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 F 1
DTSTART;TZID=Europe/Stockholm:20190613T151500
DTEND;TZID=Europe/Stockholm:20190613T154500
UID:submissions.pasc-conference.org_PASC19_sess156_msa249@linklings.com
SUMMARY:Domain Decomposition for On-Node Performance: libCEED and Beyond
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Engin
 eering\n\nDomain Decomposition for On-Node Performance: libCEED and Beyond
 \n\nMarin, Brown, Barra, Thompson, Smith\n\nDomain decomposition strategie
 s originate from the idea of decomposing the domain into smaller subsectio
 ns that can be solved independently, with appropriate boundary conditions,
  and then combined to provide the solution for the original domain. This n
 otion can be pushed to the limit and select these subdomains of small enou
 gh size (for example, a single spectral element) such that the computation
 s can be highly optimized for on-node performance.  Hereby, we explor
 e and advocate for matrix-free solvers in the context of a domain decompos
 ition approach in a framework provided by the PETSc library. A recently de
 veloped library for application of the operator at the subdomain level is 
 provided by libCEED, which can be tuned for high performance on GPUs as we
 ll as on CPUs. LibCEED is aimed at discretizations with dense and/or struc
 tured local representations (such as tensor product bases for spectral ele
 ments), however, the concept of a highly-tuned operator library expressed 
 at the subdomain level is broadly applicable.
END:VEVENT
END:VCALENDAR