You are here: Home ExaMPI14 - Exascale MPI 2014

ExaMPI14 - Exascale MPI 2014

Workshop on Exascale MPI at Supercomputing Conference 2014 - November 17 2014 -New Orleans, LA, USA

ExaMPI14: Workshop on  Exascale MPI 2014

https://www.pdc.kth.se/exampi14

New Orleans, LA

November 17, 2014

Held in conjunction with SC14 (http://sc14.supercomputing.org)

Held in cooperation with ACM SIGHPC (http://www.sighpc.org)

9.00 - 9.20 Welcome and Introduction to the Workshop - Stefano Markidis

9.20 - 10.00 Is MPI+X enough for Exascale? - Keynote Talk - William D. Gropp

As the time to the first Exascale system gets shorter, it is clear that MPI+X will be an important, and possibly the dominant, programming system for those Exascale systems.  This is because MPI has demonstrated remarkable scalability over the past twenty years and because the ability to combine MPI with other programming systems (the
"X" in MPI+X) has allowed MPI to concentrate on internode communication while "X", be it OpenMP, OpenACC, CUDA, or something else, provides an intranode programming system.  But will the simple combination of MPI with an intranode programming system be sufficient for Exascale?  This talk will review the current state of MPI in the
context of Exascale architectures and likely choices of X, and discuss some open problems for MPI+X on Exascale systems.

10.00 - 10.30 Break

10.30 - 11.00 Early Experiences Co-Scheduling Work and Communication Tasks for Hybrid MPI+X Applications by Dylan Stark, Richard Barrett, Ryan Grant, Stephen Olivier, Kevin Pedretti and Courtenay Vaughan (Regular Paper)

11.00 - 11.15 Context id allocation for end-points communicators by Daniel Holmes (Hot topic paper)

11.15 - 11.45 To INT_MAX ... and beyond! by Jeff Hammond, Andreas Schäfer and Robert Latham (Regular paper)

11.45 - 12.00 MPI Debugging with Handle Introspection by  Laust Brock-Nannestad, John Delsignore, Jeffrey M. Squyres, Sven Karlsson and Kathryn Mohror (Hot topic paper)

12.00 - 1.30 Lunch

1.30 - 2.10 MPI-IO: Time to Move On ? - Keynote Talk - Dries Kimpe

2.10 - 2.25 OmpSs Collective Offload by Florentino Sainz, Vicenç Beltran and Jesus Labarta (Hot topic paper)

2.25 - 2.40 MPI collectives at scale by Christoph Niethammer, Pekka Manninen, Rupert W Nash, Dmitry Khabi and Jose Gracia (Hot topic paper)

2.40 - 2.55 Tolerating Process Imbalance in Hierarchical Collective Operations by Benjamin S. Parsons and Vijay Pai (Hot topic paper)

3.00 - 3.30 Break

3.30 - 4.10 MPI Communication Library for Next Exa-scale Systems- Keynote Talk - Shinji Sumimoto

4.10 - 4.40 Simplifying the Recovery Model of User Level Failure Mitigation by Wesley Bland, Kenneth Raffenetti and Pavan Balaji (Regular paper)

4.40 - 4.55 A Global Exception Fault Tolerance Model for MPI by Howard Pritchard, Ignacio Laguna, Kathryn Mohror, Todd Gamblin, Martin Schulz and Nickolas Davis (Hot topic paper)

4.55 - 5.10 Running Fault-Tolerant MPI-based Applications in Unstable Systems by Edson Tavares Camargo and Elias Procópio Duarte (Hot topic paper)

5.10 - 5.25 A Personalized MPI library for Exascale Applications and Environments by Shweta Jha, Edgar Gabriel and Saber Feki (Hot topic paper)

5.25 - 5.30 Final Remarks - Stefano Markidis

 

Organizers

---------------------------------------------------------------------------------------

Erwin Laure,  KTH Royal Institute of Technology

Stefano Markidis,  KTH Royal Institute of Technology

Jesper Larsson Träff, TU Wien

Daniel Holmes,  Edinburgh Parallel Computing Centre

Mark Bull,  Edinburgh Parallel Computing Centre

William Gropp,  University of Illinois at Urbana-Champaign

Masamichi Takagi, NEC Corporation in Japan

Lorna Smith,  Edinburgh Parallel Computing Centre

Mark Parsons,  Edinburgh Parallel Computing Centre

 

Program Committee

---------------------------------------------------------------------------------------

Erwin Laure,  KTH Royal Institute of Technology

Stefano Markidis,  KTH Royal Institute of Technology

William Gropp,  University of Illinois at Urbana-Champaign

Marc Snir, University of Illinois at Urbana-Champaign

Ewing Lusk, Argonne National Laboratory

John Mellor-Crummey, Rice University

Jesper Larsson Träff, TU Wien

Torsten Hoefler,  ETH Zürich

Hans-Christian Hoppe, Intel

Daniel Holmes,  Edinburgh Parallel Computing Centre

Mark Bull,  Edinburgh Parallel Computing Centre

Masamichi Takagi, NEC Corporation in Japan

Lorna Smith,  Edinburgh Parallel Computing Centre

Mark Parsons,  Edinburgh Parallel Computing Centre

 

 

Contact

---------------------------------------------------------------------------------------

Please contact Stefano Markidis at markidis at kth.se