Core Minutes 8/10/2004
Present: Anders Borgland, Toby Burnett, Johann Cohen-Tanugi,
Seth Digel, Richard Dubois, Zachary Fewtrell, Dan Flath, Warren Focke, Riccardo
Giannitrapani,
Berrie Giebels, Navid Golpayegani, Heather
Kelly, Matt Langston, Pat Nolan, Chuck
Patterson, James Peachey, Leon Rochester, Alex Schlessinger, Tracy Usher
-
Announcements: (Richard)
- LAT Face-to-Face Mangers Meeting is set for Wednesday
- GLAST Users' Committee Meeting at GSFC Mon/Tues Aug 9 and 10th
- Heather is at SLAC for a visit Mon/Tues
- GEM and merit ntuple: (Richard)
- The Issue: Flight Software has its view of the trigger, stored
in the GEM data. Gleam uses TriggerAlg to emulate what the trigger
is doing and stores that in the merit ntuple. SVAC plans to
compare the GEM data against the results of TriggerAlg as a sanity check
of the hardware and simulation. Do we store GEM data in the merit
ntuple? How do we make it crystal clear to users that the output
from TriggerAlg is from the digis and that the GEM data reflects the
hardware trigger? What is the purpose of merit?
- Toby: Analysis involving the GEM is a hardware check and belongs
in a separate analysis. It is very easy to use ROOT's Tree Friends
to make multiple Trees appears to be one.
- Side-note: the capability to store ints, floats, and simple
arrays has now been added to ntupleWriterSvc for use in the merit
ntuple. Currently the merit ntuple contains only doubles -
this addition provides us with great flexibility.
AnalysisNtuple would have to be modified to start using this new
feature. First we need to talk to Bill to see that his IM tool
is modified to determine the type of the columns in the ntuple.
- Richard asked Anders how the SVAC ntuple and merit ntuple will
be used by I&T. SVAC ntuple will contain additional low-level
data. The merit ntuple will be used primarily, but the SVAC ntuple
is available for special studies. SVAC ntuple is generated using the
digi ROOT file. Richard then requested that I&T check that the
SVAC ntuple has one-to-one correspondence with the merit ntuple and can
use Tree Friends - to allow users to easily analyze the data.
- Anders and Toby questioned each other concerning the need for GEM data
in merit. Anders reminded us that we are only talking about 14
additional variables and feels that we should have a record of the
hardware trigger in the merit output. Toby wants to ensure that we
can perform the same analysis using merit whether the data comes from a
simulated or non-simulated source. Anders also stated that the
GltWord (the output from TriggerAlg) represents the results Or'd over
all towers - there is no record of what happened in individual
towers. There is a related quantity that reports which Tower
triggered, which only reports the lowest numbered tower - this is fine
until we start simulating/analyzing data involving multiple towers.
- External Library Naming Convention and
Organization: (Toby)
- Navid has setup a well thought out naming convention and directory
organization for the external libraries that we should all be aware of
and start using.
- Heather asked what will happen to the old external libraries and Navid
answered that they will be moved and re-named according to the new
naming convention.
- Richard asked Navid to be prepared to discuss this next week at the
core meeting and write up a slide or two detailing what he has done and
what the new rules for external libraries are.
- Building GlastRelease HEAD:
- Toby reported that there is new output from Gleam that needs to be
removed.
- GR HEAD does not build on windows due to the new RootPolicy.
Toby propose to create a new RootCintPolicy and then replace RootPolicy
to deal solely with providing access to the ROOT external library.
Toby then went on to discuss the need to remove circular dependencies
and re-iterated his desire to move tests to separate packages and start
using containers. Richard asked if the ReleaseManager can handle
containers or do we have to inform Alex about each new container as they
are created? (Alex had VRVS problems) Toby answered that
yes, as he understands it, RM needs to maintain a list of container
packages so that the checkout occurs properly.
- Toby asked whether the CAL packages are ready to be moved from LATEST
to HEAD. Berrie answered yes and will promote them using the tag
collector.
- Report from SSC: (James Peachey)
- Jim Chiang is at GSFC attending the Glast Users' Committee Meeting
(Aug 9th and 10th)
- This is Yasushi's last week at GSFC. Jim C. and James P. are spending
time with Yasushi doing a brain dump.
- Last week marked the end of Build One, the SSC schedule for internal
software development.
