Inclusion of EF Vertices into Trigger Cost Code - by Steve Sekula

Software Framework Used to Develop the Procedure

Basic Information

RELEASE: 15.6.9.37

• this release contains the MC_pp_v1 menu and has the unseeded EF chain (see https://savannah.cern.ch/bugs/?func=detailitem&item_id=74380), requested by Rustem.

PACKAGES:

• I used the HEAD (as of 11/22/2010) of
  • Trigger/TrigCost/TrigCostAlgs
  • Trigger/TrigCost/TrigCostAthena
  • Trigger/TrigCost/TrigCostBase
  • Trigger/TrigCost/TrigCostData
  • Trigger/TrigCost/TrigCostRate
  • Trigger/TrigMonitoring/TrigCostMonitor

LIST:

A dump of the StoreGate contents present when running over an MC file (mc09_7TeV.105001.pythia_minbias.digit.RDO.e517_s764_s767_d307_tid139442_00/RDO.139442._000690.pool.root.1) reveals that the VxContainer holds a few vertex lists:

• HLT_ConversionVx
• HLT_PrimVx

The list I should want is HLT_PrimVx. I then attempted to load this into Trigger/TrigCost/TrigCostAthena/src/TrigNtVertexTool.cxx using just python configuration:

    vrtx_tool = costAthena.Trig__TrigNtVertexTool('TrigCostNtVrtx')
    vrtx_tool.vertexContainerName = 'HLT_PrimVx'

Code to obtain the list of vertices from the Trigger Decision Tool (TDT):

  // get the feature container
  Trig::FeatureContainer tdtFeatures = m_trigDec->features( "EF_.*" );
  Trig::FeatureContainer::combination_const_iterator
    tdtFeature(tdtFeatures.getCombinations().begin());
  Trig::FeatureContainer::combination_const_iterator
    tdtFeatureEnd(tdtFeatures.getCombinations().end());


  // get JetCollection.
  std::vector< Trig::Feature< VxContainer > > vertexTDT = tdtFeatures.get< VxContainer >();

  if ( false == vertexTDT.empty() ) {

    // retrieve the VxContainers from the TDT
    std::vector< Trig::Feature< VxContainer > >::iterator vxcIter = vertexTDT.begin();

    // grab the first PrimVx container
    if ( vxcIter != vertexTDT.end() ) {

      const VxContainer* hltvtxColl = (*vxcIter);
.
.
.

Note: according to Jiri, the above is the wrong chain from which to grab VxCOntainers. He writes:

it should be the container from the FullScan chain - i.e. the extra
trigger chain
which is in the MC menu for you. It may be named something like InDetMon_FS
cheers

COMMAND LINE:

athena -c 'PoolRDOInput=["/tmp/sekula/mc09_7TeV.105001.pythia_minbias.digit.RDO.e517_s764_s767_d307_tid139442_00/RDO.139442._000690.pool.root.1"];setGEO="ATLAS-GEO-10-00-00";setMenu="MC_pp_v1";setDebug="true";eventMax=100' ./RunTrigCost.py >& ./RunTrigCost.log

For running the vertex study:

./runRates.py . -k TrigCost -r cost_debug.root  --test="algs vtx"

None.

The Way Forward - A Summary of Rustem's Suggested Roadmap

We need vertex reconstruction for re-weighting events with pileup for
trigger rates. There are two methods to obtain vertex counts for
trigger rate ntuples.

-- Run tracking and vertex algorithms at EF within a trigger rate job.
There, we can read vertex count similarly to how it was done for L2
vertexes from the beamspot. Tomorrow I will add a necessary trigger to
MC_pp_v1 menu. We should try to produce cost ntuples later this week.
This step will require updating TrigCostAthena package to use
corresponding EF vertex collection.

-- Run offline ID reconstruction on EnhancedBias stream. This will be
done at some T1 site with enhanced bias data replicated from CERN.
This option requires that we produce Reco_trf, as described by Jamie
below.

The Reco_trt transform will be then plugged directly into a production
system. It allows us to run standard ID reconstruction and cost ntuple
extraction as a one step. We would like to have this transform working
within a month to allow use test entire procedure. Once it is working,
we will also use this transform to extract vertex counts from already
existing enhanced bias data.

A current plan by the menu group is to produce a first draft of 2E33
menu by December. At that point, they will need trigger rate estimates
with realistic pileup scenario. These rates will our input for
discussions between the trigger and physics coordination for
optimizing trigger bandwidth for the next year run.

...

You have to use TriggerMenuPython-00-13-14 or later with the MC_pp_v1 menu:

https://savannah.cern.ch/bugs/?74380

I have posted a question to this report about a release. You need the
above tag to get an unseeded EF vertexing chain. The unseeded part
means that it runs on every event.

In a month or so, we need to switch to using offline reconstruction.
We need to provide the production team with a job transform which runs
the offline ID reconstruction and cost ntuple extraction. I am
including the email below which I sent to your earlier.

-- StephenSekula - 16-Nov-2010