CERN QMUL
L1Calo Software Meeting at RAL on 21 August 2001
Present: Bruce, Eric, Gilles, Murrough, Steve.NB a summary of the actions arising from this meeting is at the end.
Firmware in CVS
Gilles described the first moves towards storing our firmware in CVS. He showed us the directory structure he is now using which is convenient for direct access from the Renoir/Leonardo tools. So far only a couple of CPM FPGA designs have been stored. It is hoped that the use of CVS can be extended to include all our FPGA designs. The matter will be taken up at the next hardware meeting.Last word on the requirements document
Murrough asked if the draft requirements document can be considered finished. Bruce wondered if all his comments had been included. This should be checked. Murrough also mentioned that some of the email exchange had taken place when Oliver was away and he had not received any feedback from him. Murrough will contact Oliver.Reports of recent work: Bruce
Bruce showed some slides covering his recent work.The three new Concurrent CPUs are working fine and can be booted over the network, following some guidelines on how to set up diskless Linux systems. This is a convenient way of working as the crate can be switched off without shutting down the system (in fact must be). Bruce still wants to investigate the Concurrent driver. He also needs to (buy and) create a suitable disk based setup for Birmingham who will need it in the near future.
Bruce has updated the HDMC repository with the latest versions of his L1CalDaq code, a new ModuleMemory16 class, and a very simple parts file for the new TCM.
Testing of the ROD firmware (ROI and slice data) continues. The SliDAD now allows proper testing of the XON/XOFF. Both firmware variants still show (different) problems. The fact that some problems unrelated to the reformatting of the different input data appear only in one variant is a little worrying.
Two ROD related worries were expressed. Firstly there was a suggestion that the manufacture of eight more RODs had not yet started. Secondly the timescale for production of the other firmware variants needs to be clarified. Firmware for the CMM will be required fairly soon and might also be useful soon for the JEM?
On the readout issues: before he went on holiday, Norman had some phone discussions with (at least) John Strong from Royal Holloway to indicate our interest in the new UK ROBins. In answer to one of our questions it seems that the new ROBins will have the fast ("two fibre") ODIN integrated on the board.
Bruce has asked Dave Francis for the ROS software. It seems that this is still not really available as a readily installable kit. However UCL and Royal Holloway have it and could in principle help us with it.
It was suggested that a suitable time for discussions with Dave and other members of the CERN ROS group would be during next ATLAS week (mid October). Bruce is unlikely to be able to start investigating the ROS software before mid september so that would give a month or so to understand it sufficiently to ask more detailed questions.
Reports of recent work: Steve
Steve has extended his simulation software to prepare new test vectors for testing the CP FPGA designs with the Generic Test Module (GTM). This was not a simple matter as the GTM duplicates data in a complex way to produce enough inputs.Steve would like to incorporate the latest version of Bills test vector code as updated by Bruce. He noted though that Bill provides trigger towers, whereas (in places) we would like to have CPchip inputs - ie containing parity and link errors per multiplexed pair. Bills code provides data in all towers, but only half can be sent over the LVDS links at a time. An extra step of halving the data could fix this perhaps.
Future plans include: extending the existing code to other modules, saving the whole package in CVS, adding some documentation and comments in the code.
The latter is especially important if extension to other modules, eg the JEM, is actually going to be done by someone else.
Monitoring requirements document
Eric has made the draft monitoring requirements document more formal. He will distribute this soon.Discussion of packages
The "Software Organisation" document is still a draft requiring more work. Some remaining issues include how to organise and build a number of separate packages. Although the Module Services package might fit into the existing HDMC directory structure and repository, other packages should be separate. The ATLAS style is that each package gets built into a common directory tree which is separate from the working area. The common tree has lib and bin for each machine architecture and an include directory with a subdirectory for each package. Murrough should present a proposal for our packages at the next meeting.We also discussed putting names against packages as follows:
- Database & Run Control - Murrough
- Hardware Access & Interactive Diagnostics - Oliver
- Readout/ROD crate DAQ - Bruce
- Simulation & Test Vectors & System Testing - Steve
Regarding the Histogramming package, Murrough reported that Maria Smizanska has produced a first version of her JAS based package. We havent looked at it to see whether it may fulfil the requirements we identified as the basis for the ROOT based package developed by Tara. If so, and if the new package is adopted by ATLAS, we should use it.
Discussion of outstanding tasks
A few days before the meeting, Murrough had circulated an email outlining some of the major outstanding software tasks. For the last part of the meeting we discussed this.The main points arising were:
- Organisational: names against packages and action on Murrough to make a concrete proposal for internal structure of packages are described above.
- Databases: hardware configuration and connections were felt to be the most urgent tasks.
- Hardware Access Library: some HDMC changes, eg relating to Parts, Part Manager and Modules are really required before we try to make a real start on Module Services. Murrough should talk to Oliver to discuss the timescales for doing this work. Other points mentioned were improving the Address classes and the need to look at FPGA loading.
- Module Services: we should all come to the next meeting with suggestions for how these should look, ie what services we would like to see provided by each module and major subcomponent.
- Run Control: interface with Module Services needs more discussion. Murrough and Bruce had some disussions but havent yet converged...
- Readout: action on Bruce to store existing code in CVS. Action on Norman to distribute the nearly complete draft of the ROD integration tests.
- Testing and Simulation: covered in the report on Steves work above.
- Other stuff: action on Murrough to "mercilessly" insist on reviews of existing documents (one per meeting).
Next meeting
The next meeting will be at RAL on Thursday 6 September. There may be a hardware meeting on the same day. We may also try again for a video link with Heidelberg.Actions from this meeting
- Murrough: present proposal on package internal structure etc at next meeting
- Murrough: contact Oliver regarding HDMC work, also Requirements document
- Norman: distribute ROD integration test report
- Eric: distribute Monitoring Requirements document
- Bruce: send updated Bill Stokes code to Steve
- Bruce/Steve: save existing work in CVS (but may wait for agreed suggestions on package layout)
- Steve: add comments/documentation to existing code
- All: bring suggestions for Module Services to next meeting
- Tony/Viraj: status of the rest of the RODs; timescale for other ROD firmware variants, eg for CMM.
Last updated on 23-Aug-2001 by Murrough Landon