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:20181221T160728Z
LOCATION:D161
DTSTART;TZID=America/Chicago:20181112T115000
DTEND;TZID=America/Chicago:20181112T121000
UID:submissions.supercomputing.org_SC18_sess158_ws_lasalss105@linklings.co
 m
SUMMARY:Communication Avoiding Multigrid Preconditioned Conjugate Gradient
  Method for Extreme Scale Multiphase CFD Simulations
DESCRIPTION:Workshop\nAlgorithms, Heterogeneous Systems, Resiliency, Works
 hop Reg Pass\n\nCommunication Avoiding Multigrid Preconditioned Conjugate 
 Gradient Method for Extreme Scale Multiphase CFD Simulations\n\nIdomura, I
 na, Yamashita, Onodera, Yamada...\n\nA communication avoiding (CA) multigr
 id preconditioned conjugate gradient method  (CAMGCG) is applied to the pr
 essure Poisson equation in a multiphase CFD code JUPITER, and its computat
 ional performance and convergence property are compared against CA Krylov 
 methods. A new geometric multigrid preconditioner is developed using a pre
 conditioned Chebyshev iteration smoother, in which no global reduction com
 munication is needed, halo data communication is reduced by a mixed precis
 ion approach, and eigenvalues are computed using the CA Lanczos method. Th
 e CAMGCG solver has robust convergence properties regardless of the proble
 m size, and shows both communication reduction and convergence improvement
 , leading to higher performance gain than CA Krylov solvers, which achieve
  only the former. The CAMGCG solver is applied to extreme scale multiphase
  CFD simulations with ~90 billion DOFs, and it is shown that compared with
  a preconditioned CG solver, the number of iterations, and thus, All\_Redu
 ce is reduced to ~1/800, and ~11.6x speedup is achieved with keeping excel
 lent strong scaling up to 8,000 KNLs on the Oakforest-PACS.
URL:https://sc18.supercomputing.org/presentation/?id=ws_lasalss105&sess=se
 ss158
END:VEVENT
END:VCALENDAR