Beginner's GPU programming course for UK particle physicists

Europe/London
Conference Room 21 in building R1 (RAL)

Conference Room 21 in building R1

RAL

Description

This course targets UK particle physicists  with no GPU programming experience, but some experience with C++. PhD students, postdocs & staff are all welcome. 

You should bring a laptop that can connect to eduroam Wifi https://eduroam.org/about/connect-yourself/
As a backup in case of problems with eduroam, you should also register now via https://visitorwifi.stfc.ac.uk to use RAL's STFC-Guest Wifi , specifying "Tomalin" (ian.tomalin@stfc.ac.uk) as your "Sponsor".

The course offers a mix of lectures and practical sessions. We have capacity for 20 students. It runs from 12-17h on 29th + 10-14h on 30th May.

Topics covered:

  • Accelerated computing with CUDA C/C++ (8 hours) (NVidia-led course, certificate available on completion)
  • Alternative programming models (HIP, SYCL) and portability frameworks (Kokkos, Alpaka) (1 hour)
  • GPU implementations in HEP experiments (1 hour)

 

The teaching will be provided by Jony Castagna from Hartree Centre, who is the STFC NVidia partner.

RAL/PPD will book you in and pay for one nights bed & breakfast for those who need it, at Cosener's House Hotel, which is in Abingdon. The tuition is free. Attendees must pay their own travel & subsistance, though RAL people organise transport between Cosener's & RAL for those who need it.

All participants should register via indico. The school will be oversubscribed, so check if your registration is approved. When registering, you'll be asked about your accomodation needs etc.

The taxi departure times are:

  • RAL building R71 entrance to Cosener's House at 17:00 on 29th May.
  • Cosener's House to RAL at 09:00 on 30th May.

    The taxi company has the mobile phone number of course participant Mark Grimes.
     

The teaching will take place on the RAL site in Conference Room 21, which is close to the Particle Physics Dept. in building R1. A map showing you the inside of the RAL site is map_inside_RAL .

 A map & advice to help you travel to RAL is map_to_RAL . If you arrive by train, "Didcot Parkway" train station is most convenient. We'll arrange taxis from RAL back to Didcot station at the end of the course.

 

 

Participants
  • Eimear Conroy
  • Elliot Watton
  • Gavin Hesketh
  • Jakub Gajownik
  • Jony Castagna
  • Joseph Aziz
  • Jyoti Prakash Biswal
  • Komninos John Plows
  • Mark Grimes
  • Marvin Pfaff
  • Natalia Zubova
  • Niladri Sahoo
  • Santiago Paredes Saenz
  • Tailin Zhu
  • Tom Pochciol
  • Yajing Wei
  • Yiwen Yang
  • Zhenxiong Xie
    • Day 1: Day 1 (with lunch at 12:30 & coffee at 15:00)
      • 1
        Introduction
        Speaker: Jony Castagna (UKRI-STFC Hartree Centre)
      • 12:30
        Lunch
      • 2
        Accelerating Applications with CUDA C/C++
        Speaker: Jony Castagna (UKRI-STFC Hartree Centre)
      • 15:00
        Coffee
      • 3
        Managing Accelerated Application Memory with CUDA C/C++ Unified Memory
        Speaker: Jony Castagna (UKRI-STFC Hartree Centre)
    • Day 2: Day 2 ( (with coffee at 10:15 & lunch at 12:30)
      • 10:15
        Coffee
      • 4
        Asynchronous Streaming, and Visual Profiling with CUDA C/C++
        Speaker: Jony Castagna (UKRI-STFC Hartree Centre)
      • 5
        N-body simulator
        Speaker: Jony Castagna (UKRI-STFC Hartree Centre)
      • 12:30
        Lunch
      • 6
        Other GPU frameworks (SYCL, Alpaka, HIP)
        Speaker: Stewart Martin-Haugh (STFC)
      • 7
        CMS HLT GPU implementation
        Speaker: Thomas Reis (STFC)