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:20210916T132451Z
LOCATION:Ernesto Bertarelli
DTSTART;TZID=Europe/Stockholm:20210707T143000
DTEND;TZID=Europe/Stockholm:20210707T150000
UID:submissions.pasc-conference.org_PASC21_sess144_msa104@linklings.com
SUMMARY:Performance Portability of the Albany Multi-Physics Finite Element
  Code on the Road to Exascale
DESCRIPTION:Minisymposium\n\nPerformance Portability of the Albany Multi-P
 hysics Finite Element Code on the Road to Exascale\n\nTezaur, Watkins\n\nA
 s HPC architectures become more heterogeneous, climate codes must adapt to
  take advantage of potential performance capabilities. This talk focuses o
 n attaining performance-portability of the Sandia Albany multi-physics C++
  finite element code. We will discuss our efforts in transitioning the fin
 ite element assembly in Albany from an MPI-only to an MPI+X programming mo
 del via the Kokkos library and programming model. In this model, MPI is us
 ed for internode parallelism and X denotes a shared memory programming mod
 el for intranode parallelism (e.g., X=OpenMP, CUDA). With Kokkos data layo
 ut abstractions, the same code can run correctly and efficiently on curren
 t and future HPC hardware with different memory models. We will also share
  some perspectives towards performance portability of the Trilinos-based l
 inear solvers used within Albany, which are relevant for implicit problems
  implemented within the code.  We will present some performance-portabilit
 y results for the land ice modeling application implemented within Albany 
 and known as ALI.  More specifically, we will demonstrate that the same co
 de runs correctly and efficiently, with reasonable scalability, across a v
 ariety of computer architectures including NVIDIA P100 and V100 GPUs.  <br
  />[1] A. Salinger, <em>et al. </em>"Albany: Using Agile Components to Dev
 elop a Flexible, Generic Multiphysics Analysis Code", <em>Int. J. Multisca
 le Comput. Engng</em> 14(4) (2016) 415-438. <br />[2] J. Watkins, I. Tezau
 r, I. Demeshko.  "A study on the performance portability of the finite ele
 ment assembly process within the Albany land ice solver", E. van Brummelen
 , A. Corsini, S. Perotto, G. Rozza, eds.  <em>Numerical Methods for Flows:
  FEF 2017 Selected Contributions</em>, Elsevier, 2019.\n\nDomain: CS and M
 ath, Emerging Applications, Physics, Engineering
END:VEVENT
END:VCALENDAR