Verbose performance metrics

[SylvainJoube]

I've added some performance metrics for the following algorithms: finding, fitting and ambiguity resolution. I think it's an easy way to have a simple evaluation of the algorithms performance, without having to open ROOT. The code compares the reconstructed track with the truth particle data and prints basic metrics (valid/duplicates/false tracks).

My code still needs a bit of refactoring (I will clean it by the end of the day). I will also write a documentation, if you agree to merge this PR, and it you think this PR was a good idea. Here is an example of output I had:

==> Statistics ... 
- read    7830 spacepoints
- read    7830 measurements
- created (cpu)  4794 seeds
- created (cpu)  5296 found tracks
- created (cpu)  5296 fitted tracks
- created (cpu)  1780 ambiguity free tracks

Performance metrics:

===== Performance metrics for finding =====
          Valid: 1786 (34%)
     Duplicates: 2170 (41%)
          Fakes: 1340 (25%)

===== Performance metrics for fitting =====
          Valid: 1786 (34%)
     Duplicates: 2170 (41%)
          Fakes: 1340 (25%)

===== Ambiguity resolution performance metrics =====
--Among the selected tracks:
  Valid quality: 0.000561798 (should be as low as possible)
          Valid: 1766 (99%)
     Duplicates: 0 (0%)
          Fakes: 14 (1%)
--Among the evicted tracks:
          Valid: 20 (1%) (not in selected tracks)
     Duplicates: 2137 (61%)
          Fakes: 1326 (38%)

===== Performance metrics for ambiguity resolution (check v2) =====
          Valid: 1766 (99%)
     Duplicates: 0 (0%)
          Fakes: 14 (1%)

I used the following command, with the added --print-performance flag:

./bin/traccc_seeding_example --input-directory=detray_simulation/toy_detector/n_particles_2000/ --detector-file=toy_detector_geometry.json --material-file=toy_detector_homogeneous_material.json --grid-file=toy_detector_surface_grids.json --input-event=1 --track-candidates-range=3:30 --constraint-step-size-mm=1000 --check-performance --print-performance

[krasznaa]

🤔 Over the next days I intend to put together a proposal for how I believe all of this performance measurement code should be organized. As I'm really not super happy with how it is set up at the moment.

At that point, in a couple of days, let's discuss here how your development would fit into that "new landscape". 😉

[SylvainJoube]

Thanks for your feedback Attila. Okay, I’ll wait! I’m very curious about your proposal, seeing your code is always a nice way for me to discover how things should be organized and implemented! 🦊

See you here in a couple of days then :) I’ll be watching the open PRs too.

[stephenswat]

Closing this as superseded.