Software meeting at QMUL on 6 July 2005

Present: Bruce, DaveC, DaveS, Eric, Florian, Gilles, Hans-Christian, Jürgen, Kambiz, Murrough, Norman, Rainer, Richard, Stefan, Steve.

The agenda of the main meeting also shows the outline agenda for the software meeting and has links to the slides shown. These minutes, written some time after the meeting, contain only some very brief notes from the discussion. The notes are reordered to follow the agenda although some items were actually taken out of sequence.

Steve: Simulation

Steve gave a summary of the status of simulation and test vector generation software. Some simulation packages are missing support. We need to identify individuals who will be responsible for cmmSim, jemSim and ppmSim in future.

Steve feels the test vector generator scheme could do with some improvement. Its hard to write the infrastructure for new generators although the generators themselves are simple. Ideas welcome!

Points raised in the discussion:

• we would like a scheme to run a whole test, setting various other parameters as well as test vector generators (see also under Software for Production Tests)
• we need to code a complete knowledge of the final ATLAS cabling and its mapping to (eta,phi) towers. We could then create more physics style test vectors.
• the documentation on the meaning of various run types and appropriate test vector generators needs to be improved. Especially for the benefit of non-experts.

Calibration and Multistep Runs

We had a couple of reports of good progress in this area.

Richard and Norman have successfully used multistep runs and monitoring streams to do timing scans on the CMM, both for the readout pointers and the backplane and cable delays. There are a few unresolved issues:

• one is unreliable setting of some TTCrx values. This may be related to the difficulty of synchronising all the programs at run start with our present setup. This may be improved with the new TDAQ release and moving the hardware monitoring into the ROD crate DAQ controller.
• another problem is how to know that the monitoring has returned the last event. Even with GNAM, which has a better link to run control states, this may be hard as monitoring operates a "pull" architecture - the analysis program has to request events and will be stopped at end of run.
• for the moment at least, the parameter value is stored as part of the data stream. This makes it easier for the monitoring data to be stored in a file and only analysed afterwards.

Florian has also made good progress with PPM calibration. He now has a number of run types which will successively find the readout pointers and make a PHOS4 scan to find the precise pulse timing of each channel in one PPM. Although this can be improved it is already good enough to be a fallback solution for the autumn TileCal installation. He gave a full report in the main meeting. The discussion issues were:

• the present scheme involves multiple scans. It would be useful to be able to run them automatically in sequence.
• so far we only scan one parameter at a time. A multiparameter scan may be needed to explore the large BCID parameter phase space.
• as with the CMM scans, the parameter value is stored in the data stream. We need to discuss what we put in the data stream.

We agreed to discuss some of the unresolved issues further on Friday afternoon. However following the attacks in London on Thursday this discussion was cancelled. We will have a phone conference at a later date.

Databases

Murrough showed a few slides related to databases. These represent some thought and little coding. He has tried using the interfaces to the LCG software now provided in the new TDAQ release. With a few fixes required in TDAQ, LCG and Oracle installations he has successfully accessed a COOL database via Oracle at CERN from QMUL.

Meanwhile Richard and Norman have also got started with COOL developments, so far trying to run the LCG test programs on the atlun systems with a MySQL server. The lack of documentation for running away from lxplus has made for slow progress so far.

We agreed that those interested in databases should start to have more regular discussions in order to agree (and document) how we will use COOL and then how to proceed with software development.

OS/TDAQ Updates

A few days before the meeting, Bruce proposed that we set aside next week (11-16 July) for moving our software to SLC3 and the latest TDAQ release (tdaq-01-02-00). A lot of preparation has been done already in occasional tests by Murrough and Bruce.

Assuming the migration at RAL is successful, we will freeze the old versions of L1Calo software. Other sites would then migrate as soon as possible, eg within a week or two.

Bruce proposes not to distribute ready made hard disks as before, but we will look at making a tar file of other software, eg LCG and TDAQ installations which need a few fixes and customisations.

The only OS uncertainty was in the currently preferred scheme for network booting. Bruce will try and find out.

We should check with the TileCal group which TDAQ release they expect to use in the autumn installation tests. Another release (tdaq-01-03-00) is expected fairly soon.

Software for Production Tests

At the last meeting Murrough proposed trying to obtain some software from the TGC group which allowed python scripting of online software services. This could be useful in setting up simple scripts to perform a set of runs with different IS variable settings.

Unfortunately it seems this software is no longer available. However the TDAQ release does now have some command lines tools for scripting the run control, but nothing for IS variables. We would have to provide this, or ask the TDAQ group to do it.

AOB

Documentation: this is always in need of improvement. At least we need to prioritise better documentation for TileCal installation test procedures and for production tests.

Error reporting: some log files need a cleanup, removing excess debug printout. The new ERS package in the new TDAQ release may help us organise and control debug printout. Stefan volunteered to look into what ERS can do.