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:20210916T132452Z
LOCATION:Lise Girardin
DTSTART;TZID=Europe/Stockholm:20210707T150000
DTEND;TZID=Europe/Stockholm:20210707T153000
UID:submissions.pasc-conference.org_PASC21_sess149_msa156@linklings.com
SUMMARY:Cornerstone Octree: Space-Filling Curves Meet 3D Collision Detecti
 on in SPH and N-Body Problems
DESCRIPTION:Minisymposium\n\nCornerstone Octree: Space-Filling Curves Meet
  3D Collision Detection in SPH and N-Body Problems\n\nKeller\n\nOctrees ar
 e an important data structure for various kinds of hierarchical particle-b
 ased simulations, including smoothed particle hydrodynamics in an astrophy
 sical context. Together with space-filling curve (SFC) algorithms, to whic
 h they are intimately related, octrees form the basis of domain decomposit
 ion and neighbor searching in SPH and also constitute the backbone of Barn
 es-Hut tree codes and FMM with dual-tree traversal if self-gravity is incl
 uded in addition to hydrodynamics. Their naturally recursive nature poses 
 a challenge for efficient parallelization and retaining data locality requ
 ired for benefiting from the memory hierarchy. We present how this challen
 ge can be addressed by combining traditional SFC techniques with algorithm
 s for collision detection in 3D graphics. While the resulting octree const
 ruction method improves efficiency by breaking level-by-level sequentialit
 y, the resulting program code is also shorter and simpler by expressing al
 l operations either as plain "parallel for" loops with independent element
 s or as common parallel primitives, such as radix sorting and prefix sums.
  The presented method is therefore easy to implement and maintain on diffe
 rent architectures, including GPUs.\n\nDomain: CS and Math, Physics
END:VEVENT
END:VCALENDAR