Draft: T-Track vertex reconstruction

In this MR, a set of new algorithms dedicated to T-Track vertex reconstruction & selection is proposed. The aim is to complement the corresponding developments at HLT2.

Runge-Kutta improvements

• Introduced an adaptive step algorithm
• Proposed alternative make_simple_step function for extrapolation step estimation
  • loads the magnetic field vector only once per invocation, effectively reducing the number of accesses to L1/TEX

The reconstruction sequence

• Filtering of SciFi seeds with momentum cut as well as Long (and potentially Downstream) VETO.
• Seed extrapolation using the Runge-Kutta extrapolator, introduced in !1961 (merged).
  • The same approach as in Rec!4572 (merged)
  • The result is stored using a set of intermediate track states at predefined z positions.
  • A cubic-spline interpolation is used to extract a track state at an arbitrary z.
• Estimation of Point Of Closest Approach (POCA) using a ported (simplified) version of the HLT2 TrajPoca algorithm. Estimation of additional vertex parameters, such as:
  • position vector
  • parent momentum vector
  • DOCA
• Since the algorithm uses SoA containers internally, the final step is conversion from internal SoA to Allen standard track/vertex views.

Validation plots

Track extrapolation

Deviation of simplified HLT1 extrapolation from the offline implementation

Vertex reconstruction

Reconstructed vertex position (without reco cuts)

Vertex resolution

TODO

• To define physics and/or monitoring trigger lines
• Add monitoring plots/counters

References

• [1] "Physics with T-tracks using 2024 data"; General Performance Meeting; Apr. 28
• [2] "FarTreeFitter: a dedicated framework for fitting decay chains inside the magnet region"; RTA-DPA general meeting; May 2
• [3] "New T-track functors"; RTA Selections WP3 meeting; Sep. 30

FYI: @isanders @femartin