Notes from L2 meeting of 21-OCT-97: Dan Edmunds, Philippe Laurens, Jim linnemann, Roger Moore at MSU and Dan Owen at FNAL. UofM: Philippe hasn't heard from Myron/Jerry since last week. Drew's reading list: Drew can get a copy of the VME standard from VITA (Call 602-951-8866 or e-mail info@vita.com for pricing information). http://www.vita.com/ has some draft documents online. Reading out a different list of (i.e. all) geographic sections for MFP events: We can allocate one sptrg for MFP programmed to readout all geo sect, but this method would have no control over what other specific triggers fired for the event flagged as MFP. We'd need to allocate not one specific trigger but 128 new specific triggers to fully control the mix of specific triggers with this method. Run I Level 2 MFP fraction was a property of a L2 bit. Run I L1 FW was always reading out all inputs and outputs. Run I L1.5 was property of whole system. A guiding principle for Run II is to bring the least perturbance to the system for MFP events so that we actually test normal operation, and not some special handling just for MFP. We were planning to just add an event level state flag. There is still (or seems to be) a need to control (or influence) the normal mix of specific triggers in the event. A number of approaches are theoretically possible, from one Spec Trig bit reserved for MFP events with no mix control, to a number of independent groups of Specific triggers, all the way to individual Specific Trigger Control. It is unclear at the moment what is most preferable and what is achievable. The L2 group (Jim) should research what was done in Run I, why and with what success, and come up with an understaning of what functionality is needed for Run II. Note that the relevant pieces of the L1FW are being finalized *now* by the L1 group. MBT and GL2: The idea of merging the MBT card and GL2 card only makes sense if there is an alpha processor board with Magic Bus in the crate. HP G-Link: Dan reports that marvin is having serious design problem with using multiple HP G-link at the same time. one alone works, 2 shows problems, 3 worse, and 4 not at all, while Marvin needs to put 64 in each crate. Preshower Tirgger: Jim reports on the commitment from Paul Grannis and Stony Brook to work on preshower trigger. But serious involvment will not start for about 3 months. Muon Readout: Dan proposes a modified muon architecture to handle readout to L2 and L3. If one adds an MBus port to the SLIC modules, then the SLIC modules can receive an event, send it to the CNAPS processors, retrieve the answers, then blow data out on MBUS to an alpha processor without needing for further buffering on SLIC. The alpha processor can reformat/reorder the data and send it out to L2 global via MBUS and an MBT, then wait for the L2FW decision to trash the event or send the data to L3 via the VME bus and a VBD. More work for TCC: If the Run II Trigger ends up with more alpha processors besides the L2 GLobal and L2 CT PP, it seems convenient and natural that TCC provide a similar download service for executable and run parameters to these new systems too, and retrieve monitoring information from the VME crate to serve back to a host monitoring program. Simulators: Jim reports that Dr Pope and Dr Brock are currently looking for a project and expressed interest in doing some programming for the Run II simulator. Jim reports that DZero considers acceptable that hardware level simulators use FORTRAN, and the MonteCarlo GEANT events are still based on FORTRAN. One could identify some trigger computing that is fixed (i.e. non run dependent) of the L1 CalTrig (and other trigger subsystems, e.g. L2CTPP) and include them in the MC event generation. This is an attractive change to the simulator architecture because simple trigger studies could then be done directly on the MC data without the heavy machinery of L1SIM/L2SIM and TRIGPARSE/COOR. The simulator has 4 basic components: Event read/write (actually provided by the frame, e.g. D0USER), run programming (this will change for Run II), event processing (L1CT stays the same, L1FW will change), and output formatting including data access utilities (this will change for Run II). The only piece worth saving seems to be the L1CT processing, which is probably only 10-20 FORTRAN routines. Philippe will take a look and see what is worth salvaging. Further meetings this week: Oct 23 and 24 L1 Cal simulation Consulting with Amber Goal: TTowers, Etmiss, Scalar Et simulation callable in C++ with MC by March, 1998. Pope and Brock doing most of work; Laurens and Linnemann consulting. On same time scale, the schedule calls for a L1 framework to be designed (and L2 as well) and integration in June 98. Oct 25 More discussion of M+P Possible reasons for wanting a different fraction of M+P for different trigger bits include a desire to check all trigger bits equally well, rather than according to their input rate. This could be done with an overall know, plus L2Global software. Roger and Jim are continuing to study what is really desired. Dan will think about how difficult a hardware solution is, since the software solution is more complex for the triggermeisters, and absorbs L2G time. MBT with SCL, and a proposal for muon architecture Design features reviewed. Dan has developed an attractive architecture for muon assuming that the muon system can be split into 2 crates on geographical or algorithm basis, with no data flow between crates, and assuming that the SLIC cards grow to include DSP's on board. Event building is then across MBus into an alpha, and the rest of the system can look very similar indeed to a 1-worker cal crate (or rather 2 such crates). The Alpha may be able to do some postprocessing of DSP results to make Global's life easier, at the cost of some latency. Questions raised: can the M+P qualifier be passed into the SLIC cards from muon FE or Trigger crates? is there a need for data communication across SLIC cards? sure there is no need to replicate data across crates? eg need 2 copies of trigger data for merging information? Dan will write up the discussion and send it to Fortner.