Software phone meeting on 9 February 2006

Present: Bruce, Florian, Gilles, Kambiz, Murrough, Paolo, Richard, Sten, Victor.

Status reports

Birmingham

Little to report.

Heidelberg

Florian has spent time reconfiguring up the test system at Heidelberg and has also started reading up about GNAM. He has fixed a bug in the TileCal test code so that it is now possible to select any part of the readout buffer for use in the PHOS4 scan.

Victor has nearly finished his VME "byte stream decoder" package, apart from the handling of BCID flags and crate addresses. This also includes code to make some histograms. He and Murrough should consult about committing this to CVS.

Mainz

No first hand report, but Murrough reported that Rainer and Stefan had found they needed to upgrade the kernel on their local system before they could rebuild the drivers for tdaq-01-04-01. Once this is done their upgrade should be complete.

Bruce reminded us that we will need to think about which diskless booting paradigm to use in future. Meanwhile Gilles will circulate his tips for rebuilding drivers at Birmingham.

QMUL

Paolo now has both a GNAM decoding library and a library to make some simple histograms. He is still testing the latter. After that, these packages should be committed to CVS so others can start to use them as examples.

Murrough had realised during the software workshop that there might be a problem with multistep runs which make a single readout during state transitions when using ROD Crate DAQ. He has confirmed that this is the case. There are some possible solutions which need more work and discussion with the experts. However this does not affect the case where we want to make many triggered readouts during the running state which is what we need for the TileCal tests, so there should be no problem there.

RAL

Bruce mentioned that CERN will now be moving to SLC4 in the autumn (as RedHat 5 is expected to be delayed) and to gcc 3.4.4. The transition will start earlier and we could then check our software with the new versions.

Stockholm

Sten reported that they are still setting things up in Stockholm and chasing problems with CMT and CVS. Initially they installed a later version of CMT that we use elsewhere and moving to our standard version may help.

Module Services API

Following the discussions during the software workshop, Murrough had circulated a summary of the moduleServices API changes that had been mentioned there. At the meeting we discussed them a little more. Bruce suggested that the method to poll whether readout data was available might be implemented only at the module level, not by submodules. This method would be useful very soon for ROD crate DAQ. However it is clear that the methods for publishing "full statistics" (eg PPM histograms) need more thought and discussion.

There was a question whether we want to use the interrupt handling capabilities of ROD Crate DAQ. As both ROD and PPM support this we might like software to test this. However at the moment we have no other definite use cases so it remains a low priority.

PPM Readout Issue

There is a problem that PPM readout via VME of many slices (eg 100) seems sometimes to be truncated. It is thought that this is not related to buffer sizes, but more likely to the time take to perform the readout and possible interaction with state transitions in the present setup. It is hoped this will not be a problem under ROD crate DAQ.

Formatted Buffer Changes

Bruce still has changes to the formatted buffer in the infraL1Calo package waiting to be committed at a suitably quiet time...

TDAQ Patches

Bruce asked whether we needed to install further TDAQ patches on our systems. Murrough will check if there are critical ones we ought to install. It will probably be easier for people to install them at each site rather than distribute new tar balls from RAL.

Next meetings

The next phone meeting will be on Thursday 23 February at 11:00 CET (10:00 GMT).