Skip to content

update beamline PV finding to reduce fake PV rate

Florian Reiss requested to merge play_with_weights into master

changes:

  • add clean up step after PV reconstruction to remove duplicate PVs
  • change weight calculation in multi fitter
  • the CPU and GPU beamline fitting diverged at some point; bring them back together
  • clean up constants used in beamline PV
  • fix number of tracks of reconstructed PV -> is now the sum of weights

performance minbias:

Checking GPU beamline PVs 
 ============================================
 Efficiencies for reconstructible MC vertices: 
 ============================================
 
 MC PV is reconstructible if at least 4  tracks are reconstructed
 MC PV is isolated if dz to closest reconstructible MC PV >  10.0000 mm
 REC and MC vertices matched:  by dz distance
 
All                  :  0.8869( 4164 / 4695 )
Isolated             :  0.9209( 2364 / 2567 )
Close                :  0.8459( 1800 / 2128 )
False rate           :  0.0221( 94 / 4258 )
Real false rate      :  0.0221( 94 / 4258 )

performance b2phiphi:

Checking GPU beamline PVs 
 ============================================
 Efficiencies for reconstructible MC vertices: 
 ============================================
 
 MC PV is reconstructible if at least 4  tracks are reconstructed
 MC PV is isolated if dz to closest reconstructible MC PV >  10.0000 mm
 REC and MC vertices matched:  by dz distance
 
All                  :  0.8945( 4190 / 4684 )
Isolated             :  0.9263( 2325 / 2510 )
Close                :  0.8579( 1865 / 2174 )
False rate           :  0.0372( 162 / 4352 )
Real false rate      :  0.0372( 162 / 4352 )

to do:

  • check impact on throughput

Closes #58 (closed), #60 (closed) (@freiss correct me if I'm wrong about issue 60)

Edited by Dorothea Vom Bruch

Merge request reports