I have finished testing the 3D reconstruction re-write on my mac. The following lines summarize the performance on a small sample of uB MCC9 events with and without using a sliced hit collection. The overall efficiency * purity for 2D reconstruction (weighted by true kinetic energy) is 0.75 (unsliced hits) vs 0.69 (sliced hits) at a cost of 70 s/event (unsliced hits) vs 5.9 s/event (sliced hits). The processing times are for a c2 debug build which should correspond to 23 s/event and 2 s/event with a E17 prof build.
*** May 22 Fixed bug in ReconstructAllTraj Setting slc.IsValid false if StoreTraj failed
Evt 500 El 0.43 Mu 0.79 Pi 0.70 P 0.67 MuPiKP 0.75 BadEP 106 longGood 0.92 MCP cnt 2515 PFP 0.60
reco:trajcluster2:TrajCluster 0.194766 70.604 629.536 48.9219 77.1072 320
»> May 23 Repeat using sliced hits
Evt 500 El 0.40 Mu 0.71 Pi 0.69 P 0.65 MuPiKP 0.69 BadEP 138 longGood 0.90 MCP cnt 2515 PFP 0.54
reco:trajcluster2:TrajCluster 0.508129 5.85482 144.144 3.34092 10.298 320
updates for space charge effects calibration
feature/herogers_updateSCE
larreco, larevt
dunetpc, sbndcode and ubevt
Bug fixes
Updated dependencies
Change List
larsoft v08_21_00
2019-05-29 Lynn Garren : larsoft v08_21_00 for larsoft v08_21_00