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

    EP R&D Software Working Group Meeting

    → Europe/Zurich
    The Zoomiverse

    The Zoomiverse
    Live Notes
    EP R&D Software Working Group Meeting
    Zoom Meeting ID
    62893115044
    Host
    Graeme A Stewart
    Alternative host
    Andre Sailer
    Useful links
    Join via phone
    Zoom URL
    Hide
     

    Software R&D Working Meeting Minutes

    Apologies: Marco Rovere

    Introduction

    We shall contact Christian to offer a tracking talk for an up-coming EP R&D seminar (after the summer).

    CHEP2022 will have a more traditional abstract-only submission in advance of the conference.

    GSoC:

    • Unfortunately thought there were some proposals we got no good GSoC candidates this year
    • Overall programme seems to have been less successful this year (Google changes + COVID?)

    Summer Students:

    • We have a tracking student

    Tracking

    Plot on slide 5, alignments are all pre-calculated. The plot shows the fraction of wallclock time when certain numbers of threads are active. Main take-away is that events in multiple aligment IOVs can be processed very effectively in parallel.

    VecMem project - is this like Llama in way (from CASUS). Do you abstract the data structure? Idea is a container that looks the same on the host and device; will also manage allocations in the background. This was spun out of the detray detector description, which had contunguous containers of different objects, with links - already gave a massive speed-up on CPU. On the GPU had to not use STL! Connection with PODIO? Abstract description that then writes out code.

    Detray description abstracts geometry from underlying implementation. There is a prototype VecGeom backend. There’s a lot of logic to translate this to something that is indexable. Would be good to discuss this with those working on geometry for simulation and VecGeom enhancements.

    Gaudi integration? Other people do take care of that and there are enough Gaudi users to pick up on problems very quickly. There is a very lightweight farmework used in ACTS developers. Avoid the problems that Geant4 ran into with their examples. Deployment to the experiments is key to covering the use cases - e.g., the successful sPhenix integration.

    There are minutes attached to this event. Show them.