Level 1 Calo Software Meeting at RAL on Thursday 23 November 2000.
------------------------------------------------------------------------

Present: Murrough, Scott, Steve, Bill, Norman, Bruce.
The anticipated video conference with Cornelius in Heidelberg
unfortunately had to be cancelled.

The informal agenda was:

- Status of current work
- Feedback from TDAQ workshop
- Discussion on HDMC and DAQ issues
- Experience of commercial tools [Steve]
- Workplan for forthcoming tests


At this meeting we said farewell to Scott and welcome back to Steve.


Scott reported progress with decoding our ROD slice data fragments
in his event dump. His program successfully handles fragments from
Bills test vector generator. He will try to tidy things up before
he leaves us.

Bills test vector output files now conform to the letter (for Slink
output) or spirit (for CPM output) of the test vector format agreed
with Ralf Spiwoks.
However its not clear if the DSS can generate (word by word) the
various Slink control bits [39:32] which are included in the format.

We discussed (again) the requirements on the DSS firmware. We have
to restate these to the engineers. The main ones are:
- variable length specifiable for input and output buffers
- settable number of words emitted from LVDS source in response
  to an L1A
- automatic comparison of input Slink data started and stopped
  by receipt of appropriate Slink control words

Bruce has had a DSS and ROD in his hands for a short time. There is
a test plan for the ROD (paper copy from Viraj) with the required
prescription to get the two modules to talk. There are updated ROD
specs available via his home page.

Murrough has taken delivery of a new Concurrent Technologies CPU
(a VP PSE/P34). Installation of RedHat 6.2 presented no problems
apart from two previously identified (with workarounds) by Juergen.
[PS after the meeting the vmelinux driver was installed and accesses
to VME made successfully. Bus errors need checking and integration
in HDMC still to be performed].
If the QMW tests are succesful, we agreed that we should buy another
two for RAL and Birmingham.


We discussed the use of various options for DAQ in the slice tests.
From the ROD and TDAQ workshops, it seems that we dont really need
to continue using PBM. Even with our own producer software to read
event fragments from our ROD, we could inject those directly into
the standard monitoring framework and then use the monitoring skeleton
to collect them for analysis, monitoring or event dump.
However we need to understand the monitoring framework a bit better
to be sure.

Integration of ROD fragment collection into the ATLAS Readout System
(ROS) is once again a live topic. Bruce and Murrough had some positive
discussions with Dave Francis. However collection of data from several
ROD crates is still unclear.

For the 'ROS in a PC' option, we need to buy Slink/PCI cards, together
with sets of Slink source/destination cards. We will need the latter
anyway if we connect all our RODs to DSS modules. We agreed that we
should buy six Slink source/dest pairs and at least two Slink/PCI cards
which have a long lead time (three months).


At the TDAQ workshop, we also had a discussion with Beniamino Di Girolamo
who was responsible for some of the Tile Cal test beam software. He has
given us access to his code and explained some database related issues.
He was also enthusiastic about early level 1 involvement with Tile Cal
electronics, eg to explore calibration issues. We was a bit disappointed
at our current preference to complete our internal integration first.


Although, unfortunately, Cornelius wasnt able to participate, we had
some discussion around the different philosophies of HDMC and that which
may be preferable in the DAQ context. Bruce and Cornelius have had some
exchanges by email. For the moment we are committed to what Bruce has
developed within HDMC.
For the future we should consider the best way forward, eg specification
of the parts and configuration file syntax?


Steve presented his experiences of software development in the commercial
world. Typically their software was divided into a large number of
packages. Modifications to these were the responsibility of a single
developer. The use of an integrated development environment (IDE) with
a good debugger and with seamless connection to source code version
management is extremely helpful. Aspects of a formal software process
(requirements documents, reviews etc) were also helpful, but were often
not followed when deadlines approached. Frequent short meetings were
the ideal: this is clearly harder for us in a very distributed project.

We discussed KDevelop on Linux. This is not a very mature product. We
should also look at commercial Linux based tools, eg CodeWarrior?


At both Heidelberg and the TDAQ workshop, Murrough had presented a list
of the software required for the slice tests. Putting some preliminary
names to this list is a first step towards a more formal workplan. The
full(ish) time available people in the UK include: Bill, Bruce, Murrough
and Steve, with possible contributions from Norman, Reg and maybe Gilles?

- Test vectors: Bill
- Online stuff (local controller etc): Murrough
- Database: Murrough
- Histo displays: Norman (to look at Andre Bogaerts proposal), Steve
- Readout: Bruce
- DAQ end of monitoring framework, buffer manager: Bruce
- User end of monitoring, analyser, simulation: Steve + Bill
- Event dump (in above context): Steve
- Calibration: NN

We should review the required work and schedule when the ROD is working.

As for documents, Steve should take a look at our long dormant overall
requirements document; we should dig out the test vector plan for the DSS
(originally an email from Murrough to Bill?); Norman should distill his
talks about the slice tests into test plans for the CP crate tests and
the full slice tests; Murrough may extract the calibration section from
the requirements document and elaborate it into something to send out to
the calorimeters for comment; we should discuss the URD for our use of
the database.