Home

Synchronization-reducing and Communication-reducing Algorithms and Programming Models for Large-scale Simulations (January 9-13, 2012)


Organizing Committee

 

[Image courtesy of Hatem Ltaief]
Directed Acyclic Graph for the execution of a symmetric generalized eigensolver for a dense 4x4 matrix

 

Description

As concurrency in scientific computing pushes beyond a million threads and performance of individual threads becomes less reliable for hardware-related reasons, attention of mathematicians, computer scientists, and supercomputer users and suppliers inevitably focuses on reducing communication and synchronization bottlenecks. Though convenient for succinctness, reproducibility, and stability, instruction ordering in contemporary codes is commonly overspecified. This workshop attempts to outline evolution of simulation codes from today's infra-petascale to the ultra-exascale and to encourage importation of ideas from other areas of computer science into numerical algorithms, new invention, and programming model generalization.


Monday January 9th, 2012
Time Description Speaker Location Abstracts Slides
8:30 - 8:55 Welcome Coffee & Registration   11th Floor Collaborative Space
8:55 - 9:00 Welcome Jill Pipher, Director, ICERM 11th Floor Lecture Hall
9:00 - 9:45

Workshop Goals and Structure

Meeting Logistics

David Keyes, KAUST 11th Floor Lecture Hall
PDF
9:45 - 10:45 Lightning Talks
(30 of two minutes each)
  11th Floor Lecture Hall
10:45 - 11:00 Coffee/Tea Break   11th Floor Collaborative Space
11:00 - 12:00 Algorithmic Challenges of Exascale Computing Katherine Yelick, University of California, Berkeley 11th Floor Lecture Hall
PDF
12:00 - 2:00 Working Lunch  
2:00 - 3:00 Algorithm Design using Spectral Graph Theory Gary Miller, CMU 11th Floor Lecture Hall
PDF
3:00 - 3:30 Coffee/Tea Break   11th Floor Collaborative Space
3:30 - 4:00 Lower bounds for communication in linear algebra Grey Ballard, UC Berkeley 11th Floor Lecture Hall
PDF
PDF
4:00 - 4:30 Scalable preconditioners for time-harmonic wave equations Jack Poulson, University of Texas, Austin 11th Floor Lecture Hall
PDF
PDF
4:30 - 5:00 An Integrative Model for Parallelism Victor Eijkhout, University of Texas, Austin 11th Floor Lecture Hall
PDF
PDF
5:00 - 6:30 Welcome Reception   11th Floor Collaborative Space
     
Tuesday January 10th, 2012
Time Description Speaker Location Abstracts Slides
9:00 - 10:00 Multigrid at Extreme scales: Communication Reducing Data Models and Asynchronous Algorithms Mark Adams, Columbia University 11th Floor Lecture Hall
PDF
PDF
10:00 - 10:30 Coffee/Tea Break   11th Floor Collaborative Space
10:30 - 11:30 Discussion and assignment of breakout groups   11th Floor Lecture Hall
11:30 - 12:00 Initial Group Meetings   Locations will vary
12:00 - 12:05 Group Photo   11th Floor Lecture Hall
12:05 - 2:00 Working Lunch    
2:00 - 3:00 Sparse linear solvers on ever-wider machines Jonathan Cohen, NVIDIA 11th Floor Lecture Hall
PDF
3:00 - 3:30 Coffee/Tea Break   11th Floor Collaborative Space
3:30 - 4:00 Principles of algorithm-architecture co-design
Richard Vuduc, Georgia Institute of Technology 11th Floor Lecture Hall
PDF
4:00 - 4:30 Topical Talk V Dinesh Kaushik, KAUST 11th Floor Lecture Hall
4:30 - 5:00 Toward less synchronous composable multilevel methods for implicit multiphysics simulation Jed Brown, Argonne National Laboratory 11th Floor Lecture Hall
PDF
PDF
6:00 Dinner  
Local 121
121 Washington Street
Providence, RI 02903
(directions are available at the 11th floor reception desk)
Wednesday January 11th, 2012
Time Description Speaker Location Abstracts Slides
9:00 - 10:00 Finding the Happy Medium: Tradeoffs in communication, algorithms, architectures and programming models Bill Gropp, University of Illinois at Urbana-Champain 11th Floor Lecture Hall
PDF
PDF
10:00 - 10:30 Computational Challenges of Coupled Cluster Theory Jeff Hammond, Argonne National Laboratory 11th Floor Lecture Hall
PDF
PDF
10:30 - 10:55 Coffee/Tea Break   11th Floor Collaborative Space
10:55 - 11:55 Breakout Groups   Locations will vary
11:55 - 1:40 Working Lunch  
1:40 - 2:40 Synchronization and Manycore Machines John Owens, UC Davis 11th Floor Lecture Hall
PDF
PDF
2:40 - 3:10 Making Dataflow Programming Ubiquitous for Scientific Computing Hatem Ltaief, KAUST 11th Floor Lecture Hall
PDF
PDF
3:10 - 3:35 Coffee/Tea Break   11th Floor Collaborative Space
3:35 - 4:35 Preliminary plenary reports from Breakout Groups   11th Floor Lecture Hall
Thursday January 12th, 2012
Time Description Speaker Location Abstracts Slides
9:00 - 10:00 Making Good Enough...Better: Addressing the Multiple Objectives of High-Performance Parallel Software with a Mixed Global-Local Worldview John Gunnels, IBM 11th Floor Lecture Hall
PDF
PDF
10:00 - 10:30 Automatic Selection of Compiler Optimizations Using Graph-Based Program Characterization John Cavazos, University of Delaware 11th Floor Lecture Hall
PDF
10:30 - 11:00 Coffee/Tea Break   11th Floor Collaborative Space
11:00 - 12:00 Breakout Groups   Locations will vary
12:00 - 2:00 Break for Lunch  
2:00 - 3:00 Hierarchical N-body algorithms: A pattern likely to lead at extreme scales Lorena Barba, Boston University 11th Floor Lecture Hall
PDF
PDF
3:00 - 3:30 Coffee/Tea Break   11th Floor Collaborative Space
3:30 - 4:00 Group presentation I   11th Floor Lecture Hall
4:00 - 4:30 Group presentation II   11th Floor Lecture Hall
4:30 - 5:00 Group presentation III   11th Floor Lecture Hall
Friday January 13th, 2012
Time Description Speaker Location Abstracts Slides
9:00 - 10:00 Towards a Science of Parallel Programming Keshav Pingali, UT Austin 11th Floor Lecture Hall
PDF
PDF
10:00 - 10:30 Coffee/Tea Break   11th Floor Collaborative Space
10:30 - 11:30 Path forward from the point of view of mathematical challenges   11th Floor Lecture Hall
11:30 - 12:00 Workshop Wrapup David Keyes, KAUST 11th Floor Lecture Hall
12:00 Adjournment