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:20260522T150117Z
LOCATION:D174
DTSTART;TZID=America/Chicago:20181116T105000
DTEND;TZID=America/Chicago:20181116T111000
UID:submissions.supercomputing.org_SC18_sess146_ws_ftxs114@linklings.com
SUMMARY:Extending and Evaluating Fault-Tolerant Preconditioned Conjugate G
 radient Methods
DESCRIPTION:Carlos Pachajoa, Markus Levonyak, and Wilfried N. Gansterer (U
 niversity of Vienna)\n\nWe compare and refine exact and heuristic fault-to
 lerance extensions for the preconditioned conjugate gradient (PCG) and the
  split preconditioner conjugate gradient (SPCG) methods for recovering fro
 m failures of compute nodes of large-scale parallel computers. In the exac
 t state reconstruction (ESR) approach, which is based on a method proposed
  by Chen (2011), the solver keeps extra information from previous search d
 irections of the (S)PCG solver, so that its state can be fully reconstruct
 ed if a node fails unexpectedly. ESR does not make use of checkpointing or
  external storage for saving dynamic solver data and has only negligible c
 omputation and communication overhead compared to the failure-free situati
 on. In exact arithmetic, the reconstruction is exact, but in finite-precis
 ion computations, the number of iterations until convergence can differ sl
 ightly from the failure-free case due to rounding effects. We perform expe
 riments to investigate the behavior of ESR in floating-point arithmetic an
 d compare it to the heuristic linear interpolation (LI) approach by Langou
  et al. (2007) and Agullo et al. (2016), which does not have to keep extra
  information and thus has lower memory requirements. Our experiments illus
 trate that ESR, on average, has essentially zero overhead in terms of addi
 tional iterations until convergence, whereas the LI approach incurs much l
 arger overheads.\n\nTag: Resiliency, Scientific Computing\n\nRegistration 
 Category: Workshop Reg Pass\n\n
END:VEVENT
END:VCALENDAR