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:20260522T150122Z
LOCATION:D165
DTSTART;TZID=America/Chicago:20181112T163000
DTEND;TZID=America/Chicago:20181112T170000
UID:submissions.supercomputing.org_SC18_sess161_ws_pmbsf120@linklings.com
SUMMARY:Evaluating the Impact of Spiking Neural Network Traffic on Extreme
 -Scale Hybrid Systems
DESCRIPTION:Noah Wolfe and Mark Plagge (Rensselaer Polytechnic Institute),
  Misbah Mubarak (Argonne National Laboratory), Christopher D. Carothers (R
 ensselaer Polytechnic Institute), and Robert B. Ross (Argonne National Lab
 oratory)\n\nAs we approach the limits of Moore's law, there is increasing 
 interest in non-Von Neuman architectures such as neuromorphic computing to
  take advantage of improved compute and low power capabilities. Spiking ne
 ural network (SNN) applications have so far shown very promising results r
 unning on a number of processors, motivating the desire to scale to even l
 arger systems having hundreds and even thousands of neuromorphic processor
 s. Since these architectures currently do not exist in large configuration
 s, we use simulation to scale real neuromorphic applications running on a 
 single neuromorphic chip, to thousands of chips in an HPC class system. Fu
 rthermore, we use a novel simulation workflow to perform a full scale syst
 ems analysis of network performance and the interaction of neuromorphic wo
 rkloads with traditional CPU workloads in a hybrid supercomputer environme
 nt. On average, we find Slim Fly, Fat-Tree, Dragonfly-1D, and Dragonfly-2D
  are 45%, 46%, 76%, and 83% respectively faster than the worst case perfor
 ming topology for both convolutional and Hopfield NN workloads running alo
 ngside CPU workloads. Running in parallel with CPU workloads translates to
  an average slowdown of 21% for a Hopfield type workload and 184% for conv
 olutional NN workloads across all HPC network topologies.\n\nTag: Benchmar
 ks, Parallel Programming Languages, Libraries, and Models, Performance, Si
 mulation\n\nRegistration Category: Workshop Reg Pass\n\nSession Chair: Ste
 ven A. Wright (University of York, England)\n\n
END:VEVENT
END:VCALENDAR