
...include/Acts/Vertexing/FullBilloirVertexFitter.ipp	50.00%	0 Missing and 1 partial ⚠️
...include/Acts/Vertexing/GridDensityVertexFinder.hpp	83.33%	0 Missing and 1 partial ⚠️
...include/Acts/Vertexing/GridDensityVertexFinder.ipp	66.66%	0 Missing and 1 partial ⚠️
... and 1 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2948      +/-   ##
==========================================
- Coverage   48.85%   48.82%   -0.03%     
==========================================
  Files         495      496       +1     
  Lines       28906    28938      +32     
  Branches    13731    13751      +20     
==========================================
+ Hits        14121    14130       +9     
- Misses       4899     4909      +10     
- Partials     9886     9899      +13     

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

[github-actions]

📊: Physics performance monitoring for ff46291

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
• ✅ Particles fatras
• ✅ Particles geant4

[paulgessinger]

@andiwand this is ready now.

[andiwand]

Left a few general comments. I am definitely in favor of removing such templated interfaces as I do not think we gain any runtime performance from it. On the other hand we lose time during compilation and non-templated code is easier to reason about IMO.

I wonder how far we are from actually using virtual inheritance and keeping the state inside the algorithm. We could even use a templated frontend to keep the ACTS state / algorithm separation intact.

I stumbled across this vertex seed finder and vertex finder interface a few weeks ago and it did not really make sense to me. I think these should be two different interfaces but maybe I am missing something.

[paulgessinger]

@andiwand The main argument against internal state is thread-safety, in my opinion. This separation defines well which parts are mutable and which ones are not, and we don't need to create / recreate the tools frequently.

[paulgessinger]

I seem to have broken physmon with the updates I added.