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:20181221T160728Z
LOCATION:D173
DTSTART;TZID=America/Chicago:20181112T153000
DTEND;TZID=America/Chicago:20181112T155000
UID:submissions.supercomputing.org_SC18_sess175_ws_opensu102@linklings.com
SUMMARY:Driving Asynchronous Distributed Tasks with Events
DESCRIPTION:Workshop\nArchitectures, Collaborative Environments, Parallel 
 Programming Languages, Libraries, and Models, Simulation, Workflows, Works
 hop Reg Pass\n\nDriving Asynchronous Distributed Tasks with Events\n\nBrow
 n\n\nOpen-source matters, not just to the current cohort of HPC users, but
  also to potential new HPC communities, such as machine learning, themselv
 es often rooted in open-source. Many of these potential new workloads are,
  by their very nature, far more asynchronous and unpredictable than tradit
 ional HPC codes and open-source solutions must be found to enable new comm
 unities of developers to easily take advantage of large scale parallel mac
 hines. Task-based models have the potential to help here, but many of thes
 e either entirely abstract the user from the distributed nature of their c
 ode, placing emphasis on the runtime to make important decisions concernin
 g scheduling and locality, or require the programmer to explicitly combine
  their task-based code with a distributed memory technology such as MPI, w
 hich adds considerable complexity. In this paper we describe a new approac
 h where the programmer still splits their code up into distinct tasks, but
  is explicitly aware of the distributed nature of the machine and drives i
 nteractions between tasks via events. This provides the best of both world
 s; the programmer is able to direct important aspects of parallelism whils
 t still being abstracted from the low level mechanism of how this parallel
 ism is achieved. We demonstrate our approach via two use-cases, the Graph5
 00 BFS benchmark and in-situ data analytics of MONC, an atmospheric model.
  For both applications, we demonstrate considerably improved performance a
 t large core counts and the result of this work is an approach and open-so
 urce library which is readily applicable to a wide range of codes.
URL:https://sc18.supercomputing.org/presentation/?id=ws_opensu102&sess=ses
 s175
END:VEVENT
END:VCALENDAR