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:20210916T132449Z
LOCATION:Ernesto Bertarelli
DTSTART;TZID=Europe/Stockholm:20210706T123000
DTEND;TZID=Europe/Stockholm:20210706T130000
UID:submissions.pasc-conference.org_PASC21_sess128_msa205@linklings.com
SUMMARY:Spectral Element Based Flow Simulations on the SX-Aurora TSUBASA
DESCRIPTION:Minisymposium\n\nSpectral Element Based Flow Simulations on th
 e SX-Aurora TSUBASA\n\nJansson\n\nFollowing the recent transition in the h
 igh-performance computing landscape to more heterogeneous architectures, a
 pplication developers are faced with the challenge of ensuring good perfor
 mance across a diverse set of platforms. We present our work on porting a 
 high-order spectral element based incompressible flow solver to the recent
  vector architecture SX-Aurora TSUBASA. Using the mini-app Nekbone, we for
 mulate suitable loop transformations of key matrix-vector and matrix-matri
 x multiplication kernels, allowing for better vectorisation, achieving clo
 se to half the peak performance of a single SX-Aurora core. The mini-app a
 lso served as the basis to formulate a new, nearly twice as fast, implemen
 tation of Nek5000's gather-scatter library with mesh topology awareness fo
 r improved vectorisation via exploitation of the SX-Aurora's hardware gath
 er-scatter instructions. Based on the experience gained from the mini-app,
  a detailed description of the integration of the tuned kernels and optimi
 sed gather-scatter routines into the full flow solver is given together wi
 th a performance study, comparing both single node performance and strong 
 scalability characteristics of turbulent flow simulations, running across 
 multiple SX-Aurora cards.\n\nDomain: CS and Math, Physics, Engineering
END:VEVENT
END:VCALENDAR
