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:20181221T160742Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181113T171500
DTEND;TZID=America/Chicago:20181113T190000
UID:submissions.supercomputing.org_SC18_sess343_drs115@linklings.com
SUMMARY:Scalable Non-Blocking Krylov Solvers for Extreme-Scale Computing
DESCRIPTION:Doctoral Showcase\nWorkshop Reg Pass, Tutorial Reg Pass, Tech 
 Program Reg Pass, Exhibits Reg Pass, Exhibits - Exhibit Hall Only Reg Pass
 \n\nScalable Non-Blocking Krylov Solvers for Extreme-Scale Computing\n\nEl
 ler, Gropp\n\nThis study investigates preconditioned conjugate gradient me
 thod variations designed to reduce communication costs by decreasing the n
 umber of allreduces and overlapping communication with computation using a
  non-blocking allreduce. Experiments show scalable PCG methods can outperf
 orm standard PCG at scale and demonstrate the robustness of these methods.
 <br /><br />To develop the most optimal Krylov methods we need a clear und
 erstanding of the factors limiting performance at scale. Detailed timings 
 and network counters are used to more thoroughly measure the performance o
 f these methods. Performance models with penalty terms are developed that 
 provide reasonable explanations of observed performance and guide developm
 ent of optimizations. The effectiveness of scalable PCG methods and these 
 performance analysis tools is demonstrated using Quda and Nek5000, two HPC
  applications seeking improved performance at scale.
URL:https://sc18.supercomputing.org/presentation/?id=drs115&sess=sess343
END:VEVENT
END:VCALENDAR