Software workshop: 30-31 January 2006

Present: Bruce, Florian, Gilles, Kambiz, Marianne (part time), Murrough, Norman, Rainer, Richard, Sten (part time), Steve, Victor.

Introduction

The software workshop was held over two days with discussion initiated by a few talks. The talks can be found via the CERN agenda system.

These minutes do not attempt to summarise the talks, but contain only brief notes arising from the discussions afterwards.

CPM: Gilles

Gilles showed example configuration files for the PMP presenter. It was not clear if the configuration file will scale well when we have thousands of histograms.

The strategy and timescale for moving the existing CPM kicker to a RCD/GNAM style needs more discussion. Certainly it will not be until CPM production has finished and there are good working examples for other modules.

The present backplane timing scans are single crate activities. But timing scans of incoming LVDS signals might be multicrate.

PPM: Florian

The main point from Florians talk was that the present scheme for TileCal tests has scripts and database quite separate. Development of a more integrated "multirun plan" to drive a set of procedures is really needed.

When considering how to move PPM calibration data into COOL/RAL databases, Florian would like to keep the present structure as seen by ppmServices - this is rather close to the hardware, not an abstraction as would be preferred by some.

The VME readout will always need different RemFPGA firmware from that required for ROD readout due to the internal buffer organisation - at least for readout of a large number of slices. If we want to use just one firmware variant for everything that would imply some software changes eg in the FIFO scan.

CMM: Richard

The main questions from Richards talk are how we choose what data to load (in run controllers) and how it should be passed around. Filling something like the existing objects was the general preference.

When taking new calibrations, there should be separate programs that monitor trends of calibrations over time and that select whether a new calibration should actually be used.

We had a short discussion about the use of COOL channel IDs. It is not clear we will be able to have individual PPM channels as separate channel IDs in COOL (due to limititations in ATHENA) we do want to have an encoding that identifies crate, module, and channel within module. Richard will look at the prototype in the coolL1Calo package.

Richard is continuing to gain experience with COOL and RAL. It is clear we need further detailed discussion on how to use them.

JEM: Rainer

Rainers talk triggered a discussion as to whether we need readout for the full crate test of JEMs. It seems this may not be done for the equivalent CPM full crate test.

ROD Crate DAQ: Murrough

Murroughs initial implementation of ROD Crate DAQ readout is (mis)using the TDAQ beam crate source ID for labelling fragments read from VME. We need to discuss and agree some other scheme that will allow us to distinguish between VME and ROD fragments for bytestream decoding.

There was a request that the present VME readout format be documented: Bruce agreed to do this.

The ROD Crate DAQ readout really needs an API change in moduleServices to indicate when an event (in PPM or ROD) is ready for readout. Other moduleServices API changes were mentioned during the meeting (see below) in addition to the expected addition of (dis)connect methods associated with the change to the run state model. We need to consider the issues further via an email discussion on the subject.

Monitoring: Steve

Steve asked if we should have a scheme to avoid hard coding names of histogram servers. Murrough agreed to provide something quickly.

There were various questions about GNAM, some of which were subsequently answered by Paolo. Are the commands from the presenter to rebin or reset histograms implemented by the GNAM framework? Here the answer is yes. Only really custom commands need to be implemented by the detector developers. However it is still not clear if one GNAM program can collect data from more than one monitoring stream.

There was a discussion on how to handle the histograms from the PPM. These should probably not go via the same monitoring stream as we are currently using for events read from VME. One suggestion would be to use the publishFullStatistics method of ROD Crate DAQ ReadoutModules for publishing larger volumes of data less frequently than our existing statistics monitoring (eg error counters) which is sent to IS. The PPM histograms could be published directly to OH. This would need an additional change to the moduleServices API.

Next meetings

The next phone meeting will be on Thursday 9 February at 11:00 CET (10:00 GMT).