- This event has passed.
17 November 2013 - 22 November 2013
PRACE – The European HPC Infrastructure at SC13
PRACE offered some interesting activities to visitors of SC13 and the PRACE booth:
Every year at SC HPCWire announces the winners of the HPCwire Reader’s Choice Awards. HPCWire asks readers across the HPC community to recognize the most outstanding organizations and individuals in the industry. This year, 6 PRACE partners, systems or projects are nominated in 6 categories, which shows how PRACE members and their researchers are working at the forefront of science and engineering. Whether in manufacturing, automotive, “edge HPC”, storage, or Big Data, PRACE members are nominated for their excellence! The nominees can be found here The winners will be announced on Monday 18 November at 6:00 PM (MST) at the HPCWire booth (#3510).
HLRS together with DataDirect Networks for large scale problem resolution in automotive and physics received the 2013 Reader’s Choice Award for Best use of HPC application in manufacturing.
The 2013 Reader’s Choice Award for Best Application of “Big Data” in HPC went to GENCI CURIE for its DEUS (Dark Energy Universe Simulation) project.
The highly coveted HPCwire Readers’ Choice and Editors’ Choice Awards winners are selected by a polling of HPCwire’s global audience for the Readers’ Choice, combined with winners selected by a panel of editors, staff executives and HPC luminaries for the Editor’s Choice. The formal presentation of the awards takes place during the week of the Supercomputing Conference each year, which focuses upon high performance computing, hardware, software, networking, storage, and scientific breakthroughs. Widely recognized as one of the most prestigious awards presented during the conference, the awards honour demonstrated excellence and outstanding technological advancements achieved by the HPC community.
SC13 Best Paper
The paper entitled Scalable Domain Decomposition Preconditioners For Heterogeneous Elliptic Problems by Pierre Jolivet, Frédéric Hecht (both Universite Pierre et Marie Curie and INRIA, France), Frédéric Nataf (CNRS, Universite Pierre et Marie Curie and INRIA, France), Christophe Prud’homme (Université de Strasbourg, France) is nominated for the Best Paper Award at SC13. Their paper is linked to the PRACE supported projects HPC-PDE (4th Call) and HPC-PDE 2 (6th Call).
About the paper
Domain decomposition methods are, alongside multigrid methods, one of the dominant paradigms in contemporary large-scale partial differential equation simulation. The project is oriented towards the design of efficient domain decomposition solvers for various problems, such as how solid objects deform and become internally stressed due to prescribed loading conditions. When such objects are made of materials with different proprieties, such as steel and rubber, traditional solvers tend to perform poorly when increasing the number of subdomains. Jolivet’s paper at SC’13 details the construction of a scalable preconditioner for solving on more than 16,000 threads various highly heterogeneous problems, yielding unstructured discrete linear systems of few billion unknowns. It was nominated for the best paper award out of the 92 papers accepted by the technical program committee.
SC13 Best Student Paper
The paper entitled Parallel Design and Performance of Nested Filtering Factorization Preconditioner by Long QU (Université Paris Sud, France), Laura GRIGORI (INRIA and Universite Pierre et Marie Curie, France), and Frederic NATAF (CNRS, Universite Pierre et Marie Curie and INRIA, France) is nominated for Best Student Paper at SC13. Their paper is linked to the PRACE supported projects HPC-PDE (4th Call) and HPC-PDE 2 (6th Call).
About the paper
Solving large sparse linear systems of linear equations Ax = b is an operation used in many industrial and academic simulations. Given the complexity of today’s numerical simulations and the sizes of the linear systems that need to be solved, the design of linear solvers that have limited memory requirements and that can scale to a large number of processors is of major interest. Qu et al. paper at SC’13 focuses on nested filtering factorization (NFF), a multilevel parallel preconditioning technique based on a recursive factorization which is a key feature in allowing NFF to have limited memory requirements and also to be very well suited for hierarchical parallel machines. Besides that, NFF is also able to preserve some directions of interest of the input matrix. The paper details the parallel design of NFF which scales up to 2048 threads, and is up to 2.6 times faster than the domain decomposition preconditioner Restricted Additive Schwarz implemented in PETSc for a boundary value problem with high heterogenous coeifficents. It was nominated for best student paper finalist out of the 92 papers accepted by the technical program committee.
PRACE 2.0 Strategy – the future of PRACE
Presenters: Catherine Rivière and Sergi Girona
Date: Tuesday 19 November 2013
Time: 3:45 PM (MST)
HPC training perspectives and collaborations from PRACE (Europe), XSEDE (US) and RIKEN AICS (Japan)
Presenters: David Henty
Date: Wednesday 20 November 2013
Time: 3:45 PM (MST)
PRACE Industrial offering – SHAPE highlights & Open R&D
Presenters: Giovanni Erbacci and Stéphane Requena
Date: Thursday 21 November 2013
Time: 11:00 AM (MST)
HPC training plays a vital role in fostering an ever-growing community of researchers and developers who conducts their work on supercomputers. In this BOF session, representatives from the Partnership for Advanced Computing in Europe (PRACE), the Extreme Science and Engineering Discovery Environment (XSEDE), and the RIKEN Advanced Institute for Computational Science (AICS) will each present their HPC training programs in Europe, the US, and Japan, respectively. The presentations will be followed by a discussion of the respective experiences and lessons learned, opportunities for sharing materials, processes for assessing and certifying knowledge acquisition, and a discussion of opportunities for collaboration.
The winners of the PRACE Treasure Hunt were:
- Run Wei Cheng (BGP)
- Yan Lui (UIUC)
- Oiangian Shi (Tsinghua University)