- Very successful - all software was delivered as scheduled, which
included:
- rspgen (U14 tool) which was already working for some time but
new features have been added
- Delivered a new Pulsar Tool named glbary
- st_graph a new graphing package that uses ROOT
- Build Two is now underway which will involve work on the A3 tool
for periodic searches and work on visualization and plotting tools.
- Toby asked if there was a new ScienceTools release coming. James
responded that he will talk to Jim C. today
- James P went on to give some details about the Glast User's Committee
Meeting.
- Monday morning was spent eliminating confusion concerning the new
organization chart for the SSC.
- Monday afternoon was spent on talks concerning the science behind
the Science Tools. Jim Chiang gave a talk on Likelihood that
impressed the crowd. Data Rights have been discussed.
James suggested Seth could provide more details. Richard asked
that Seth provide more information at the ground software meeting on
Thursday.
- PeoplePower: (Richard)
- http://www-glast.slac.stanford.edu/software/core/GLAST-offline-functionChart.pdf
This was originally presented at the Science Operations Center
Review. It represents what the Offline group does. Note that
Release Manager appears three times. Under Infrasture it refers to
the tool, the Release Manager. Under ScienceTools and LAT Sim/Recon
it refers to the person coordinating the package.
- Green boxes represent areas that we have pretty well covered.
- Shocking Purple boxes are areas we need some help
- Yellow boxes are areas we believe SCS could provide aid.
- Richard mentioned that they have an interest in an all-memory
machine that would have TBs of memory that could gooble up all data
at once so data processing would not have to deal with disk latency
issues.
- We then went over the purple boxes
- Richard detailed our need for someone to oversee the
Infrastructure. He has asked the lab to hire someone to be at
SLAC who can work on this.
- Could use more ROOT expertise. Babar has asked the lab to
mimic what Fermi lab has done and have a ROOT team member reside
locally.
- Sim & Recon both need more oversight. Find people with a
system-wide view of the simulation and reconstruction. We
would probably fill this need from within the collaboration.
Leon went on to ask for more details. Richard explained that
we need someone to coordinate the overall performance.
Currently we have a loose confederation of the subsystems.
- Toby then asked for clarification of the new FTE for
Infrastructure. Richard elaborated stating that this person
would answer to and be paid by SCS, but would functionally be
dedicated to GLAST.
- Flight I&T Support Report to Face-to-Face
Managers Meeting:
- Richard went over his slides to be presented tomorrow (Aug 11th):
http://www-glast.slac.stanford.edu/software/core/SAS-Deliveries-to-INT-200408.pdf
- Pipeline will be used by Online to do FITS conversion of the real
data, among other things. Dan has given SVAC direction on how to
wrap their scripts (SVAC is now ready). We had a false start about
a month ago, where it was found that the directory structure used by the
pipeline and in use by Online was orthogonal - now fixed. Dan
reported last night that Navid has provided a batch interface.
I&T could be in a position to test the pipeline this afternoon or
evening. Matt has been working on web displays for the pipeline
view (looking at running processes) and the datasets. The first
version of the web displays are due today. The pipeline is due
Friday and we believe we are on track.
- CAL will make its delivery after Labor Day.
- TKR Hot/dead strip handling is ready. Splits are well in
hand and will be ready for an Aug 13th release. Hiro's TKR
Alignment has been reorganized as Gaudi Algorithms and Leon is preparing
to ingest the alignment constants into reconstruction.
- Geomety: Eduardo has requested a full review.
- CAL (Mark) has completed the CAL geometry review and written up a
report that points out the areas that need modification in our CAl
geometry model. A number of small corrections, most of which
should produce no noticeable effect on the simulation. Richard
suggests that in the short-term, we correct the easy items such as
the CsI density.
- TKR Leon is confident of the TKR geometry's integrity. It
has been difficult to schedule time with the engineers. It is
hoped that will happen soon, but it not felt that this is a pressing
issue.
- LDF support: GEM is on the way. Heather has two requests
for I&T:
- A new test file from Online containing all of the event
contributions: GEM, TEM, Trigger Primitives, AEM.
- Continued support for the C++ version of LDFdump.cpp in Online's
LDF library.
- EngineeringModel is being prepared for release in mid-August.. we may
suffer from feature-creep in the coming weeks, as new updates to the
packages are made to satisfy needs for I&T.
- No time to go over JIRA issues - next week
H. Kelly Last Modified:
2010-06-01 15:48:38 -0700