Disabled L,S corrections in CaloFutureShowerOverlap in baseline HLT2 based on the studies of gamma and pi0 resolution (documented below). Performance of CaloFutureShowerOverlap is improved by about 60%. Reference output of several Moore tests needs to be updated ( RecoConf.hlt2_reco_baseline_multi_threaded, RecoConf.hlt2_reco_brunelesque, RecoConf.hlt2_reco_brunelesque_with_data_monitoring, RecoConf.hlt2_reco_brunelesque_with_mcchecking, RecoConf.hlt2_reco_calo_from_dst )
Added tests RecoConf.hlt2_reco_calo_resolution_gamma, RecoConf.hlt2_plot_calo_resolution_gamma to produce plots with x,y and E cluster resolution (calculated by matching to MC gammas or pi0s from B -> K* gamma, B -> pi+ pi- pi0 MC samples. To run, one needs Rec!1880 (merged), Lbcom!431 (merged), LHCb!2380 (merged) and lhcb-datapkg/PRConfig!113 (merged) (also Brunel!971 (merged) to pass Brunel tests). The output of CellularAutomaton (raw clusters), CaloFutureShowerOverlap and faster CaloFutureShowerOverlap (with the stopping criteria, or without L,S corrections) is benchmarked.
Also added algorithms RecoConf.hlt2_reco_calo_resolution_pi0, RecoConf.hlt2_plot_calo_resolution_pi0 to do similar benchmark for pi0 reconstruction.
To produce resolution plots:
make fast/Moore/test ARGS="-V -R _calo_resolution_"
See output of hlt2_plot_calo_resolution_gamma, hlt2_plot_calo_resolution_pi0 for produced plots (plots are produced for several values of matchFraction and also as a function of it (see below in the discussions for the description of what is matchFraction), it makes sense to look at plots with matchFraction ~ 0.9, e.g. Res_summary_matchFraction0.9.png).
By default 100 events are processed, while much more is needed for good plots (~10K, will run a couple of hours).
Rosen Matev authored
