Track by Track Comparison

Data samples:

• StiCA vs Sti, Stv vs Sti, Stv vs StiCA.
  1. Y2004_AuAu200
  2. Y2009_pp500
  3. Y2010_AuAu200 and p_T for different sectors.
  1. CPU usage:
     • Debug Y2004_AuAu200, Y2009_pp500, Y2010_AuAu200: Sti is a little slower than StiCA because it has more outliers (some outlier has been removed my no. of hit per sector cut). Stv is slower by a factor of ~2.
     • Non Debug: Y2009_pp500: Sti is slower by a factor ~1.5 than StiCA. Stv is slower than StiCA by a factor of ~3.
• Sti vs StiCA, Stv vs StiCA, StvCA vs StiCA
  1. Y2010_AuAu200 with StvCA
  1. CPU usage: Non Debug: Y2010_AuAu200 with StvCA: Stv has been run in Debug mode, all other in NoDebug. From plots we can see that it was used 3 different type of CPU.

Summary

• Sti vs StiCA: The results have confirmed the conclusions made by Maksym Zyzak at al. talk on August 25, 2010 @ S&C meeting and in October published in CHEP 2010 proceedings:
  1. StiCA is a factor 1.5 faster than Sti for events with high TPC occupancy,
  2. In average StiCA tracks have ~1 additional hit with respect to Sti.
  3. Efficiencies for Global tracks for StiCA ~7% higher that for Sti. These efficiencies have less dependence on ref. multiplicity.
  4. Efficiencies for Primary tracks for StiCA ~1÷2% higher that for Sti. These efficiencies have less dependence on ref. multiplicity.
  5. The reconstructed global and primary track kinematic al parameters are practically the same for Sti and StiCA.
• Stv/StvCA vs StiCA:
  1. Stv is a factor 1.5 ÷ 3 slower than StiCA depending on TPC occupancy.
  2. In average StvCA tracks have ~1 additional hit with respect to StiCA.
  3. Stv efficiency is higher than StiCA for global (~2%) and primary (~1%) tracks. (This observation contradict to comparison based on MC.)
  4. Stv global and primary tracks have momentum higher than Sti ones. The relative difference increases with p_T.