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:20260202T201803Z
LOCATION:260
DTSTART;TZID=America/Chicago:20251116T120000
DTEND;TZID=America/Chicago:20251116T123000
UID:submissions.supercomputing.org_SC25_sess206_ws_exampi107@linklings.com
SUMMARY:MPI Collectives with Programmable Smart Switches
DESCRIPTION:Thomas Erbesdobler (Technical University of Munich) and Amir R
 aoofy, Ehab Saleh, and Josef Weidendorfer (Leibniz Supercomputing Centre (
 LRZ))\n\nProgrammable smart network devices are heavily used by cloud prov
 iders, but typically not for HPC. However, they provide opportunities for 
 off-loading computations, in particular for collective operations, which a
 re important for data intensive workloads in classic HPC and ML training. 
 In this paper, we present a prototype called mpitofino to enable offloadin
 g MPI collectives (in particular reductions) onto smart switches over an E
 thernet fabric. We target Intel’s programmable Ethernet switches equipped 
 with a Tofino ASIC, and we use the P4 programming language to process coll
 ective packets on the chip’s low-latency data path. We demonstrate how the
  flexibility of P4 enables us to use RoCEv2 as protocol, utilizing RDMA ha
 rdware support on the nodes’ NICs. Furthermore, we implement mpitofino as 
 a collective provider in Open MPI and discuss its desirable scaling charac
 teristics. Finally, we demonstrate that mpitofino can achieve data through
 put close to the 100GBit/s line rate.\n\nRecording: Livestreamed, Recorded
 \n\nRegistration Category: Technical Program Reg Pass, Workshop Reg Pass\n
 \nSession Chairs: Matthew G. F. Dosanjh (Sandia National Laboratories); Wi
 lliam Schonbein (Sandia National Laboratories); Amanda J. Bienz (Universit
 y of New Mexico); and Joseph Schuchart (Stony Brook University, Institute 
 for Advanced Computational Science (IACS))\n\n
END:VEVENT
END:VCALENDAR