Reconstruction tutorial: Part 2

1
Reconstruction tutorial Part 2

• General introduction on reconstruction
• Web/documentation tour
• Running Reconstruction with no user code
• Customize jobOptions
• Browsing the ntuple

2
A brief history of Atlas Reconstruction

• Some algorithms development started more than 10
  years ago!
• Pursued through the various detector Technical
  Design Reports till  Physics TDR  in 1999
  (still the most relevant reference document)
• This was  atrecon , mostly fortran code in
   slug , a zebra based framework
• Then migration to C then to Athena
• validation of Athena Reconstruction this year
  plus on-going development (Detector Description,
  Event Data Model, new algorithms)
• Plenty of things to do!

3
Some acronyms

• All atlas offline code is stored in CVS, which
  manage the evolution of the source files
• The build of the binaries (compilation option,
  include files, libraries) as well as the
  run-time environment are managed by CMT
• A new version of the code is entirely rebuilt
  approximately every 3 weeks ( developer
  release  e.g 5.2.0) with a  major release 
  approximately every 6 months.

4
Brief tour of the web/ documentation

• Mailing lists (sw-help, sw-reconstruction,
  sw-developers)
• Reconstruction web page
• Software Development web page
• Howtos
• viewcvs/lxr
• release plan
• release status

5

Moore flow
TES

• Each step is driven by an Athena top-algorithm
• Transient objects are passed via TDS/StoreGate
• Independent algorithms, the only coupling is
  through the transient objects

MooAlgs
RPC/TGC digits
MooMakePhiSegments
PhiSegments
MooMakeRZSegments
MDT digits
MooMakeRoads
CrudeRZSegments
MooRoads
MooMakeiPatTracks
MooiPatTracks
MooStatistics
MooMakeNtuples
Ntuples
G.Cataldi
6
Typical reconstruction dataflow
Tracker digits
Tracker digits
Storegate Transient Event Data Store
Tracking
Tracks
Tracks
Calorimeter Digits
Calorimeter Clustering
Calorimeter clusters
Clusters
Tracks, Clusters
Electron/photon Identification
Apparent dataflow
Electron/photon
Real dataflow
Electrons/photons
7
Available algorithms (5.2.0)

• (Some of these correspond to several Athena
  Algorithms)
• Truth interface
• xKalman (tracking)
• iPatrec (tracking)
• LArg Reconstruction
• Tile Reconstruction
• Muonbox (Muon reconstruction)
• Moore (Muon reconstruction)
• Jet
• E/gamma identification
• Tau identification
• missing ET
• Vertexing (primary vertex, interface of secondary
  vertices)
• Conversion
• Energy flow (new)
• Coming soon the possibility to run Atlfast (fast
  simulation) in parallel to reconstruction

8
RecExCommon

• All the available reconstruction algorithms are
  run from package RecExCommon, which is code free
• (RecExCommon is systematically run to check a
  release is (not) working)
• (almost) each Reconstruction algorithm fills a
  block of the combined ntuple (CBNT) with all the
  variables necessary at analysis level
• This allow easy checking of basic quantities
• However, the recommended model is not to write
  long kumacs to analyse the ntuples, but rather to
  do the analysis (i.e look for Z candidates and
  fill ntuple with Z variables) in the Athena
• We do not have persistency (besides ntuple) so
  reconstruction need to be rerun anytime the
  analysis is changed
• Description of variables (to be updated)
  Computing ?Organisation ?CSG domain ?
  Reconstruction?Reconstruction in Athena ?
  CBNT_Athena

9
Documentation

• Poor right now, but we try to improve it
• Where is the list of reco algorithms available ?
• RecExCommon/share/RecExCommon_jobOptions is most
  likely to be up to date
• Web page Reconstruction?Reconstruction in Athena
  (being updated)
• Knowing the algorithm name how can i have
  information on the algorithm, and on its output
• Same web page
• Documentation hub (Software Development?TagCollect
  or then Package menu) exist but barely filled,
  on-going
• Find the code Software Development?LXR (very
  useful to find where is a class/method defined)
• Since release 4.6.0, Doxygen generated class
  diagram in ltpackagegt/doc/Doxygen/html/
• http//atlas.web.cern.ch/Atlas/GROUPS/SOFTWARE/OO/
  dist/5.2.0/Reconstruction/egammaEvent/egammaEvent-
  00-01-01/doc/Doxygen/html/
• Also reachable through tagcollector
• CBNT_XXX algorithms to fill combined ntuple are
  usually good examples to see how to use object
  XXX

10
Use of CBNT ntuple

• Ntuple ntuple.hbook is created in run area
  Reconstruction/RecExample/RecExCommon//run
• Look at the truth
• paw
• Pawgt his/file 20 ntuple.hbook
• Pawgt cd cbnt
• Pawgt scan 3333 ! ! ! ! type ptgen etagen kmothnt
• (type is PDG code 11ele, 13muon, 23Z, 25H
• Kmothnt the index of the mother in the ntuple)
• Can you guess what kind of events you are looking
  that ?
• Plot reconstructed track pT
• Pawgt nt/plot 3333.1/ptinvvert

