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:D221
DTSTART;TZID=America/Chicago:20181113T114500
DTEND;TZID=America/Chicago:20181113T120000
UID:submissions.supercomputing.org_SC18_sess277_drs115@linklings.com
SUMMARY:Scalable Non-Blocking Krylov Solvers for Extreme-Scale Computing
DESCRIPTION:Doctoral Showcase\nComputational Biology, Exascale, GPUs, Grap
 h Algorithms, Linear Algebra, Machine Learning, Sparse Computation, Worksh
 op Reg Pass, Tutorial Reg Pass, Tech Program Reg Pass, Exhibits Reg Pass, 
 Exhibits - Exhibit Hall Only Reg Pass, Doctoral Showcase\n\nScalable Non-B
 locking Krylov Solvers for Extreme-Scale Computing\n\nEller, Gropp\n\nThis
  study investigates preconditioned conjugate gradient method variations de
 signed to reduce communication costs by decreasing the number of allreduce
 s and overlapping communication with computation using a non-blocking allr
 educe. Experiments show scalable PCG methods can outperform standard PCG a
 t scale and demonstrate the robustness of these methods.<br /><br />To dev
 elop the most optimal Krylov methods we need a clear understanding of the 
 factors limiting performance at scale. Detailed timings and network counte
 rs are used to more thoroughly measure the performance of these methods. P
 erformance models with penalty terms are developed that provide reasonable
  explanations of observed performance and guide development of optimizatio
 ns. The effectiveness of scalable PCG methods and these performance analys
 is tools is demonstrated using Quda and Nek5000, two HPC applications seek
 ing improved performance at scale.
URL:https://sc18.supercomputing.org/presentation/?id=drs115&sess=sess277
END:VEVENT
END:VCALENDAR

