CERN QMUL
Mins-01-17
L1Calo Software meeting at RAL on 17 January 2002
Present: Bruce, Eric, Gilles, Murrough, Norman, Pete, Steve.Actions from previous meetings
- Data formats: there are still missing data formats, but Norman now wants to revisit the existing ones with the aim of improving data compressibility of the JEM data.
- Steve has started documenting his simulation package - this process has prompted some improvements to the code.
- Bruce has provided examples of using his module services package - but accompanying explanations are still outstanding.
- We still need to identify key HDMC issues where we would like urgent developments. Bruce will interact with Oliver.
- Other actions are carried over...
Murrough: progress with run control and databases
Murrough has changed the way run types are handled in the run control to a data driven style, removing more algorithmic ideas. The database now includes encapsulation of both run type dependent settings in objects for both general and module type specific settings. He has also started work on more dynamic run parameters, ie which live in IS and can be changed in the IGUI.All recent work, including changes to the database on moving to the latest Online release (0.0.16) are now in CVS.
It is not clear what is the best way to handle the mix of static configuration data, run parameters, calibration, etc data and how to pass it to the modules. Some must be passed at initialisation, some should come with the object that encapsulates the run type, some might fit into either category.
Bruce: module services and ROS developments
Bruce has finished and stored in CVS the modifications to the HDMC parts file syntax to support submodules. He would like to make further tests for its robustness.Testing of the ROD with the Slink receiver code from the ROS has shown up a flow control problem. Further tests, with logic analyser, are required to be certain this is at the ROD end.
Dave Francis has announced that a beta of the official ROS distribution will be available soon (perhaps a month?). We should test this as soon as it is available.
CERN has initiated an order for a many PCI slot industrial PC. This has a backplane with pluggable motherboard. The configuration they ordered has 4 standard PCI slots, 4 64 bit slots and 2 64 bit 66MHz slots. We also need ten slots, but probably prefer all standard slots. However there is now a 64 bit PCI adapter for the ODIN Slink receiver cards. For the CP subsystem tests we only need three slots, so we can wait a bit before deciding exactly what to order for the slice tests.
Gilles: CPM module services
Gilles has tried to use Bruces example "Bpm" module, but would like more explanation of the syntax. He has also had some problems building it since the example code lives in the l1calo/moduleServices area while everything else is in HDMC. He will send a list of changes he had to make to Bruce.Although it might at the moment be easier if the module services were part of HDMC, we expect that when module services start having dependencies on our other packages and the Online software that it will be easier to keep it separate.
Norman: progress with CMM and ROD crate DAQ taskforce
Norman has produced a list of the database and run parameter requirements for the CMM which he will send to Murrough. But we should decide where it is best to check the integrity of the data coming from the database. At the moment OKS does not support ranges of validity (this feature has been requested). Meanwhile the data could be check by the DAL or at the point of use. For example the RegBuild generated classes could throw exceptions if the data was outside the range of a bit field. Though some parameters, eg number of slices, have ranges which arent complete bit fields.Norman also reported on the first (phone) meeting of the newly formed ROD crate DAQ taskforce. The taskforce has a web page which has the two presentations made at the this meeting which we are encouraged to read and comment on. The mandate of the taskforce is to produce a specification for ROD crate DAQ to be presented at the next ATLAS week (mid February). So far it all looks encouraging - though Norman had to reiterate our requirement for collecting events from multiple crates.
Steve: simulation
Along with the progress with documentation, mentioned above, Steve has made a few updates. He has also split the code into subdirectories. He has not had further contact with the Stockholm diploma students.We briefly discussed test vector organisation. Steves original idea had been to generate both input data and expected output online. However this is likely to take time and we will probably want to run the same test many times, so saving and using files will be useful. However this means we also have to save to conditions (run parameters, calibration, trigger etc) used in producing the output.
We proposed to have a brainstorming session around this whole issue at the next software meeting. An email discussion beforehand will probably be useful to provoke ideas. We should also write down the detailed lists of tests and the test vectors required. Gilles has a draft document for the CPM which he will circulate more widely.
Murrough: Online software news
The latest version, 0.0.16, was released just before christmas and contains some useful improvements in the database area. This release also contains the new Online Histogramming package (really just for the transport of histograms). This component (like other Online software components) has a web page where you can find detailed guides aimed at end users and developers.Discussion: number of CPUs
At present we have three of the new Concurrent CPUs plus an older one at QMW (with incomplete bus error handling). For the CP subsystem tests Birmingham will need three crates and RAL will need at least two, probably three. We should buy another three of the same model with the QMW one as a spare.Schedule
Norman will send back to Murrough the draft software schedule as imported into MS Project. However for it to be really useful, it still needs links with the hardware schedule.Next meeting
Possible dates for the next meeting are constrained by various peoples holidays. We proposed to meet on Monday 4 February at Birmingham with a video conference in the morning and test vector brainstorming in the afternoon.Actions
Actions from this and previous meetings:- Norman: find/define remaining data formats
- Steve: documentation for simulation package
- Bruce/all: investigate software development tools
- Bruce: complete Module Service document
- Bruce/Murrough: define required HDMC improvements
- Murrough: complete run control design document
- Murrough: check D32 and D16 access with VMEBusEltec on QMW CPU
Last updated on 17-Jan-2002 by Murrough Landon