Difference between revisions of "GlueX Offline Meeting, January 9, 2013"
(→Data Challenge Status)
|Line 88:||Line 88:|
==EVE event display: Dmitry==
==EVE event display: Dmitry==
Revision as of 14:02, 22 January 2013
GlueX Offline Software Meeting
Wednesday, January 9, 2013
1:30 pm EST
JLab: CEBAF Center L207
- 1 Agenda
- 2 Communication Information
- 3 Minutes
- AmpTools upgrade
- Review of minutes from the last meeting: all
- Reconstruction sub-group reports
- PID RF Beam Bunch Selection: Paul
- Data Challenge Status
- CCDB status: Dmitry
- EVE event display: Dmitry
- New CPU architecture from Intel: David
- Action Item Review
- Review of recent repository activity: all
Talks can be deposited in the directory
/group/halld/www/halldweb1/html/talks/2013-1Q on the JLab CUE. This directory is accessible from the web at https://halldweb1.jlab.org/talks/2013-1Q/ .
- CMU: Paul Mattione, Curtis Meyer
- FSU: Nathan Sparks
- IU: Matt Shepherd
- JLab: Mark Ito (chair), Simon Taylor, Dmitry Romanov, Beni Zihlmann
Matt reviewed his recent announcement of a new version of AmpTools, now at version 0.5. Although the package is under development all the time, recently the user interface has been streamlined and should be simpler to use. However, user code must change to be compatible with this new version. Matt has changed the relevant code in the repository, so to use it one must have the new AmpTools version.
The functionality is nearly the same as before, a few new features have been added.
Matt reports that he is now collaborating more closely with the group at INFN Genoa and they have submitted various feature requests, in particular to do fits in terms of moments. Some recent work has also been in collaboration with Adam Szczepaniak.
Reconstruction sub-group reports
PID RF Beam Bunch Selection
Paul reported on his scheme for finding multiple RF beam buckets in multi-track events. Reconstructed tracks vote on their preference for the event bucket based on their times at the BCAL or TOF. See his wiki page for details.
Mark asked that at some future meeting Paul present some quantitative measures of the performance of the algorithm.
Future of reconstruction sub-group reports
We decided that although very useful in the past, we will drop the reports as standing agenda items. Progress in these areas will continue to be reported, but on an as-needed basis.
This raised the issue of our current offline task list. Now that the bulk of reconstruction factories have been written, the major issues don't fit our task list as well as they used to. Mark will propose a list of offline projects that we will track in the future. We can then discuss both composition of the list and its scope.
Dmitry pointed out that there are systems to facilitate software project management on a high-level, above that supported by the Mantis system we are using. We thought that to start out, we would use some ad-hoc system. We can get more formal later if we have to.
Data Challenge Status
The data challenge is finished now, final tally:
- CMU: 370 M events
- JLab: 1 G event
- Grid: 4 G events
The next tasks:
- Distribute the events. Richard had written a HOWTO on using grid tools to access our data. Several of us are learning how to do it.
- Archive the events. Mark will try to use the grid tools to copy the data to JLab for archiving on the tape library at JLab.
- Write a report. We need to capture the things we did and the lessons learned from the excercise.
- Check the quality. We should do some overall quality checks on the data and include them in the report.
- Produce reduced-sized data sets. Either by doing skims or writing an even more compact format or both.
- Evaluate the effect of the level-1 trigger, as configured for the data challenge.
CCDB deployment proceeds; Dmitry and Mark gave a report.
- There is a wiki page for all things CCDB now.
- The website is up and running. This can be used for browsing the structure of the database and the content of the constant sets.
- Documentation is starting to come together on the wiki page.
- A new release of sim-recon that uses the CCDB by default us being prepared.
- Three access methods are available now.
- MySQL server at JLab inside the firewall (mysql://email@example.com/ccdb)
- MySQL server at JLab available read-only, world-wide (mysql://firstname.lastname@example.org/ccdb). Note that this has the same URL as in (1) but actually resolves to a different server on a different subnet at the Lab.
- SQLite database. This is a local file that can be used, with the full functionality of the MySQL database (sqlite:///<path-to-local-file>). Since it is used from the local disk, no network connection is necessary. See the wiki page for instructions on where to find the file.
- Note that the old ascii file-based method will continue to work (file://<path-to-calib-directory>)