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:20190719T085744Z
LOCATION:HG D 1.2
DTSTART;TZID=Europe/Stockholm:20190613T124500
DTEND;TZID=Europe/Stockholm:20190613T131500
UID:submissions.pasc-conference.org_PASC19_sess167_msa142@linklings.com
SUMMARY:A New Ocean Barotropic Solver for the High-Resolution NEMO Model
DESCRIPTION:Minisymposium\nClimate and Weather\n\nA New Ocean Barotropic S
 olver for the High-Resolution NEMO Model\n\nYang, Song, Zhou\n\nThe Ocean 
 general circulation models have developed towards higher spatial resolutio
 n and higher temporal resolution, which need more computational resources.
  As the processors increased, the barotropic solver in Nucleus for Europea
 n Modelling of the Ocean (NEMO) becomes the bottleneck on large scale clus
 ters. The most time-consuming part is the solutions of elliptic equation, 
 one of which is Preconditioned Conjugate Gradient (PCG) method in NEMO. Fi
 rstly, we demonstrate that the communication cost in barotropic solver is 
 the key bottleneck for scalability. In order to solve this problem, we imp
 lemented the Communication-Avoiding Krylov Subspace Methods to PCG method 
 (called CA-PCG). For CA-PCG, every 8 iterations need inner product operati
 ons, and the global communication cost decreases a lot and then the high c
 ore counts scalability is improved. We find out that the new solver can im
 prove the scalability effectively. Moreover, we also prove that the ocean 
 outputs using CA-PCG are consistent with that using the original PCG via a
 n ensemble-based statistical method.
END:VEVENT
END:VCALENDAR