BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20101116T193000Z
DTEND:20101116T201500Z
LOCATION:388
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Petascale machines with hundreds of thousands of cores are being built. These=0Amachines have varying interconnect topologies and large network diameters.=0AComputation is cheap and communication on the network is becoming the=0Abottleneck for strong scaling of parallel applications. Most parallel applications=0Atypically have a certain communication topology. Mapping of tasks in a parallel=0Aapplication based on their communication graph, to the physical processors on=0Athe machine can potentially lead to performance improvements.=0A=0AMapping of the communication graph for an application on to the interconnect=0Atopology of a machine while trying to localize communication is the research=0Aproblem under consideration. Performance improvements for applications such as=0AWRF and NAMD will be presented to motivate the work. Building on these =0Aideas, the talk will discuss algorithms and techniques for automatic mapping=0Aof parallel applications to relieve the application developers of this burden.=0AThe automatic mapping framework is a suite of algorithms with capabilities to =0Achoose the best mapping for a problem with a given communication graph. This=0Aframework will save much effort on the part of application developers to=0Agenerate mappings for their individual applications.=0A=0AMore details on my research available at:=0Ahttp://charm.cs.illinois.edu/~bhatele/phd/
SUMMARY:Mapping Your Application on Interconnect Topologies: Effort versus Benefits
PRIORITY:3
END:VEVENT
END:VCALENDAR