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:20260522T150116Z
LOCATION:D168
DTSTART;TZID=America/Chicago:20181116T092000
DTEND;TZID=America/Chicago:20181116T093500
UID:submissions.supercomputing.org_SC18_sess144_ws_pawatm112@linklings.com
SUMMARY:Efficient Algorithms for Collective Operations with Notified Commu
 nication in Shared Windows
DESCRIPTION:Muhammed Abdullah Al Ahad (KTH Royal Institute of Technology),
  Christian Simmendinger (T-System Solutions for Research), Roman Iakymchuk
  (KTH Royal Institute of Technology), Tiberiu Rotaru (Fraunhofer ITWM), an
 d Erwin Laure and Stefano Markidis (KTH Royal Institute of Technology)\n\n
 Collective operations are commonly used in various parts of scientific app
 lications. Especially in strong scaling scenarios, collective operations c
 an negatively impact the overall applications performance: while the load 
 per rank decreases with increasing core counts, time spent in e.g. barrier
  operations will increase logarithmically with the core count. \n\nIn this
  article, we develop novel algorithmic solutions for collective operations
  -- such as Allreduce and Allgather(V) -- by leveraging notified communica
 tion in shared windows. To this end, we have developed an extension of GAS
 PI which enables all ranks participating in a shared window to observe the
  entire notified communication targeted at the window.  By exploring benef
 its of this extension, we deliver high performing implementations of Allre
 duce and Allgather(V) on Intel and Cray clusters. These implementations cl
 early achieve 2x-4x performance improvements compared to the best performi
 ng MPI implementations for various data\n\nTag: Parallel Programming Langu
 ages, Libraries, and Models, Productivity\n\nRegistration Category: Worksh
 op Reg Pass\n\n
END:VEVENT
END:VCALENDAR