BEGIN:VCALENDAR PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN VERSION:1.0 BEGIN:VEVENT DTSTART:20111116T003000Z DTEND:20111116T010000Z LOCATION:WSCC 611/612 DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: This talk outlines some development challenges on clusters based on multi- and many-core node architectures and in clusters with GPU co-processors. It highlights three tools that make tackling those challenges easy: ThreadSpotter, TotalView, and ReplayEngine.=0A =0AThe ThreadSpotter cache memory optimization tool simplifies finding, understanding and resolving memory bandwidth and cache coherency issues, including false sharing. Problems that usually require a guru are now solvable, and if you do happen to be a performance guru you'll be able to solve them really quickly. The TotalView scalable graphical debugger provides the control and visibility into hybrid and GPU-accelerated MPI applications that you need to make short work of crashes, hangs, deadlocks, and out-of-memory problems. With the ReplayEngine reverse debugging add-on even intermittent race condition type problems become tractable. =0A =0AWe’ll cover the key features that make these tools indispensable in multicore and GPU development. SUMMARY:Memory Optimization and Debugging on Multi- and Many- Core Systems PRIORITY:3 END:VEVENT END:VCALENDAR