[Halld-offline] Offline Software Meeting Minutes, August 20, 2014
Mark M. Ito
marki at jlab.org
Fri Aug 22 14:35:46 EDT 2014
Find the minutes below and at
GlueX Offline Meeting, August 20, 2014
* CMU: Will Levine, Paul Mattione, Curtis Meyer
* FSU: Aristeidis Tsaris
* IU: Ryan Mitchell, Matt Shepherd
* JLab: Alex Barnes, Mark Ito (chair), David Lawrence, Nathan Sparks,
Mike Staib, Beni Zihlmann
* MIT: Justin Stevens
* NU: Sean Dobbs
* UConn: Richard Jones
* David has released JANA 0.7.2. Among other features it includes
verification of MD5 checksums for resource files.
Review of Minutes from August 6
We looked at the minutes.
Mark "fixed" the broken trunk problem, induced by a change in the table
definitions for tagger counters in the CCDB, by simply removing the
"exit" in the code when an error is detected in hitTag.c. David
reported that Simon Taylor reported that this induced an effective
memory leak in the code where the unsuccessful attempt to fetch the
constants was repeated for each event, using a small amount of memory
each time. Richard thought that rather than trying to fix this, we
should wait for the new version of hitTag that actually requires the
aforementioned CCDB table change. This will come when the development
branch (discussed below) is checked onto the trunk.
Tagger Simulation Update
Richard and Paul have been working on a branch of sim-recon, namely
sim-recon-rj-pm, to add realism to the global timing scheme of the
HDDM Calls: Conversion to C++ API
All invocations of the HDDM API have been upgraded to use the C++
version rather than the old C version. This change comes in many places
in the sim-recon tree. The exception is the GEANT-3-based code, in
particular the current HDGeant; that conversion will come with the
conversion to Geant4.
HDDM Integrity Checks
The HDDM library now supports "integrity checks" where a cyclic
redundancy check (CRC) code is computed event-by-event and can be
written with the event. Downstream programs can check that incoming
data has not been corrupted.
The template for simulation output has been changed to rationalize the
arrangement of data in two areas: (1) the mix of truth and hit
information and (2) the detector component hierarchy. A generic example
of the latter is using enclosing elements to indicate "layer" rather
than encoding "layer" as an attribute of each hit. Note that changes of
this nature are indicated in the HDDM_s version in the template
contained in every data file.
As he has been making these changes, Richard has been periodically
merging changes made to the trunk onto his branch. In this way
integration testing can proceed on the branch before merge of the
branch back onto the trunk.
CCDB tagger tables have been rationalized and new C++ classes
introduced in sim-recon to accommodate the new scheme. The start
counter was used as a model for the low-level hit classes in the DAQ
plug-in. There were some modifications needed to HDGeant (GEANT3) code
to write the new classes.
Analysis Library Changes
Paul will need to change the analysis library to deal with the
ambiguity of the "correct" tagger hit. For now he will use the factory
tag for the tagger to get the true tagger hit. That makes things more
or less as things were before. Once the merge onto the trunk is done,
he will implement the necessary changes.
Paul and Will have found a problem on the branch where TDC hits from
the upstream BCAL are missing. This will have to be fixed by Richard
before the merge-back is done.
Beni asked if the new library will support named pipes connecting HDDM
producing and consuming programs, e. g. HDGeant to mcsmear. Richard
thought that this should have been working all along, but Beni has had
trouble getting it to go. We decided to take further discussion
Gxtwist is a stand-alone simulation of the tagger hall. Richard has
updated the geometry based on the latest drawings. This has been
checked in on the trunk. He has also added reality to the radiator with
separate positions for the diamond and the amorphous radiators.
There was a discussion on the email list about re-creating tracks
with mass hypotheses that are not present in the set coming from track
reconstruction. This can be very time-consuming. In these cases usually
the fit failed when using the mass assumption in question. Simon is
working on solving this problem in tracking so a full set of mass
hypotheses are available in the REST file. See the email discussion for
Commissioning Geometry Simulation
An effort has started by Sean and Simon to simulate the detector with
the commissioning geometry. A branch of HDDS has been made for this
work and Simon has started putting in the commissioning target.
Analysis of Six Final States Using Data Challenge 2 MC
Ryan has looked at the following reactions generated in Data
1. \gamma p \rightarrow \pi^+\pi^- p
2. \gamma p \rightarrow \pi^+\pi^-\pi^0 p
3. \gamma p \rightarrow K^+K^- p
4. \gamma p \rightarrow \pi^+\pi^-\pi^+\pi^- p
5. \gamma p \rightarrow \pi^+\pi^-\pi^+\pi^-\pi^0 p
6. \gamma p \rightarrow \pi^+\pi^-\pi^+\pi^-\pi^+\pi^- p
The data set is comprised of 10 M events at 1×10^7 photon rate, about
1% of the events generated for DC2.
He is using analysis code he developed for use at CLEO and BES. (As a
check, Matt has been using Paul's analysis library in a parallel
analysis to compare and contrast results. It was in this context that
he discovered the track re-creation time sink mentioned above.)
Please see his slides for the plots.
The mass distributions for the multi-pion systems in many reactions
peaks much higher than people expected.
Data Challenge 3
Mark has succeeded in creating EVIO format simulation data. Sean told
us that reconstruction of this data should not be a problem.
* Fix BCAL bug on tagger reconstruction branch and merge onto trunk.
(Richard and Paul)
* Ask Simon about track re-creation. (Mark)
* Develop the simulation for the commissioning geometry. (Sean,
* Will will send problematic BCAL timing events to Richard. (Will)
Mark M. Ito, Jefferson Lab, marki at jlab.org, (757)269-5295
More information about the Halld-offline