DØ Level 3
Algorithms Meeting: 3rd July 2002 in the Farside
Talks
-
Status of Trigger Examine: Elliot Cheu
Elliot showed example plots from the L3 trigger examine program run
during datataking online.
Documentation will be provided with the aim of getting shift captains
to start running it very soon.
There are a number of open questions of how this should develop in the
short term:
-
Breakdown of histograms by trigger type would be useful, but it's not
clear how to organize this. Elliot is currently working on a scheme
that would make separate histograms for each tool and each individual
trigger in which that tool is called. This works automatically and
doesn't require maintenance. However, it generates a lot of
histograms. What we probably want is to group related triggers
together, but how to define this and maintain it is not obvious.
-
Perhaps separate histograms for each stream would be more appropriate?
Compared with individual triggers there are a
smaller number of streams, each of which contains a logically related
set of events that it would probably make sense to group together for
histograming purposes.
We should keep the trigger_examine code version consistent with that
running in L3 farm. (Let's hope L2 and L3 keep roughly in phase.)
Discussion Items
-
Status of p12 preparations
-
CFT-only tracking release delayed, but pushing hard to go into release
by the end of the week.
-
It is not at all clear that the improved shower shape cuts for
electrons will make it in time. Ulla Blumenschein and Volker Buescher
are still studying
algorithms in the context of offline. L3 coding has not yet started
and they are both in Oklahoma next week.
We shall review the situation in the L3 meeting after Oklahoma.
However, there is a new spirit of quick convergence and not developing in
the production branch unless absolutely essential.
We should be prepared to stick with what is running currently
unless there is a significant improvement and we can convince
ourselves the changes can be made without disruption to the p12 schedule.
-
L3 Thumbnail: Peter Tamburello has rejoined DØ as a postdoc
with Arizona and has agreed to take responsibility for implementing L3
physics objects into the thumbnail for p13.
-
Primary vertexing in z:
-
Question of falling efficiency with p_t cut in W->munu events still open.
-
It would be very useful for Chris Barnes to make a comparison of the
performance of the track-based and smt-hit-based primary vertex finders.
-
Answering questions posed by run 2b review committee
wrt. justification of resources needed at L2 and L3.
Please think about this! We'll discuss it again next time.
-
Jim Linnemann would like us to come up with hare-brained ideas
for migrating L3 algorithms to L2.
A couple of first ideas:
-
Impact parameter tag runs quickly (by L3 standards). L2 will have run-by-run
vertex and L2STT.
Hopefully we shall be testing this online in p12 (see above).
For L2 we would need from Per Jonsson a better
estimate for how long it takes to run
on Linux with maxopt.
-
Primary vertexing in z: L2 doesn't have 3D tracking. The comparison
of track-based and smt-hit-based primary vertex finders (see above)
would give us an idea of whether it might be worth investing effort in
getting smt-hit-based primary vertex finding at L2.
-
L3 resource estimates needed for Run 2b.
Difficult to come up with anything concrete when we don't know what
we're going to be doing for run 2a.
However a couple of things might be relevant:
-
We are drifting towards the fairly expensive strategy of running full
reconstruction on the events. See for example electrons, where we are
already doing full calorimeter and track reco. Can we do "zoned"
reconstruction? My preliminary answer would be no. We need all tracks to find
the primary vertex in z to get best possible E_T resolution. We need
all calorimeter towers to do missing E_T, which we will add soon (in
association with a loose electron) to improve redundancy for W->enu events.
-
We should run the standard trigger list on some events with large
number of min.bias to get an idea of the increase in cpu time.
Scribe: Terry Wyatt.