CERN QMUL
Home
|
L1Calo Software meeting on 11 June 2003
Present: Bruce, Cano, Gilles, Jürgen, Murrough, Norman, Thomas.JEM software status: Cano
Mainz have experienced some problems after updating HDMC (segment violations) following the distribution of the updated RedHat and Online software which they copied to a new larger disk. [A workaround - but not yet a proper fix - for this was found after a separate phone conference after the meeting].Cano also reported seeing unreliability in the behaviour of the DSS in recent tests. This is not understood. It was thought not to be related to recent changes to the DSS software that was requested from Bruce to facilitate use of run types not using the DSS.
The planned visit to RAL next week is still not fixed and depends on rapid progress with further tests of the readout controller.
CMM software: Norman
Norman has been concentrating on the CMM simulation. He has completed the real time path: reader to crate merger to system merger to writer, including cable input, for the CP mode. He has also implemented the generation of Glink readout data from both crate and merger FPGAs including Steves new L1A and multi slice readout builder.He is still working on interfacing the top level containing to the database and the mechanism of selecting and connecting up the components depending on the firmware mode. Implementation of the jet and energy variants is not yet started.
Another task will be the analysis of events from a multi step timing calibration from which the trends and optimal settings need to be extracted.
Norman also raised some questions:
- Q: What happens if modules listed in the database are not present? A: Modules can be dynamically disabled via IGUI panels. If this is not done, the run controller may get errors when the DaqModule is created. To be checked. One suggestion may be to get the run controller to test a nominated register (eg ModuleID) for each module before trying to create the complete set of DaqModule parts.
- Q: How to debug the simulation when it is invoked under run control? A: There is a convoluted way to debug a run controller. But to debug the operation of the simulation configured from the database it would be easier to use the standalone testdbsim program from the dbSim package. (Steve when he returns may have other suggestions).
CPM software status: Gilles
The Birmingham setup now uses a CPU in each crate as at RAL, ie the CPM is now in a crate containing a CPU rather than the Bit3 interface. Copying the RAL setup also reproduces the problems seen at RAL and consequently Gilles has been mostly chasing hardware problems recently.The latest changed to the CPM module services was to take FPGA information from the database. However Gilles reported a problem with the pathnames returned by the database.
He has also noticed some other problems:
- The run control log files still go to /tmp despite making changes to the .onlinerc file.
- It seems necessary to start and exit play_daq on each crate CPU before the play_daq on the PC works properly. Gilles thinks the ssh setup is done OK
- Sometimes the HDMC Register GUI isnt updated properly on reading another register.
JEM simulation: Jürgen
Jürgen reported that the realtime path for both jet hits and the energy sums was now working correctly. These can read and write files in the formats used by Cano. A suspected problem with the jet algorithm code has been reported to Stockholm.Jürgen has also implemented the readout - but this is not yet debugged. He hopes he would be able to provide something for JEM tests at RAL if they happen next week.
The next steps are to check the operation of more than one module and to configure the simulation with settings from the database.
DSS, ROD: Bruce
After changes to the new RedHat/Online software and the move to the new DaqModuleFactory Bruce has noticed problems with the CPROD Kicker. It now seems to require a delay at startup in order to work properly. With this, tests on the ROD firmware variants continue.The DSS module services were changed (under pressure) so that the DSS can be left unused (while remaining enabled) by some run types in a database configuration.
Bruce also mentioned the new DaqModuleFactory being used in Kicker and also the new dbhdmc. He said these appear to be in good shape.
Pressure of other activities mean that Bruce has not yet had time to look again at the ROS software.
Database and HDMC: Murrough
Murrough also mentioned the recent development DaqModuleFactory. This required a few changes also in the database and run control packages.Other recent changes to the database code include support for updating calibrations with new data. This does mean that we need to decide where to keep calibration data as it would no longer make sense to have the present empty files as part of the database source files in CVS.
It was pointed out that we now have several types of "non-repository" data and we need to decide how best to manage them all. A brainstorming session on this is required.
There have also been some other updates to HDMC. The existing DataSink part was udpated to allow dumping data from memories into a file. Also a new SlinkDataSink part was added specifically to find and dunmp Slink packets from memories into byte swapped binary files suitable for reading in the event dump.
Prompted by Normans imminent need to monitor events from timing calibrations, Murrough also started looking at the event monitoring package from the Online software. It doesnt look too hard to get the CPROD Kicker program to store events into the monitoring system.
Event dump: Murrough for Dave
Dave has now added display of JEM data fragments to the event dump. Up to now only CPM data fragments were supported. The standalone ed.sh script to start the display is also fixed.Dave had asked for feedback. Bruce remarked that since the data formats were still in flux and subject to review it might be best to add functionality to the existing displays rather than try to support more format which might change.
Next meeting(s)
Wednesday 2 July at QMUL.Last updated on 12-Jun-2003 by Murrough Landon