Go to the Indico Home Page Oldest event Older event Up to category Newer event Newest event Printable version
  • Compact style
  • Indico style
  • Indico style - inline minutes
  • Indico style - numbered
  • Indico style - numbered + minutes
  • Indico Weeks View
    • Switch to the management area of this event
    Login

    Developer's meeting

    → Europe/Zurich
    Hide

    Giovanna (in absentia), Enrico, Roland: updates at EHN1

    • the FLX-711 card has been connected to the patch panel
    • plan is to connect the WIB to the patch panel tomorrow and test the links
    • then, run the actual DAQ
      • should be able to use WIB internal generation

    Other updates: Overlay class implementation by Milo is now working

    • implementation w/o compression
    • to be communicated to Antonino and tested with WIB data soon

    Schedule:

    • 5 WIBs arriving soon; to be used for Cold-box testing
      • expect to be able to hook up FELIX to these, even if only for shorter periods of time
      • Eric: does this use the same firmware? Dan: yes, just need JTAG to configure for FELIX instead of RCE
    • Frank: order for 3 Mellanox (dual 100 GB/s) cards and 2 corresponding 5m cables going out ~ now; can hopefully be at CERN in ~ 1.5 weeks
    • Roland: two separate BoardReader PCs (Dell PowerEdge R740; 14 cores; high-end Xeon Silver or low-end Xeon Gold) to be ordered
      • expect these to be at CERN in a few weeks
      • possibility to use QAT (and not under non-disclosure agreement anymore)

    Planning (near- and mid-term):

    • start running without compression
    • then use QAT for compression
      • this uses deflate (see https://en.wikipedia.org/wiki/DEFLATE), which combines LZ77 and Huffman coding
      • Frank: what compression factor did this yield in the past? Roland: a factor of ~ 2
        • clearly well below that of Milo's s/w based compression, but on the other hand a lot better than not compressing at all
      • Frank: this shouldn't compress artdaq Fragment headers? Roland: it won't
      • Milo: a lot depends on the ordering of the data. Roland: indeed, the above factor of 2 is w/o any re-ordering.
      • we can try re-ordering the data in s/w and then using QAT for the compression
    • other compression schemes can be looked into later
    • related (potential) Overlay issues:
      • idea is to start using a separate FELIX fragment type, but to converge to the more generic TPC fragment type
      • distinguishing between different compression schemes (and between FELIX and RCE) will still be possible as the relevant metadata is not limited to the fragment type

    Next meeting: tentatively planned for October 18 (we can at least discuss further the progress at EHN1)

    There are minutes attached to this event. Show them.
      • 18:00 → 18:20
        EHN1 progress 20m
        Speakers: Enrico Gamberini (CERN), Roland Sipos (Institut of theoretical physics Eotvos Lorand University (HU))
      • 18:20 → 18:40
        Near-term planning 20m