BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20131120T203000Z
DTEND:20131120T210000Z
LOCATION:401/402/403
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Stencil computations are a common class of operations that appear in many=0Acomputational scientific and engineering applications. Stencil computations=0Aoften benefit from compile-time analysis, exploiting data-locality, and=0Aparallelism. Post-processing of discontinuous Galerkin (dG) simulation=0Asolutions with B-spline kernels is an example of a numerical method which=0Arequires evaluating computationally intensive stencil operations over a mesh.=0APrevious work on stencil computations has focused on structured meshes, while=0Agiving little attention to unstructured meshes.  Performing stencil operations=0Aover an unstructured mesh requires sampling of heterogeneous elements which=0Aoften leads to inefficient memory access patterns and limits data=0Alocality/reuse.  In this paper, we present an efficient method for performing=0Astencil computations over unstructured meshes which increases data-locality and=0Acache efficiency, and a scalable approach for stencil tiling and concurrent=0Aexecution.  We provide experimental results in the context of post-processing of=0AdG solutions that demonstrate the effectiveness of our approach.
SUMMARY:A Scalable, Efficient Scheme for Evaluation of Stencil Computations over Unstructured Meshes
PRIORITY:3
END:VEVENT
END:VCALENDAR