11
Structure of jobOption file

• Three parts in RecExCommon_jobOptions.txt
• Setting up athena zebra event reading
• User control number of events, printout level,
  ntuple files
• Algorithms list (system reconstruction, combined
  reconstruction additional services)
• In general include joboption fragments
• include "VERTEXPRIMARYROOT/share/Vertex_jobOptio
  ns.txt"
• CBNT_Athena.Members "CBNT_Vertex"
• VERTEXPRIMARYROOT is an environment variable set
  up by cmt to point to the VertexPrimary package
  (in release or in user area)
• The jobOption fragment is part of the Algorithm,
  should be maintained as carefully as the C code
  

12
Structure of the log file

• A sample log file is available in
  RecExCommon/Testing/Reference_RecExCommonTesting.l
  og
• dump of all nested jobOptions files
• Dynamically load of libraries
• Starting Athena services
• Read data file header and first event
• Algorithms initialisation
• Algorithms execution
• readexecute for each event
• Algorithms finalize
• Dump of CPU time per algorithm
• Dump of ntuple structure
• If anything goes wrong look for first occurrence
  of ERROR or WARNING (unfortunately some are
   normal )

13
Tuning the jobOption (1)

• Switch off/on algorithms comment out the
  corresponding jobOption fragment (it turns out
  muon reconstruction is not functional in release
  5.2.0 we are using)
• // AMDB database is needed by muonbox
• ////include "AMDBSIMRECATHENAROOT/share/Amdbsimr
  ecAthena_jobOptions.txt"
• //
• ////include "MUONBOXATHENAROOT/share/Muonbox_job
  Options.txt"
• //new f90 version of muonbox (at cern only so
  far)
• //include "MUONBOYATHENAROOT/share/Muonboy_jobOp
  tions.txt"
• //
• ////CBNT_Athena.Members "CBNT_Muon"
• //
• ////include "MOOREROOT/share/Moore_jobOptions.tx
  t"
• ////CBNT_Athena.Members "MooiPatNtuple"

14
Tuning the jobOptions (2)

• Change the printout level in RecExCommon_jobOption
  s.txt
• // Set output level threshold
• //(0NIL 1VERBOSE, 2DEBUG, 3INFO, 4WARNING,
  5ERROR, 6FATAL )
• MessageSvc.OutputLevel 2 //for
  everything (but do not do it now)
• egamma.OutputLevel 2 // just for algorithm
  egamma
• Use NameAuditor (print a message about entering
  and leaving each algorithm)
• AuditorSvc.Auditors "NameAuditor"
• Use StoreGate dump to see the list of all
  collections in each event
• StoreGateSvc.Dump true
• Try now and check you have indeed switched of
  muon reconstruction, requested debug printout to
  CaloRecsw, and switched on the name auditor
• Note that to save time you can switch off the
  scrolling of the log file on the screen

15
Tuning the jobOption (3)

• Use root tree rather than paw ntuple (root users
  raise your hand)
• // Hbook Ntuple output
• ApplicationMgr.DLLs "HbookCnv"
• //// NTupleSvc.Output "FILE1
  DATAFILE'ntuple.hbook' OPT'NEW'"
• //// ApplicationMgr.HistogramPersistency"HBOOK"
• ////HistogramPersistencySvc.NPAWC 800000
• // Root Ntuple output
• ApplicationMgr.DLLs "RootHistCnv"
• ApplicationMgr.HistogramPersistency "ROOT"
• NTupleSvc.Output "FILE1 DATAFILE'ntuple.root'
  OPT'NEW'"
• root
• rootgt TBrowser()
• (you are free to pursue the tutorial with either
  hbook or root ntuples pick the one you are most
  familiar with)
• For the tutorial an alias setroot is available to
  set up root
• RecExCommon/share/dump.C is a very simple example
  to use of root ntuple)

16
Tuning the jobOptions (4)

• Change algorithm properties
• I.e in share/CBNT_Truth_jobOptions.txt
• // add a selector if nothing specified, tak
  everything within
• // PtMin and EtaMax
• CBNT_Truth.Members "CBNT_TruthSelector/TakeAl
  l"
• TakeAll.PtMin-1. // minimum pt (if negative,
  no threshold)
• Note
• TakeAll is another instance of sub-algorithm
  CBNT_TruthSelector
• These two lines can also be added in
  RecExCommon_jobOptions.txt after
• include  CBNT_TRUTHROOT/share/CBNT_Truth_jobOpt
  ions.txt 
• In general the order in the jobOptions does not
  matter except
• If conflicting settings the last one wins
• always appends to the end (so, in particular,
  the order of algorithms/subalgorithms is the
  order they appear in the jobOptions)
• (it is possible to set the property of an
  algorithm before creating it)
• (there will be a fatal error if you try to set a
  property which is unknown to the algorithm)
• (there will be no error if you set a property to
  a non existing algorithm)
• Please try now
• check that the new instance is run grep TakeAll
  AskAthena.log check that you have more particles
  in the ntuple