GlueX Offline Meeting, December 1, 2009
- 1 Location
- 2 Telephone
- 3 Video Conferencing
- 4 Slides
- 5 Agenda
- 6 Minutes
The meeting will be at 2:00 pm in CEBAF Center F326/7
To connect by telephone:
- 800-377-8846 : US
- 888-276-7715 : Canada
- 302-709-8424 : International
- enter participant code: 77438230# (remember the "#")
Talks can be deposited in the directory
/group/halld/www/halldweb/html/talks/2009-11 on the JLab CUE (you have to be a member of the "halld" Unix group). This directory is accessible from the web at https://halldweb.jlab.org/talks/2009-12/ .
- Review minutes from last meeting: all
- Nightly builds, now building without errors
- Three platforms: RHEL5 (32-bit only), Fedora 8 32-bit, CentOS 5 64-bit
- non-debug and debug versions
- Other platforms? (need ssh access, scratch disk mounting a plus)
- Auto-notification of errors? Warnings?
- New tagged release: release-2009-12-01
- Three platforms
- release notes, set-up scripts to come
- New Subversion server
- New virtual machine
- Newer version of subversion: 1.4 vs. 1.1 on the old server
- Will allow locking of tagged releases: no new check-in's allowed
- Cut-over scheduled for two weeks from today
- Nightly builds, now building without errors
- Presentation from Simon role of DParticle_factory
- Presentation from Matt, Blake, or Mihajlo on photon reconstruction chain
- Discussion and decision on how to unify high-level objects
- Offline Computing Project Management: Mark
- Coding Conventions vote: all
- Action item review: all
IU: Mihajlo Kornicer, Matt Shepherd
JLab: Mark Ito (chair), David Lawrence, Sascha Somov, Simon Taylor, Elliott Wolin
Review of minutes from last meeting
We reviewed the minutes from the November 17th meeting. No significant remarks were generated.
- Nightly builds: The builds are now going on now building without errors, on three platforms: RHEL5 (32-bit only), Fedora 8 32-bit, and CentOS 5 64-bit. Both debug and non-debug versions are created.
- We discussed whether we want to do the builds on other platforms, especially MacOS. For that we would need hardware. No clear path forward presented itself, but we will think about it.
- We also discussed whether we should have automatic notification of errors and warnings. One idea is to have a dedicated email list so people could subscribe the error message stream. Again, on this there was not a consensus; we will raise it again later.
- David noted that a couple of feature that used to be present in the nightly builds have been lost: doxygen documentation tree generation and svn statistics. Mark will look into restoring them.
- New tagged release: release-2009-12-01 This has been built on the three above-mentioned platforms. Release notes and set-up scripts are to come.
- New Subversion server The CNI group is building us a new subversion server. It will be a new virtual machine running subversion 1.4 vs. 1.1 on the old server. The new version will allow locking of tagged releases: i. e., no new check-in's will be allowed for the locked files. Cut-over is scheduled for two weeks from today.
DPhoton, DParticle and how to unify high-level objects
Simon outlined the philosophy behind his recent work on the DParticle factory.
- BCAL clusters, FCAL clusters and charged tracks are all represented as particles
- track matching is done to the outer detectors
- beta is calculated from time of flight
- particle ID is done; a mass is assigned
- time-based tracking results are used
- the DParticle name used to be used by the object now known as DTrackTimeBased
Sascha thought there should be a way to go back to the original low-level objects. David mentioned that this could be accomplished via associated objects.
Matt then led us through the structure of the DPhoton object as illustrated on his chart.
- It inherits from DKinematicData
- DPhoton calculates the error matrix for its kinematic quantities in standard global coordinates.
- It would benefit from a event vertex factory.
- At present the center of the target is assumed and the resultant uncertainty is folded into the error matrix.
- Matching with charged tracks is done to identify neutral clusters (photons).
Clearly there is some redundancy in the current services provided by DParticle and DPhoton. General discussion:
- Sasha thought that any particle ID likelihoods should be available.
- Related to that Mark pointed out that different analyses will want to cut on particle ID criteria differently, depending on the requirements of each analysis on purity. One size will not fit all.
- Matt remarked that on CLEO fits to pi, K, and proton hypotheses were always done during reconstruction and the results of all three saved.
- Matt also mentioned that both "entrance fits" and "exit fits" were done, the former doing the Kalman algorithm from the outside in to get good resolution on momentum at the event vertex, the latter going from the inside out to get good resolution on momentum at the RICH detector.
- It was pointed that the DPhoton should now use DTrackTimeBased for charged-particle vetoing.
- There was some question about whether wire-based was adequate for neutral cluster identification. Most felt that is was, but this should be studied.
David and Simon agreed to present a plan for modifying the current class structure taking into account the discussion we had and present it at the next meeting.
Finally Matt showed us the documentation he wrote on HOWTO extract photons or pi0's from the framework.
Offline Computing Project Management
Mark showed us the recent work he has done on revising the Offline Computing component of the official 12 GeV contruction plan. He has created a wiki page where the information is posted. He showed:
- a Gantt chart
- the list of tasks and their attributes including work required and cost
- a cost comparison between the old plan and the new
- a manpower profile for both Lab staff and contributed labor
The timeline is still a bit arbitrary, but tools have been developed to produce these reports automatically if modifications are made in the future.
Coding Conventions vote
We dicussed the Coding Conventions a bit before voting. David mentioned that there were still some modifications based on discussion at the last meeting that need to be incorportated. Mark express that the requirement that code be thread-safe was a bit vague and might be hard to support. David and Elliott said a specific definition was hard to state concisely; that in this case the proof of the pudding is in the eating [not their exact words].
We voted. The conventions were adopted unanimously.
Action Item Review
We skipped this item. We will review the list next time.
New Action Items
- Explore options for a MacOS build.
- Implement a system for auto-notification of errors from nightly build. -> Mark
- Look into restoring doxygen generation and subversion statistics generation to the nightly build. -> Mark
- Use DTrackTimeBased in DPhoton for neutral cluster identification. -> Matt
- Study whether wire-based tracking is adequate for neutral identification.
- Propose a new high-level class structure for charged particles and photons. -> David, Simon