Report from the L3 PDR committee

Dave Brown, LBL
Bob Jacobsen, LBL
Usha Mallik, Iowa 
Helmut Marsiske, SLAC 
Fernando Ferroni, Roma

Below are the findings of the L3 PDR committee.

Progress:

The L3 group has, since the last CDRR + PDR, decidedly made progress in all of
the right directions. It was impressive.


Fulfillment of Requirements:

The proposed architecture has been shown to be feasible, and the initial
implementations are functional. The actual design shows good rejection against
a variety of backgrounds that seems to cover all the known sources.
From the point of view of physics the design, based on L3DCH and L3EMT trigger 
modules, looks conceptually simple, very effective and robust.

The strawman trigger satisfies the physics selection and background
rejection requirements.  The strawman use of computer resources was
estimated to be close enough to the 10ms/event design target as to
satisfy the requirements.  

Near Future Priorities:

The L3 group should concentrate all their resources into being ready for
the cosmic ray data taking in the OEP environment: i.e, a minimal chain
of data taking mode using L3 and OEP should exist and be tested by this fall.
The L3 group should ascertain that all the infrastructure is in place,
including the L3/OEP interface, the L3/L1 interface, and a simple version
of TrgConfig (see below).  The developers of
the system can, in the process, find any unpleasant surprises and address
them. This is also the best way to pass their expertise on to the new people.

We recommend concentrating on infrastructure and interface development
as a first priority for the L3 group.  In particular, development of
a next-generation strawman should be defered until the full chain of
the trigger has been demonstrated to function in the online environment.

The L3 group should periodically monitor the computer resource consumption of
their code as they develop it, to ensure their design can continue to
meet the computer resource use requirements.



Infrastructure and Design:

The L3 group has provided solid justification for the need for TrgConfig
functionality.  Since the proposed design involves major changes to existing
code, we recommend that the L3 group provide a basic implemention
capable of fulfilling the cosmic ray requirements (L1-L3-Oep chain all
working consistently together) as soon as possible.  In particular,
driving the system from the configuration database might be best
left for a second-generation implementation.


Algorithms:

The DC algorithm has shown sufficient results that it should be put to
rest for the next couple of months in favor of higher priority for
infrastructure work.  In the Fall, it should receive any necessary updates
and be prepared for a first look at cosmic data from the real DC.

The Level-3 group has made impressive progress in utilizing L1EMT
information to efficiently select physics while suppressing background.
It is, however, becoming clear the one has to go beyond L1 primitives
-- further progress can only be made by using crystal and cluster
information. While this should not be the immediate priority of the 
core Level-3 group, strategies to unpack relevant (i.e., above-threshold) 
crystal information and fast clustering need to be investigated,
together with other promising signatures to distinguish signal from
background. It is necessary that the EMC group contributes significantly
to this effort, and they should make every effort to find the manpower
asap.

The Level-3 group is keenly aware of the requirements imposed by the need
to handle various types of events (e.g., Bhabha and muon pair events)
for luminosity, and calibration and monitoring purposes. They have made 
good progress on trying to identify Bhabhas with high efficiency, while
keeping them out of the physics lines(s). While the current design does 
not quite meet the stringent requirements for a Bhabha veto, the core
Level-3 group should not devote any more effort to this issue. Rather, 
it is crucial to secure additional effort from the EMC group which can
be directed at this problem. (NB: The EMC has indicated that additional
manpower to address this issue will become available over the next one
to two months.)

Any optimization of the algorithms for faster processing or
better coding should be defered until the minimal L1-L3-OEP chain is
working.

Manpower:

The manpower issue, as presented by L3, concerned the committee.
The L3 core group needs two postdocs, relatively experienced in software,
to replace the outgoing experts. The transfer of knowledge might best be
accomplished by geographic proximity of the relevant people.

In Addition, subsystems  (DCH
and EMC) should be encouraged to devote some of their software
manpower and expertise to the question of specific algorithms.  In particular,
additional EMC personel who could work on the EMC trigger and
BhaBha selection seems appropriate, as was mentioned earlier.


MDC2:

Nothing specific is required of L3 for MDC2 or MDC2p.  None-the-less, they
should continue to ensure interoperability with the rest of the online and
offline systems by continuing to build and release their code.



FDR Schedule:

The committee feels the FDR should take place before the start
of the cosmic ray run.  As that date is subject to optimization, we cannot
propose a specific date at this time.  This situation should be reviewed
mid-summer and an appropriate date chosen at that time.