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:20260202T201805Z
LOCATION:260
DTSTART;TZID=America/Chicago:20251116T110000
DTEND;TZID=America/Chicago:20251116T113000
UID:submissions.supercomputing.org_SC25_sess206_ws_exampi108@linklings.com
SUMMARY:Accelerating Intra-Node GPU Communication: A Performance Model for
  Multi-Path Transfers
DESCRIPTION:Amirhossein Sojoodi (Queens University) and Mohammad Akbari, H
 amed Sharifian, Ali Farazdaghi, Ryan E. Grant, and Ahmad Afsahi (Queen's U
 niversity)\n\nOptimizing GPU-to-GPU communication is a key challenge for i
 mproving performance in MPI-based HPC applications, especially when utiliz
 ing multiple communication paths. This paper presents a novel performance 
 model for intra-node multi-path GPU communication within the MPI+UCX frame
 work, aimed at determining the optimal configuration for distributing a si
 ngle P2P communication across multiple paths. By considering factors such 
 as link bandwidth, pipeline overhead, and stream synchronization, the mode
 l identifies an efficient path distribution strategy, reducing communicati
 on overhead and maximizing throughput. Through extensive experiments on va
 rious topologies, we demonstrate that our model accurately finds theoretic
 ally optimal configurations, achieving significant improvements in perform
 ance, with the average of less than 6\% error in predicting the optimal co
 nfiguration for very large messages.\n\nRecording: Livestreamed, Recorded\
 n\nRegistration Category: Technical Program Reg Pass, Workshop Reg Pass\n\
 nSession Chairs: Matthew G. F. Dosanjh (Sandia National Laboratories); Wil
 liam Schonbein (Sandia National Laboratories); Amanda J. Bienz (University
  of New Mexico); and Joseph Schuchart (Stony Brook University, Institute f
 or Advanced Computational Science (IACS))\n\n
END:VEVENT
END:VCALENDAR