Argonne: R. Blair
UCL: R.Cranfield
CERN: D. Francis
Cracow: P.Golonka, K.Korcyl
Michigan: M. Abolins
NIKHEF: J.Vermeulen
1. ROS modelling: R. Cranfield has written the code for modelling partial event building in the ROS modular unit and is trying to get it running. The calibration is to be discussed with B. Gorini, as soon as the model is running. Due to the partial event building multi-frame messages may occur, P. Golonka has to look into this as this type of messages may not be handled properly.
2. Data Collection modelling, HLT modelling, at2sim: A model for the testbed has been set up, with 128 FPGA emulators connected to 4 concentraing switches. The switches in the model are non-blocking switches, the SFIs are single processor machines without output to the EF. The next step is to run the model, and study the effect of varying the interrupt coalescence in the SFI and the number of credits. A script for the Alteon ROS emulators also exists. In a configuration with two EB switches and associated SFIs the LVL2 system can be simulated by the second EB system. This is more straightforward to implement than a LVL2 + EB system, but the LVL2 system should also be modelled more realistically. For the testbed on average 1.6 RoIs per event are assumed and data is requested from ROBs chosen randomly. This should be straightforward to implement in the model. For the full system the mapping of the detector onto the ROBs, the LVL1 trigger menu and the details of the LVL2 processing have to be taken into account. J. Vermeulen will provide details on the implementation of event generation in simdaq with respect to picking a RoI position and determining which ROBIns do need to receive RoI requests. D. Francis remarked that the reviewers feel that distibutions for the fragment sizes should be used in the models. J. Vermeulen will provide the distributions generated by R. Scholte to P. Golonka and K. Korcyl and the simdaq code for handling these distributions.
3. Simdaq: J. Vermeulen referred to he distribution for the number of threads simultanesoulsy active per L2PU (with a single processor) which was presented in the March Trigger/DAQ week (generated with simdaq using the paper model parameterization). For the case studied for less than 0.1% of the events a third thread is needed.
4. Paper modelling: J. Vermeulen mentioned that the paper model results presented in the draft TDR are about final. M.Abolins is investigating the correctness of the assumption that the LVL1 RoI rate for each RoI position does only depend on the area in eta-phi space contributing to the rate for that position and not on the absolute value of eta and phi. He has tried to get relevant information from the LVL1 and PESA groups and from D0 without success. However, the work of the CTEQ ("The Coordinated Theoretical-Experimental Project on QCD") collaboration may provide clues. Distributions for jets of dN/dpT for different eta provide support for the assumption, but some further work is needed.
5. Workplan: K. Korcyl will attend the testbed meeting of Friday 4 April and present a status report in the DataFlow meeting of Monday 7 April.
6. Next meeting: The next telephone meeting is planned for Tuesday 1 April, 15.30 h CERN time.
7. AOB: None
J. Vermeulen, 10 April 2003