DØ Level 3
Algorithms Meeting: 28th August 2002 in the Farside
Talk
-
Status of timing/profiling study: Han Do.
Results were presented
from the profiler of time usage per L3 tool based on 1000
events from a recent data run.
The next step will be to make a more detailed breakdown of where the
time is being used for the 10 most time consuming tools.
Discussion Items
-
"Effort reports"
Harry Melanson is putting together a summary of the current and
projected effort for the D0 algorithms group.
In order to provide input from L3 we'd like to bring our L3 activities
file:
http://www-d0.fnal.gov/computing/algorithms/level3/status.html
back up to date. Please take a look at that file and let us know
if the activities and %FTE numbers that appear under your name are
correct. If you are working on L3 related stuff that we have missed out,
please let Dan and Terry know.
-
There are several reasons why we maintain a list of available L3 projects:
-
Provides specific suggestions for how newcomers to L3 can get involved.
-
Helps us ensure that important jobs don't get forgotten about.
-
Helps document the need for extra manpower for the L3 algorithms group.
Please take a look at the file
http://www-d0.fnal.gov/computing/algorithms/level3/projects.html.
If there jobs that should be on the list that have been left out,
please let Dan and Terry know.
-
A couple of the urgent items on the available L3 projects list
are specifically relevant to jets:
-
Look at data collected with global_CMT-8.xx to study effectiveness of
the primary vertex corrected jet E_T.
-
Perform studies to evaluate the effects of calorimeter non-linearity
corrections and hot cell killing on jets.
We (L3 algorithms group) have less direct contact with the jets group,
e.g.,
through shared personnel, than with most other ID groups.
Action item: Terry Wyatt and/or Volker Buescher
to talk to Greg Davies about getting
these studies performed.
-
p12 readiness for online:
-
Per Jonsson reported that the jet tool appears to have a memory leak.
Jobs run out of memory after ~15K events unless jet tool switched off.
This is surprising since nobody could remember any changes to the jet tool
between p11 and p12.
Action item: Volker will investigate.
-
Jon Hays and Gustaaf Brooijmans have encountered a number of problems
in trying to run p12 online in tests.
-
Crash in monitoring: Moacyr Souza investigating
-
Problems in geometry manager.
Conclusion: p12 online is probably 2-3 weeks away. (This is
anyway consistent with the desire to run stably for a few weeks with
the current recent global trigger
list global_CMT-8.20, which has only recently been run online.)
-
L3 trigsim bugs/features that need to be addressed for p13: Dugan O'Neil
-
rdc run config manager (needed for L1CTT) is incompatible with L3,
which has its own run config manager.
-
We need to eliminate as many as possible of the trigger list
parameters to do with picking up the correct geometry and calibration
files.
The event data contain enough information to enable the correct files
to be selected.
This is one of the main causes of problems when running trigsim.
Volker has implemented the necessary changes for the calorimeter
unpacker for p13.
Action item: In the short term
Andre Turcot will provide cookbook-level instructions of a similar
nature to those provided by Daniel Whiteson for the L3 tracking.
-
As part of the trigsim verification project, Josh Dyer has implemented
the bug reporting mechanism "bugzilla".
The relevant tool authors will be notified automatically by email when
a bug is reported.
-
It was agreed by all that for the L3 part of the trigsim verification
project, Josh Dyer should start with the same set of histograms that have been
defined by Elliot Cheu for the L3 trigger examine.
Action item:
Elliot will get in contact with Josh to tell him where to find the
appropriate histogram definitions.
-
Erich Varnes (Arizona) has started to provide "user"-level
analysis examples and documentation for L3.
-
Lee Sawyer is continuing efforts to study the performance of the
missing E_T tool:
-
with/without NADA
-
good/bad calorimeter runs
-
with/without non-linearity corrections
-
before/after change in calorimeter threshold suppression
As far as the trigger list goes,
the initial aim is to add a missing E_T filter script (possibly plus
a very loose calorimeter electron) to provide extra redundancy for
W->enu events.
The minimum requirements to get a missing E_T filter script added IN
PARALLEL to
the already-existing "single electron" scripts should not be set too
high. We should adopt a similar philosophy to the one we adopted for
introducing L3 tracking in
parallel to the calorimeter-based single electron filter scripts.
It should be sufficient to demonstrate:
-
that an L3 missing E_T filter performs similarly to the offline missing E_T,
-
that it has a reasonable efficiency (a few times 10% is sufficient),
-
that the extra rate to tape is small
-
that it is relatively insensitive to the presence of hot cells
(Of course, all the usual technical
stuff concerning time/memory consumption, no
crashes, etc, is taken for granted.)
Scribe: Terry Wyatt.