CERN QMUL
L1Calo Software meeting on 2 October 2002
Present: Bruce, Eric, Gilles, Murrough, Norman, Steve, Thomas.Calibration Document
We reviewed the draft calibration document. There were a number of suggested clarifications and an extra section (describing our requirement for offline or event filter analysis) was proposed. Norman will send some text to Thomas. Other minor corrections should be sent to Thomas directly as soon as possible. He will try to get a final draft out to the calorimeter representatives by thursday evening.Progress report: Norman
Norman has been trying to estimate the time taken to process our calibration events in real time. His simple model estimates about 60 us per event per preprocessor ROD (handling 512 channels) for the energy calibration. That just includes histogramming the pulse height in each channel. However a single program analysing the data from all 16 RODs could not keep up with the expected trigger rate of the standard LAr calibration. The timing calibration is several times slower, taking about 1 ms per event per ROD. We could consider adding a DSP to the preprocessor RODs. In this case we could cope with the LAr rate if they changed to run all eight of their partitions synchronously.Progress report: Bruce
The ROD test software has stabilised and after some working sessions together with Murrough and Steve the "kicker" program runs under the run control.Together with Murrough, some memory and initialisation problems in HDMC have been identified using the valgrind tool. However another HDMC crash reported by Gilles is still present so further investigation is still required.
Bruce also got the diskless RedHat 7.2 CPUs to boot over the network. His experiences, hints and instructions are now on the web here.
Progress report: Murrough
There have been a few database changes to provide access to the test description ("Bill file") used to generate test vectors. However more work is still needed to allow easy selection from the IGUI.We also need further discussion to sort out the architectural issue of how and where to uniquely define some key parameters (eg number of slices) which at present are specified in both the Bill file and the database.
The Online group recently recommended a memory debugging tool called valgrind. Details of valgrind can be found via its website. Murrough has fixed a few problems in the database and (with Bruce) in HDMC. It looks like a very useful tool and is easy to use and install.
We recently obtained some AFS disk space at CERN at /afs/cern.ch/atlas/project/tdaq/level1/calo. At present this is used to make nightly builds of our software using the CMT based test release of the Online software.
Progress report: Steve
Apart from working on the PPRP presentation, Steve has done a little work on the ROD test integration. In particular looking at the CPROD simulation compared with the actual ROD behaviour. Some further inconsistencies between documented and observed behaviour have been identified.Progress report: Gilles
Gilles has been making good progress testing the CPM. He is mostly using a standalone program (akin to the CPROD test Looper program) together with HDMC for diagnostics. The standalone program is based on the CPM module services.As mentioned above, there are some problems with segment violations in HDMC which can be worked around, but which have still to be understood properly.
The next steps are to look at the CPM output data via logic analyser and DSS. Also, the standalone program currently reads thresholds etc from flat files. Integration with the database is still required.
Gilles also raised the question of how we should organise firmware loads. This should be the subject of a future brainstorming session.
Progress report: Thomas
Testing of the JEM is continuing. There are about 10 test programs which are shell scripts directly calling VME driver utilities. These have recently been converted to use the CERN VME driver which is now installed on both Concurrent CPUs in Mainz - however Thomas reported a problem where the system randomly hangs requiring a reboot. No such problem has been observed at RAL (except for a separate problem with interrupts) so perhaps this is a hardware related problem.Thomas has tried setting up diskless network booting of their VME CPUs from a new server PC, following Bruces instructions. However this preocedure does not yet work.
Next steps: CMM, CPM, JEM
Norman now has the complete CMM memory map and can now proceed with the remaining CMM services and CMM simulation.CPM tests need software for the DSS LVDS daughtercard.
Thomas estimated that it may be two or three months before the JEM is integrated into the module services framework. Meanwhile the Stockholm students have restarted work on the JEM simulation using Steves framework, but it is not clear how far they have progressed.
Some further work is required integrating the test vector generation for the JEM into Steves framework. Steve will discuss this with Thomas. Extension of the CPROD simulation to cope with JEM data is also needed. This is now thought to be difficult.
At this point, the video conference with Mainz ended.
Release strategy
At a previous meeting we had agreed to make a "release" of our software to include the functionality needed to allow Tony and Eric to use the software to test the remaining ROD and DSS modules.After some discussion about our priorities, we reiterated this decision and agreed the additional functionality we still want to add and the timescale on which we aim to implement it.
- The database and IGUI should allow the user to select a test description file interactively. However for the moment this will required separate interactive changing of the number of slices parameter as we havent properly sorted out the duplication of information between the database and the test description file.
- The test vector generation should issue MRS messages if there is an incompatibility between database and description file.
- The kicker program should also issue MRS messages when it detects errors in the data stream. It should also update status information into IS.
- The timescale for these is the end of next week (11 October).
Documentation tool
We agreed that in future any collaborative document should be written in Word. However we did not achieve a concensus on whether documents provided by individuals (eg user guides to software packages) must also use Word.Documentation status
We briefly reviewed the status of our documentation. Some user guides to the database packages, the run control and module services have been on the action list for an embarassingly long time and have still not been written. The developers concerned must make writing these a priority.We also felt that we should write an overall user guide to the software. Initially this would be aimed at the CPROD test setup but would be expanded in future. This should be at the non-expert level. We dont feel we need to turn existing web pages on how to build or install our software and tools, but links to these pages could be added in a short section in the user guide. Norman offered to write this document (in Word) when the time came.
Next meeting(s)
It is not clear if we will want to have another software meeting before the Birmingham joint meeting. If so, this will be arranged later.It was proposed that the agenda for the Birmingham software meeting be organised as a "mini review" of our status and future steps.
Actions
Actions from this and previous meetings:- Bruce: complete Module Service document and submodule syntax description
- Bruce/Murrough: define required HDMC improvements
- Murrough: tackle project management issues: eg schedule (Gantt chart, stages of evolutionary delivery), risks, etc
- Murrough: write database user guide
- Murrough: complete run control design document
Last updated on 02-Oct-2002 by Murrough Landon