BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121114T223000Z
DTEND:20121115T000000Z
LOCATION:155-F
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: A very important issue for most parallel applications is efficient load-balancing. My current focus in PhD research is load-balancing for multicores and clusters of multicores. We have shown that sometimes an optimistic parallelization approach can be used to avoid the use of locks and atomic instructions during dynamic load balancing on multicores. We have used this approach to implement two high-performance parallel BFS algorithms based on centralized job queues and distributed randomized work-stealing, respectively. Our implementations are highly scalable and faster than most state-of-the-art multicore parallel BFS algorithms. In my work on load-balancing on clusters of multicores, I have implemented distributed-memory and distributed-shared-memory parallel octree based algorithms for approximating polarization energy of molecules by extending existing work on shared-memory architectures. For large enough molecules our implementations show a speedup factor of about 10k w.r.t. the naïve algorithm with less than 1% error using as few as 144 cores (=12nodesX12cores/node).
SUMMARY:Dynamic Load-Balancing for Multicores
PRIORITY:3
END:VEVENT
END:VCALENDAR