The APOS and HOPSA projects are concerned with the exploitation of next generation high-performance computing (HPC) systems. Three of the top five systems in the November 2011 Top500 list consist of both CPU and NVIDIA GPUs. Within the next year, Intel's MIC platform will likely be added to that mix. In addition, CPUs are becoming increasingly heterogeneous themselves with complicated memory architectures. While APOS investigates the methods and technologies used to program such systems (e.g. HMPP, OpenACC or OmpSs), HOPSA focusses on performance tools that can aid programmers to realise their full potential (e.g. Paraver, Scalsca, ThreadSpotter or Vampir). The Partners in the two projects include experts from the EU and Russia, and represent supercomputing centres, tools developers and end-users of HPC systems. As such we are uniquely placed to give an insight into the challenges faced by scientists and programmers alike and offer advice on how to embrace the changes in hardware design and move forward.
|14:00 - 14:30||Lawrence Mitchell, EPCC - The University of Edinburgh||"Producing fast, portable unstructured mesh programs"|
|14:30 - 15:00||Francois Bodin, CAPS entreprise||"Autotuning with the HMPP workbench"|
|15:00 - 15:30||Vladimir Bakhtin, Keldysh Institute for Applied Mathematics||"Extension of the DVM Parallel Programming Model for Clusters with Heterogeneous Nodes"|
|15:30 - 16:00||Colin Glass, HLRS||"Quantitative Research in High-Performance Molecular Dynamics"|
|16:00 - 16:30||- COFFEE BREAK -|
|16:30 - 17:30||Andreas Knuepfer, TU Dresden||"Parallel Application Monitoring Meets System Monitoring - the HOPSA Project"|
This workshop is co-located with the HiPEAC 2013 conference, which will be held at the Radisson Blue Hotel in Berlin.
EPCC, The University of Edinburgh
EH9 3JZ Edinburgh, UK
e-mail: m.weiland at epcc.ed.ac.uk
tel: +44 131 651 3580