BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211207T055404Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T080000
DTEND;TZID=America/Chicago:20211114T170000
UID:submissions.supercomputing.org_SC21_sess188_tut138@linklings.com
SUMMARY:Efficient Distributed GPU Programming for Exascale
DESCRIPTION:Tutorial\n\nEfficient Distributed GPU Programming for Exascale
 \n\nHerten, Garcia de Gonzalo, Kraus, Hrywniak, Oden\n\nOver the past year
 s, GPUs became ubiquitous in HPC installations around the world. Today, th
 ey provide the majority of performance of some of the largest supercompute
 rs (e.g. Summit, Sierra, JUWELS Booster). This trend continues in upcoming
  pre-exascale and exascale systems (LUMI, Leonardo; Frontier): GPUs are ch
 osen as the core computing devices to enter this next era of HPC.\n\nTo ta
 ke advantage of future GPU-accelerated systems with tens of thousands of d
 evices, application developers need to have the proper skills and tools to
  understand, manage, and optimize distributed GPU applications.\n\nIn this
  tutorial, participants will learn techniques to efficiently program large
 -scale multi-GPU systems. While programming multiple GPUs with MPI is expl
 ained in detail, also advanced techniques and models (NCCL, NVSHMEM, …) ar
 e presented. Tools for analysis are used to motivate implementation of  pe
 rformance optimizations. The tutorial combines lectures and hands-on exerc
 ises, using Europe’s fastest supercomputer, JUWELS Booster with NVIDIA A10
 0 GPUs.\n\nTag: Online Only, Accelerator-based Architectures\n\nRegistrati
 on Category: Tutorial Reg Pass
END:VEVENT
END:VCALENDAR