(Updated 9/30/26)
Building 402, Lower Gallery, APS Conference Center
Thursday, October 1
8.30-8.45 Welcome, Introduction to Workshop Goals and Outcomes (Rajeev Thakur, Argonne National Laboratory)
8.45-10.00 Plenary Talks (25 min each)
-
8.45-9.10 "Message Passing R&D: How 40 Years of Scalable Systems Research, Standardization, and Quality Open-Source Implementations Enable Exascale Science & Engineering, and AI/ML Applications", Anthony Skjellum, Tennessee Tech University
-
9.10-9.35 "Communication Requirements for HPC and AI", George Bosilca, Nvidia
-
9.35-10.00 "Towards Scalable CFD: MPI Requirements and Challenges", Paul Fischer, University of Illinois and Argonne National Laboratory
10.00-10.30 Coffee Break
10.30-11.45 Plenary Talks (25 min each)
-
10.30-10.55 "MPI for HPC, AI, and Computing Continuum: The MVAPICH Perspective", DK Panda, Ohio State University
-
10.55-11.20 "Current Development on NWChem and Global Arrays", Bruce Palmer, Pacific Northwest National Laboratory
-
11.20-11.45 "Communication Requirements for Connecting Scientific Instruments: Observations from the FedSci & INTERSECT Projects", Thomas Naughton, Oak Rige National Laboratory
11.45-1.15 Lunch (on your own), Argonne Guest House Restaurant
1.15-2.45 Breakout Topic 1 (two concurrent groups)
-
Group 1 (Rm. E1100-1200), Lead: Martin Schulz, Technical University of Munich
-
Group 2 (Lower Gallery), Lead: David Bernholdt, Oak Ridge National Laboratory
2.45-3.15 Coffee Break
3.15-4.45 Breakout Topic 2 (two concurrent groups)
-
Group 1 (Rm. E1100-1200), Lead: Andrew Lumsdaine, Pacific Northwest National Laboratory
-
Group 2 (Lower Gallery), Lead: Joseph Schuchart, Stony Brook University
4.45-5.00 Plenary Talk: "LCI: (Finally) A Universal Low-Level Active Communication Layer", Sanjay Kale, University of Illinois
Dinner on your own (Map of restaurants near Argonne)
Friday, October 2
8.30-10.00 Breakout Topics 3 and 4 (separate groups)
-
Topic 3 - Performance Portability, (Lower Gallery), Lead: Sanjay Kale, University of Illinois
-
Topic 4 - Resilience, etc., (Rm. E1100-1200), Lead: Ignacio Laguna, Lawrence Livermore National Laboratory
10.00-10.30 Coffee Break
10.30-12.00 Report outs from the breakout groups (15 min each)
12.00-1.30 Lunch (on your own), Argonne Guest House Restaurant
1.30-2.45 Plenary Discussion: Research Priorities and R&D Investment Strategy (Lead: Rajeev Thakur, Argonne National Laboratory)
Identify the highest-priority research directions for communication software over the next 5–10 years. How should DOE, national laboratories, academia, industry, and the open-source community coordinate investments to address them.
2.45-3.00 Workshop Report Writing Process & Next Steps
3.00 Workshop Concludes
Breakout Topics
# |
Topic |
Questions |
1 |
Communication for Post-Exascale Architectures |
What communication patterns, volumes, latencies, synchronization requirements, and data-movement challenges will future applications impose? How must MPI, PGAS, CCLs, and other communication runtimes evolve for advanced networks, accelerators, heterogeneous nodes (including quantum), and other emerging architectures? Are multiple MPI implementations still needed? |
2 |
Convergence of HPC and AI Communication |
What can HPC communication libraries learn from AI CCLs, and vice versa? Where should the ecosystems converge or remain distinct? |
3 |
Performance Portability and Communication Optimization |
How can communication software automatically adapt to different hardware, networks, accelerators, and application behaviors? |
4 |
Resilience, Reliability, Security, and Energy Efficiency |
What new communication challenges arise from failures, variability, security requirements, and energy constraints at extreme scale? |