BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121114T180000Z
DTEND:20121114T181500Z
LOCATION:155-F
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The increasing size and complexity of High Performance Computing systems is making it increasingly likely that individual circuits will produce erroneous results, especially when operated in a low energy mode.   We describe one general approach for converting applications into more error tolerant forms by recasting these applications as numerical optimization problems.  We also show how both intrinsically robust algorithms as well as fragile applications can exploit this framework and in some cases provide significant energy reduction.  We also propose a set of algorithmic techniques for detecting faults in sparse linear algebra.  These techniques are shown to yield up to 2x reductions in performance overhead over traditional ABFT checks.  We also propose algorithmic error localization and partial recomputation as an approach for efficiently making forward progress.  This approach improves the performance of the CG solver in high error scenarios by 3x- 4x and increases the probability of successful completion by 60%.
SUMMARY:Algorithmic Approaches to Building Robust Applications for HPC Systems of the Future
PRIORITY:3
END:VEVENT
END:VCALENDAR