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:20210916T132451Z
LOCATION:Michel Mayor
DTSTART;TZID=Europe/Stockholm:20210707T123000
DTEND;TZID=Europe/Stockholm:20210707T130000
UID:submissions.pasc-conference.org_PASC21_sess142_msa142@linklings.com
SUMMARY:Readying HemeLB and SCEMa Codes for Exascale with POP and E-CAM Ce
 ntres of Excellence
DESCRIPTION:Minisymposium\n\nReadying HemeLB and SCEMa Codes for Exascale 
 with POP and E-CAM Centres of Excellence\n\nCoveney, Patronis, McCullough,
  Sinclair, Vassaux...\n\nAs supercomputers approach exascale performance i
 t is necessary to ensure that current simulation codes can continue to eff
 iciently generate results on the latest machines. This presentation will o
 utline how collaborations between the Centre for Computational Science (UC
 L) and the POP and E-CAM CoEs have improved performance characteristics on
  current generation Tier-0 HPC facilities. POP assessments of the scalabil
 ity of the CompBioMed CoE [www.compbiomed.eu] flagship code HemeLB have sh
 own excellent strong scaling performance beyond 300,000 cores on BlueWater
 s (NCSA) and SuperMUC-NG (LRZ). However, load balance efficiency can be im
 proved for larger systems, and this is being addressed in collaboration wi
 th the E-CAM CoE [www.e-cam2020.eu] which is developing a load-balancing l
 ibrary suited to application requirements. Furthermore, POP has benchmarke
 d the multi-model workflow implemented in SCEMa for multiscale prediction 
 of material properties. They have highlighted the strong scaling capabilit
 ies of the code, but also demonstrated a low overall efficiency. They have
  identified bottlenecks appearing in the execution of individual MD simula
 tions using LAMMPS, caused by communication inefficiencies on large alloca
 tions. Recommendations from POP for improving SCEMa’s efficiency inc
 lude reducing the minimum allocation possible for an MD simulation, and co
 mbining MPI and OpenMP to further improve load balance.\n\nDomain: CS and 
 Math, Emerging Applications, Chemistry and Materials, Climate and Weather,
  Physics, Solid Earth Dynamics, Life Sciences, Engineering
END:VEVENT
END:VCALENDAR