|Name:||Programming Models & Tools|
|Time:||Monday, June 23, 2014
04:00 pm - 05:00 pm
CCL - Congress Center Leipzig
|Chair:||Barbara Chapman, University of Houston|
|Abstract:||With rapid innovation in the hardware used to build the nodes of large-scale computers, traditional approaches to programming HPC platforms are no longer adequate. Such nodes are now configured with an increasing number of cores that share memory and which may be augmented with specialized cores or accelerators.
Several programming interfaces have emerged that aim to support the application developer's need to obtain high performance on current and emerging HPC systems with a reasonable development cost. OpenMP and OpenACC are two compatible, rapidly evolving, directive-based approaches that are already used in the creation of HPC code to exploit node capabilities. Several so-called "Partitioned Global Address Space" (PGAS) programming models have been proposed that promise to provide high performance, potentially across the entire system, via a good match with architectural characteristics. With expert presentations in each of these areas, this session gives an overview of some of the most important advances and trends in high-level programming models for HPC.
|Presentations:||OpenACC & the Evolution of the Modern GPU
04:00 pm - 04:20 pm
|Duncan Poole, NVIDIA|
|Is PGAS Ready for Prime Time?
04:20 pm - 04:40 pm
|Michèle Weiland, EPCC|
|OpenMP 4.0 & Beyond
04:40 pm - 05:00 pm
|Christian Terboven, RWTH Aachen University|