Software meeting at Birmingham on 23 November 2004

Present: Alan, Bruce, Eric, Ethan, Gernot, Gilles, Jürgen, Murrough, Norman, Stefan, Steve, Victor.

The agenda has some transparencies with a set of points for discussion. These minutes are brief notes from the discussion.

Timescales and Priorities

The discussion focussed mainly on software for production testing.

• Software migrations (ROD crate DAQ and monitoring?): software for module production may come before at least some of these migrations
• Module production needs more automated software, including storage of results
• Production testing software for CMM will be needed first. May serve as an example to work to
• Online DVS style tests only useful for very simple tests. Scriptable run control would be useful

Software Review

• The meeting was in favour of having a review, fairly light on documentation, perhaps with offline software people giving "external" input
• Style should be 1-2 day workshop with presentations on major components and plenty of time for discussion
• Date around mid January was most favoured

HDMC Optimisation

The proposal to optimise HDMC parts management should be tried. If lists inside DaqModules are also too large further optimisation of the composites strategy may be needed.

VME read-modify-write

We presently use a read register, modify one bit field, write back whole register scheme to update bit fields in a register. This can have problems if the read is unreliable, eg PPM at the testbeam, or JEM in the past. There was plenty of discussion on this issue.

• Is it correct to maintain register persistence in a software model?
• What about access by run control and HDMC? Our VME access layer has no concept of access control and arbitration of it: would need mods to the driver
• Module services, invoked by run control, should be sure to explicitly set every bit. It could set whole registers and use software model for bit fields. HDMC diagnostics would still have to use read modify write cycles. A read-only HDMC might be useful
• One radical suggestion: remove all regbuild classes, do all bit manipulations in higher level code
• We should start an email discussion to explore ideas further and make a decision at the software review

ROD Crate DAQ

• The full agenda for the workshop is available
• Benedetto presented suggested ROS/RCD architecture changes, eg moving configuration from run controller to IOManager
• He was hoping for more specific input for architecture changes than was given in the subdetector talks (which formed the bulk of the workshop)
• Further input will be requested by email
• Markus Joos presented a new Concurrent CPU, the VP315
• Of the subdetectors, muons and calorimeters were happy with proposals to use RCD during installation. SCT and Pixels will not to VME readout?