ESSEX-II – Equipping Sparse Solvers for Exascale

ESSEX project documentation

Inhalt

ISC15 Workshop “Performance Modeling: Methods and Applications”

IMG_20150716_105807.jpg

Jeffrey Vetter (ORNL/GATech) presenting at the Performance Modeling workshop.

Our workshop at ISC High Performance 2015 in Frankfurt on July 16, 2015, was well attended, and was considered by all speakers a great success. Performance modeling experts from different groups across the world presented their view of this important field, and interesting (and sometimes heated) discussions emerged. We saw a broad spectrum of methods, from analytical modeling to frameworks and from scaling analysis to machine learning approaches. Power and energy consumption also received due attention. In his keynote, Bill Gropp (UIUC) emphasized the importance of knowing the inherent hardware bottlenecks of numerical software in order to be able to put performance figures into perspective – a scalable application can nevertheless be extremely inefficient (in fact, there is a high probability that it is).

All slides can be found on the workshop page.

Workshop “Sparse Solvers for Exascale: From Building Blocks to Applications”

Exascale15_participants.jpgThe Workshop takes place in  Greifswald, Germany, March 23-25, 2015. It focuses on sparse linear solvers for eigenvalue problems with applications in mathematics and natural sciences, and it highlights important aspects in designing and implementing numerical methods and software building blocks for future exascale systems. Performance engineering, advanced parallel programming techniques and fault-tolerance strategies crossing the traditional boundaries of software layers will be addressed. The workshop website has additional information.

The workshop is organized by the Institute of Physics of the University of Greifswald (IfP), and held in the Alfried-Krupp Wissenschaftskolleg Greifswald.

The workshop is supported by the Stiftung Alfried-Krupp-Wissenschaftskolleg Greifswald, by the German Research Foundation (DFG) through Priority Programme “Software for Exascale Computing” (SPPEXA), and by the Intel corporation.