Introducing TC-LVT vertices into DAOD_PHYS
This merge request includes changes to the code in Reconstruction/VKalVrt/SoftBVrtClusterTool to streamline and adapt the TC-LVT vertices (please see ATLAS-CONF-2019-027 for details - three working points in total) to be written to DAOD_PHYS. It includes also the changes to DerivationFrameworkExamples/PHYS.py to have the overtaxing code run and the output persistified.
Disk Usage: the code has been tried running on 10000 events on ttbar semileptonic DSID 410470. These are the results on the output file size when running with and without the additional TC-LVT vertices:
No vertices --> -rw-r--r-- 1 iv41 iv41_g 412741331 Dec 17 01:50 DAOD_PHYS.18778365._000001.pool.root.1 With vertices --> -rw-r--r-- 1 iv41 iv41_g 415299962 Dec 17 19:12 DAOD_PHYS.18778365._000001.pool.root.1
The output of checkXAOD on the latter shows: 3254.129 kb 595.932 kb 0.060 kb 5.461 10000 SoftBVrtClusterTool_Tight_Vertices (DataVectorxAOD::Vertex_v1) 3550.590 kb 808.430 kb 0.081 kb 4.392 10000 SoftBVrtClusterTool_Medium_Vertices (DataVectorxAOD::Vertex_v1) 3885.955 kb 1018.602 kb 0.102 kb 3.815 10000 SoftBVrtClusterTool_Loose_Vertices (DataVectorxAOD::Vertex_v1)
for the three working points. Overall they introduce an XAOD size change below 1%.
CPU-wise, this is the output at the end of the log file for the execute method of the Loose working point (the slowest of the three):
AODtoDAOD 19:12:50 SoftBVrtClusterToolLoose:Execute INFO Time User : Tot= 97.4 [s] Ave/Min/Max=0.00974(+-0.0137)/ 0/0.309 [s] #=10000
so, a tenth of a second/event on average.
This is the perfmon output:
AODtoDAOD 19:12:50 PMonSD n cpu max@evt vmem max@evt malloc max@evt component
AODtoDAOD 19:12:50 PMonSD [evt] 9999 10 310@660 0 0@0 63 230@417 SoftBVrtClusterToolLoose
AODtoDAOD 19:12:50 PMonSD [evt] 9999 9 230@660 0 0@0 44 220@3369 SoftBVrtClusterToolMedium
AODtoDAOD 19:12:50 PMonSD [evt] 9999 8 160@660 0 0@0 24 221@8735 SoftBVrtClusterToolTight
Log file of the run with vertices on 10000 top events attached.
Please let me know if further information is needed. log_withvertex.log