CERN QMUL
Home
|
L1Calo Software meeting on 8 April 2003
Present: Cano, Gilles, Murrough, Steve, Thomas.By phone: Bruce, Norman.
CPM software status: Gilles
Gilles has added an I2C interface part to HDMC and uses this to provide access to all TTCrx registers. VME can now be used to set TTCrx deskew settings, but broadcast commands are still needed to reset counters etc when running test programs.Some two to four run types (using the new database scheme) are being used. The Birmingham setup is now regularly running tests with four DSSes and two CPMs.
The different database styles used in creating module objects between the kicker (DAQ list) versus the run control (online database) is becoming an issue.
JEM software status: Cano
Some of the JEM module services have been rewritten to make it a little more common. Algorithms for processing readout data are now in a separate class from the module services code itself. The analysis uses a buffer filled by the standard readout() method. Some further integration with the run control and more use of the database was implemented during Canos visit to QMUL.The database method to provide masks of backplane fanin from neighbouring modules would be useful for the CPM too. Similar masks for input channels (not yet implemented) would be useful. Also perhaps for Glinks?
Database and HDMC: Murrough
Some initial work has been done to add the possibility of defining which TTC broadcast commands are required at each state transition for different run types. [However the database library doesnt yet return this information to any potential implementation in the TTCvi module services].It was also realised that the scheme should be extended to allow L1As. Maybe there should be a TTC command object instead of the currently proposed simple list of broadcast command values?
The memory display GUI in HDMC has been rewritten to permit migration to Qt version 3. Unfortunatety the new class doesnt work with Qt version 2, so there are some ugly hacks in the HDMC Makefile to select between the two versions of the code. Murrough has also successfully compiled all the L1Calo software with gcc 2.95 and online release 00-19-00 on a RedHat 7.3 system - though none of this has been tested on real hardware.
RODs, ROS, OS migration: Bruce
Bruce has mainly been bug fixing while continuing testing ROD firmware variants.Bruce also reported on recent discussions at CERN with Dave Francis, Benedetto Gorini, Beniamino di Girolamo and Ralf Spiwoks concerning our data acquisition needs for the slice tests. we had been worried that the event monitoring facilities in the ROS had disappeared. They have now been reinstated, but it was suggested that we use the DataOut facility of the ROS IOManager instead.
It was also pointed out that the overhead in servicing ten Odin Slink PCI cards in one PC was very high. We were recommended to consider moving to the new generation of links (HOLA) and the FPGA based FILAR PCI card. Further discussions will be held with the ROS group.
Regarding OS migration towards the official test beam releases, Bruce intends to download Ralfs ROD crate DAQ test kit and to understnd how to use the ROS components. Meanwhile he has installed CERN RedHat 7.3 on the two new server PCs at RAL. One will be set up as boot server for the VME CPUs, the other will be a test platform for the new software. The date of the next JEM test visit to RAL seemed like a reasonable time to aim for.
Steve remarked that it would be useful if the Linux installation of a PC acting as a disk server to diskless CPUs could be provided (by Bruce), but he was happy to do the other software upgrades himself.
As part of other discussions Bruce was asked about the ordering of DaqSubmodules in SubmoduleLists which is reversed compared with the order in the DAQ lists. The CPM and JEM code take different approaches in dealing with this. Bruce should make a proposal on how (or whether?) to change this.
Simulation: Steve
Steve has received the JEM simulation code from Daniel and Asa. It is claimed to work for a single JEM but not for backplane communication between two JEMs. Steve suspects that this may be due to the way in which the Stockholm students have not always used his framework in the recommended way.The intention is that Steve will reorganise the code to use his framework as intended and then pass it on to Juergen for further development, such as the DAQ readout, RoIs, implementation of playback/spy memories etc. Juergen will also check the energy tree.
CMM software status: Norman
Norman has updated the CMM register model with the latest changes from Ian - though the module is now back in Ians lab. He hopes too look to the readout data via the ROD and intends to discuss the next steps with Bruce.The cmmSim package was cleaned up and updated by Steve some time ago. It should be possible to add it to the default list of packages.
TTC broadcasts
We discussed the issue of TTC broadcasts. The new TTCrx chips dont enable the deskew2 clock by default. We need a broadcast command for this - probably we should always do this for all modules. The JEM also needs a broadcast to set the TTCrx control register. We should define a common setup for the whole system.We also realised that we dont use the broadcasts consistently. The DSS uses the short form, the JEM the long form, while the CPM can use either.
Multiple crates
Bruce felt that our problems extending tests to multiple crates might mean the HDMC NetBus may need to be revived.allModules package
As raised earlier by Gilles, the different ways in which parts of our software create the moduleServices objects should be resolved.We have in the past discussed the need for an "allModules" package which (a) brings together all the individual module services package and provides a common HDMC binary and (b) could provide a single common way for standalone tests, "kicker" type readout program and the run control to initialise all the required DaqModules.
Murrough had proposed - to Bruce - an API for a DaqModule factory class which all our software could use. Bruce was happy with this approach - it just needs to be implemented... In the shorter term an allModules package should be created to provide the common HDMC binary. Later HDMC itself could be extended to initialise parts from the database.
Event dump
In Dave Kants absence, Murrough reported that Dave had made some progress with customising the event dump. We have also had some interactions and provided some feedback to the Online group who had taken on board our proposals.Next meeting(s)
To be arranged.Actions
Actions from this and previous meetings:- Bruce/Murrough: write and circulate proposal for handling multistep runs in DaqInterface and run controller
- 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: update database user guide
- Murrough: complete run control design document
Last updated on 11-Apr-2003 by Murrough Landon