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:20260202T201312Z
LOCATION:260
DTSTART;TZID=America/Chicago:20251116T090000
DTEND;TZID=America/Chicago:20251116T173000
UID:submissions.supercomputing.org_SC25_sess206@linklings.com
SUMMARY:ExaMPI25: Workshop on Extreme Scale MPI
DESCRIPTION:MPI is currently the de facto standard for programming HPC sys
 tems and parallel applications. Development of the MPI standard continues 
 in earnest, with version 4.1 being released within the past year, and feat
 ures for version 5.0 under active discussion. The aim of this workshop is 
 to bring together researchers and developers to present and discuss innova
 tive algorithms and concepts within the Message Passing programming model,
  and to create a forum for open discussions on the future of the Message P
 assing Interface (MPI) in the post-exascale era. Possible workshop topics 
 include and are not limited to algorithms for collective operations, MPI o
 ptimization for artificial intelligence and machine learning workloads, da
 ta-centric models, scheduling, fault tolerance, MPI optimization in hetero
 geneous systems, interoperability of MPI with other programming models (e.
 g., PGAS), integration of task-parallel models in MPI, the role of MPI in 
 "smart" networks, and the use of MPI in large-scale simulations.\n\nPanel:
  The future of MPI in a world driven by AI\n\nJoseph Schuchart; Joseph Sch
 uchart (Stonybrook); Hari Subramoni (Ohio State University); Rich Graham (
 NVIDIA); and Edgar Gabriel (Advanced Micro Devices, Inc. (AMD))\n---------
 ------------\nOpening Remarks\n---------------------\nAfternoon Break - Ex
 aMPI25\n---------------------\nLarge-Message All-to-All Communication at F
 rontier Scale\n\nNear the full scale of exascale supercomputers, latency c
 an dominate the cost of all-to-all communication even for very large messa
 ge sizes. We describe GPU-aware all-to-all implementations designed to red
 uce latency for large message sizes at extreme scales, and we present thei
 r performance using 6...\n\n\nJames B. White III (Oak Ridge National Labor
 atory (ORNL), National Center for Computational Sciences)\n---------------
 ------\nMPI Collectives with Programmable Smart Switches\n\nProgrammable s
 mart network devices are heavily used by cloud providers, but typically no
 t for HPC. However, they provide opportunities for off-loading computation
 s, in particular for collective operations, which are important for data i
 ntensive workloads in classic HPC and ML training. In this paper...\n\n\nT
 homas Erbesdobler (Technical University of Munich) and Amir Raoofy, Ehab S
 aleh, and Josef Weidendorfer (Leibniz Supercomputing Centre (LRZ))\n------
 ---------------\nInvited Talk: Patrick Bridges\n\nPatrick Bridges\n-------
 --------------\nScaling All-to-All Operations Across Emerging Many-Core Su
 percomputers\n\nPerformant all-to-all collective operations in MPI are cri
 tical to fast Fourier transforms, transposition, and machine learning appl
 ications. There are many existing implementations for all-to-all exchanges
  on emerging systems, with the achieved performance dependent on many fact
 ors, including messa...\n\n\nShannon Kinkead (Sandia National Laboratories
 , University of New Mexico); Jackson Wesley (University of New Mexico); Wh
 it Schonbein (Sandia National Laboratories); David DeBonis (Los Alamos Nat
 ional Laboratory (LANL)); Matthew Dosanjh (Sandia National Laboratories); 
 and Amanda Bienz (University of New Mexico)\n---------------------\nMPPI -
  Type safe C++ Datatypes for MPI\n\nMPI provides a flexible C-API to commu
 nicate data of various types between a set of distributed processes over h
 igh-speed interconnects in HPC systems. Data buffers are described using M
 PI-Datatypes, which specify the type and layout of the data to be transmit
 ted. To construct these datatypes, users...\n\n\nMike Söhner and Christoph
  Niethammer (University of Stuttgart, Germany)\n---------------------\nExa
 MPI25: Workshop on Extreme Scale MPI\n\nMPI is currently the de facto stan
 dard for programming HPC systems and parallel applications. Development of
  the MPI standard continues in earnest, with version 4.1 being released wi
 thin the past year, and features for version 5.0 under active discussion. 
 The aim of this workshop is to bring togethe...\n\n\nMatthew G.F. Dosanjh 
 and William Schonbein (Sandia National Laboratories), Amanda Bienz (Univer
 sity of New Mexico), and Joseph Schuchart (Stony Brook University)\n------
 ---------------\nMorning Break - ExaMPI25\n---------------------\nOn the I
 ntegration of Lightweight Tasks with MPI using the C++26 std::execution `S
 enders' API\n\nIntegrating asynchronous MPI messaging with tasking runtime
 s requires careful handling of request polling and dispatching of associat
 ed completions to participating threads. The new C++26 Senders (std::execu
 tion) library offers a flexible collection of interfaces and templates for
  schedulers, algori...\n\n\nJohn Biddiscombe (ETH Zürich / Swiss National 
 Supercomputing Centre (CSCS)); Mikael Simberg, Auriane Reverdell, Raffaele
  Solca, Alberto Invernizzi, and Rocco Meli (Swiss National Supercomputing 
 Centre (CSCS)); and Joseph Schuchart (Innovative Computing Laboratory, Uni
 versity of Tennessee)\n---------------------\nInvited Talk: Ken Raffinetti
 \n\nKen Raffenetti (Argonne National Laboratory (ANL))\n------------------
 ---\nLunch break (on your own)\n---------------------\nAccelerating Intra-
 Node GPU Communication: A Performance Model for Multi-Path Transfers\n\nOp
 timizing GPU-to-GPU communication is a key challenge for improving perform
 ance in MPI-based HPC applications, especially when utilizing multiple com
 munication paths. This paper presents a novel performance model for intra-
 node multi-path GPU communication within the MPI+UCX framework, aimed at d
 et...\n\n\nAmirhossein Sojoodi (Queens University) and Mohammad Akbari, Ha
 med Sharifian, Ali Farazdaghi, Ryan E. Grant, and Ahmad Afsahi (Queen's Un
 iversity)\n\nRecording: Livestreamed, Recorded\n\nRegistration Category: T
 echnical Program Reg Pass, Workshop Reg Pass\n\nSession Chairs: Matthew G.
  F. Dosanjh (Sandia National Laboratories); William Schonbein (Sandia Nati
 onal Laboratories); Amanda J. Bienz (University of New Mexico); and Joseph
  Schuchart (Stony Brook University, Institute for Advanced Computational S
 cience (IACS))
END:VEVENT
END:VCALENDAR