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:20210916T132447Z
LOCATION:Lise Girardin
DTSTART;TZID=Europe/Stockholm:20210709T120000
DTEND;TZID=Europe/Stockholm:20210709T123000
UID:submissions.pasc-conference.org_PASC21_sess190_pap_dec119@linklings.co
 m
SUMMARY:Massive Scaling of MASSIF: Algorithm Development and Analysis for 
 Simulation on GPUs
DESCRIPTION:Paper\n\nMassive Scaling of MASSIF: Algorithm Development and 
 Analysis for Simulation on GPUs\n\nKulkarni, Kovacevic, Franchetti\n\nMicr
 omechanical Analysis of Stress-Strain Inhomogeneities with Fourier transfo
 rms (MASSIF) is a large-scale Fortran-based differential equation solver u
 sed to study local stresses and strains in materials. Due to its prohibiti
 ve memory requirements, it is extremely difficult to port the code to GPUs
  with small on-chip memory. In this work, we present an algorithm design t
 hat uses domain decomposition with approximate convolution, which reduces 
 memory footprint to make the MASSIF simulation feasible on distributed GPU
  systems. A first-order performance model of our method estimates that com
 pression and multi-resolution sampling strategies can enable domain comput
 ation within GPU memory constraints for $3$D grids larger than those simul
 ated by the current state-of-the-art Fortran MPI implementation. The model
  analysis also provides an insight into design requirements for further sc
 alability. Lastly, we discuss the extension of our method to irregular dom
 ain decomposition and challenges to be tackled in the future.\n\nDomain: C
 S and Math, Physics, Life Sciences
END:VEVENT
END:VCALENDAR