BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111116T190000Z
DTEND:20111116T193000Z
LOCATION:WSCC 611/612
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Future-generation HPC systems comprising many-core sockets and GPU accelerators will impose increasing challenges in programming, efficiency, heterogeneity and scalability for extreme-scale computing. =0A=0AEmerging execution models using event-driven, task-based parallelism; dynamic dependency and constraint analysis; locality-aware computation; and resource-aware scheduling show promise in addressing these challenges and meeting the needs of the future of supercomputing. Applications utilizing these innovative runtime systems have shown significant gains in performance and utilization of computational resources over conventional methodologies.  On the Graph 500 challenge, for example, SWARM (SWift Adaptive Runtime Machine) technology has shown 2-10x speed-up over MPI implementations and has demonstrated the highest per-node performance over all multi-node runtime systems. This presentation will articulate the challenges that extreme-scale systems pose and highlight some of the solutions to these problems, including new programming models that allow developers to express concurrency, locality, and synchronization in a concise manner.
SUMMARY:Runtimes and Applications for Extreme-Scale Computing
PRIORITY:3
END:VEVENT
END:VCALENDAR