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:20260202T201804Z
LOCATION:267
DTSTART;TZID=America/Chicago:20251117T093000
DTEND;TZID=America/Chicago:20251117T100000
UID:submissions.supercomputing.org_SC25_sess198_ws_pmbsf116@linklings.com
SUMMARY:A Peak Performance Model for All-to-all on Hierarchical Systems an
 d Its Applications
DESCRIPTION:Rohini Uma-Vaideswaran (Georgia Institute of Technology), Josh
 ua Romero (NVIDIA Corporation), Daniel Dotson (Georgia Institute of Techno
 logy), David Appelhans (NVIDIA Corporation), and P. K. Yeung (Georgia Inst
 itute of Technology)\n\nAn accurate measure of communication performance i
 s a key component of optimizing large scale high performance computing app
 lications. This paper presents a model for the peak performance of all-to-
 all communication, in the context of systems composed of a hierarchy of in
 terconnect bandwidths; a common trait of multi-GPU per node systems. We de
 monstrate an application of the model to distributed transposes, such as t
 hose encountered in distributed three-dimensional Fast Fourier Transforms.
  The model is validated on three different network architectures, using a 
 variety of communication libraries, by measuring all-to-all and distribute
 d transpose performance in a pseudo-spectral code for direct numerical sim
 ulations of 3D fluid turbulence. Both the model and the validation results
  provide insights into the impact of fast communication links located lowe
 r in the network hierarchy, the expected scaling for all-to-all bound prob
 lems, and performance considerations when selecting slab (1D) or pencil (2
 D) domain decompositions.\n\nRecording: Livestreamed, Recorded\n\nRegistra
 tion Category: Technical Program Reg Pass, Workshop Reg Pass\n\nSession Ch
 airs: Steven A. Wright (University of York, England); Simon Hammond (Natio
 nal Nuclear Security Administration (NNSA)); and Sascha Hunold (Technical 
 University of Vienna)\n\n
END:VEVENT
END:VCALENDAR
