Present: Graeme S, Andrea V, Andy B, David L, Peter R, Servesh M, Stefan H, Taylor C
Michelangelo has confirmed the TH Conference Room, 26-27 November (except for 14.00-15.30 Tuesday when there is a TH seminar).
Discussed the agenda:
- We agreed to cover software performance optimisations as a topic, this should be a basic introduction into how to exploit modern CPUs successfully. Servesh has good material on how to write code in a way that lets the compiler do the hard work through auto-vectorisation (Graeme noted experiment experience had not been so positive, but maybe we did it wrong!)
- A general introduction to generators, what the codes do and which algorithms are involved will also be really useful, especially as there are many different parts of the process, e.g., matrix elements to hadronisation.
- Combining these two pieces to give a walkthrough of how to structure an algorithm for concurrency would be ideal.
- Servesh could help with that, suggesting to work with Stefan on their O(100) lines of pseudo-code.
- HPCs can be covered as offering particular challenges to exploit efficiently.
- A session on more effective use of the generator codes by the experiments would be very useful.
- We should think about Run 3 as well, and stepping stone improvements - not rewriting everything for HL-LHC.
- Want to cover NNLO calculations - slicing / antennas (Alexander Huss).
- Sherpa and Madgraph the most important generators to have presentations from.
- We don't want to open up for abstracts (time is too short), but put contact information down for suggestions.
- Open question about hackathon session.
Actions:
- Graeme to put up a straw-man agenda and improve indico to the point where registration can be opened.
- Gather suggestions for speakers as well as people to specifically invite in the TODO list google doc.
There are minutes attached to this event.
Show them.
