BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230831T095746Z
LOCATION:Flüela
DTSTART;TZID=Europe/Stockholm:20230627T173000
DTEND;TZID=Europe/Stockholm:20230627T180000
UID:submissions.pasc-conference.org_PASC23_sess129_msa144@linklings.com
SUMMARY:Impressions from the LUMI Porting Program
DESCRIPTION:Minisymposium\n\nPeter Larsson (KTH Royal Institute of Technol
 ogy, PDC Center for High Performance Computing)\n\nThe LUMI porting progra
 m was created to help scientists and application developers port applicati
 ons to LUMI and AMD GPUs. Six applications were selected in different scie
 ntific domains: two from high energy physics (tmLQCD and PLEGMA), two from
  quantum chemistry/materials science (GPAW and Quantum Espresso), one hydr
 odynamic model (SLIM), and one machine learning project using the Megatron
  language model. This talk highlights some outcomes, experiences and lesso
 ns learned from the first round of the program. The actual porting of e.g.
  CUDA code to HIP turned out not to be the major problem facing most proje
 cts, instead the difficulties were related to getting access to the target
  GPUs (AMD's MI250x GPUs), immaturity of compilers and development tools, 
 and source code organization issues, i.e. how to adapt the code base to us
 e two different GPU backends.\n\nDomain: Computer Science, Machine Learnin
 g, and Applied Mathematics 
\n\nSession Chair: Aniello Esposito (HPE)
END:VEVENT
END:VCALENDAR