Michigan: M. Abolins
Argonne: R. Blair
UCL: R.Cranfield
Cracow: K. Korcyl
NIKHEF : J. Vermeulen
CERN: S.Wheeler, P.Golonka
1. Paper modelling: A C++ program replacing the spreadsheet has been developed by J. Vermeulen. The motivation has been that it has become difficult to make modifications to the spreadsheet without introducing errors. This is due to the complexity of the spreadsheet and the mixing of input parameters, calculations and results which is typical for a spreadsheet (the program has become a few days after the meeting available via the modelling web page). No further development of the spreadsheet is foreseen, output of the program should be documented and made available via the web page or in DAQ notes. V. Vercesi has been asked to co-author future documents on parameters and results. He has provided new parameters, J. Baines also provided a new B-physics strategy, a study of which probably would require major modifcations in the spreadsheet. Now the program is available the effect of the parameters and new B-physics strategy are to be studied
2. ROS modelling: There were no new developments to report. Some time was spent on discussing how to proceed with the model required for the performance testbed. J.Vermeulen suggested to have a look at the ROS performance test results and see how we could use these for a simple model, possibly building on existing at2sim or simdaq models. R. Cranfield and J.Vermeulen intend to discuss this issue in the week of the 17th of June in detail.
3. Data Collection modelling and HLT modelling: P. Golonka has developed a new model of the LVL2PU, based on the new Data-Collection software and for this reason not based on the model for the LVL2 processor implemented by S. Wheeler. For calibration of the model the API for time-stamping developed by the ROS group is to be used. This introduces an overhead of about 800 ns per time-stamp. clock synchronisation is still an open issue. The new model is also to be used for the supervisor emulator, P. Golonka will send R. Blair information on the model once it is running. R. Blair remarked that for further development of the real supervisor / RoIBuilder it is necessary to define error handling pocedures and that using the emulator probably is good enough for the performance testbed. Finally, P. Golonka mentioned that he has tried the new release of Together but still found problem areas.
K. Korcyl reported to be working on a description of the parameterized switch model and hoped to produce in a about a week after the meeting a first draft. He also mentioned problems with at2sim and Ptolemy, which are described below (agenda item on at2sim).
S.Wheeler is looking after the DFM and SFI models. S.Gadomski has not done further measurements on the DFM and is away for a few weeks. For the SFI contact with R. Mommsen has been established. For event-building the TCP-IP protocol has to be used. It has to be discussed with K.Korcyl and P.Golonka how this can be done without the complications of the detailed TCPBroker model. The document on the SFI model should be finished in about a week from now, it is not clear who will do the required characterization measurements. Concerning the DFM model: this has two input message queues, one for input from LVL2 and one for input from the Event Builder. Messages arriving in the LVL2 queue are handled with priority. It has to be checked with H.P Beck whether this (two queues, one with priority) is correct. It is probably possible to use P.Golonka's framework for implementing the model of the DFM.
P.Golonka's scheme is also to be used for the ROB emulators, for the FPGA based ROB emulators Asgeir Bjorlykke's model is to be used.
4. at2sim: K. Korcyl reported to have not been able to solve the problems with compiling at2sim and Ptolemy itself with version 3 of the Gnu compiler, it seems that there is no solution for running under RedHat 7.2. It is hoped that the problems will be gone under RedHat 7.3. In the mean time a machine with an older version of RedHat has to be used for running at2sim or at2sim has to be compiled with optimization switched off. For large models the Ptolemy scheduler is responsible for consuming a large amount of computing time. In Krakow Anna Kaczmarska will help to see whether it is possible to reduce this computing time so that large models can be run at an acceptable speed.
5. Simdaq: No news
6. Chiba City: No news
7. Workplan: Modelling is not known to be reviewed during the July T/DAQ week. However, it seems wise to be prepared for answering questions during the week, reason for having another meeting before that week, see the next agenda item.
8. Next meeting: The next telephone meeting is planned for Tuesday 25 June, 15.30 h CERN time.
J. Vermeulen, 11 June 2002