Future Approaches to Data Centric Programming for Exascale
to be held at the
IEEE International Parallel and Distributed Processing Symposium
May 20th, 2011
This meeting is aimed at bringing together researchers, practitioners, and developers,both from the computer science and application areas, that are considering data centric approaches for the efficient utilization of future extreme-scale computing systems. This is currently a very active area of research given the expected increases in future system complexities as we embark on the road to Exascale computing.
The meeting is focused on a number of top ics including: current status in programming models (including UPC, CAF, Global Arrays OpenMP and MPI); current practises; advances in Global Arrays; applicatio n experiences; application needs and their impact on programming models for the future.
8:30 Opening Remarks
8:40 "Global Arrays: Now", Manojkumar Krishnan, Pacific Northwest National Lab.
9:10 "Global Arrays: Into the Future", Daniel Chavarria, Pacific Northwest National Lab.
9:30 "The Future of OpenMP", Barbara Chapman, University of Houston.
10:45 "Chapel's Data Centric Approach to Parallelism and Locality", Brad Chamberlain, Cray.
11:30 "Deterministic Use of Global Address Space via Multi-phase Shared Arrays", Laxmikant Kale, University of Illinois Urbana-Champaign.
1:30 "Coarray Fortran for Exascale Systems?", John Mellor-Crummy, Rice University.
2:15 "Unified Communication Runtime for MPI and PGAS (UPC): Experience with MVAPICH2", Dhabaleswar K. (DK) Panda, Ohio State University.
2:40 "MVAPICH2-GPU: Optimized GPU to GPU Communication for Infiniband Clusters", Sayantan Sur, Ohio State University.
3:30 Panel Discussion "Programming Models at Exascale: Are we ready for the Challenges?"
Brad Chamberlain, Cray Barbara Chapman, Unversity of Houston Daniel Chavarria, Pacific Northwest National Lab Laxikant Kale, University of Illinois John Mellor-Crummy, Rice University DK Panda, Ohio State University
4:30 Closing Remarks
This meeting is sponsored by the eXtreme Scale Computing Initiative (XSCI) at Pacific Northwest National Laboratory (PNNL). XSCI is developing core computer science capabilities as well as new domain science methods, algorithms, and software that can effectively be used on the computing resources that will soon be three orders of magnitude more capable than those available today. The primary driver for PNNL to invest in high-performance computing is to solve problems our society is facing in areas such as energy, the environment, and national security. The fundamental capabilities developed by the XSCI enable scalability and fault tolerance in modeling and simulation applications as demonstrated by the molecular and subsurface science applications that are specifically targeted in the Initiative, and for which PNNL already holds a leadership position, and provide tools for programmability, performance diagnostics and performance and power modeling. A significant portion of the research in the Initiative is focused on the use of Global Arrays/ARMCI as the programming model for exascale application development.
The meeting will take place on May 20th in conjunction with IPDPS in Anchorage, Alaska. Please contact the meeting organizers for further details.Adolfy Hoisie Darren J. Kerbyson T.P. Straatsma
PNNL is a Department of Energy Office of Science national laboratory where interdisciplinary teams advance science and technology and deliver solutions to America’s most intractable problems in energy, the environment, and national security. PNNL employs 4,900 staff, has an annual budget of nearly $1.1 billion, and has been managed by Ohio-based Battelle since the lab’s inception in 1965. PNNL is located in Richland, WA.