Title: Scalable Parallel Primitives for Massive Graph Computation

Speaker: Aydin Buluc, University of California, Santa Barbara

Date/Time: Wednesday, January 20, 2010, 10:00 am        

Location: CSRI Building, Room 279 (Sandia NM) and 915, Room W133

Brief Abstract: Software development for large scale graph and data mining applications is a formidable task that requires an enormous amount of human expertise. In contrast to numerical computing, a scalable software stack that eases the application programmer’s job does not exist for computations on graphs. In this talk, I will introduce the Parallel Combinatorial BLAS, which consist of a small but powerful set of linear algebra primitives specifically targeting graph and data mining applications. In particular, I will talk about the usage and implementation of sparse (generalized) matrix-matrix multiplication as a coarse-grained parallel primitive for tightly-coupled computations on massive graphs. Our focus will be on the new algorithmic techniques that make the primitives scalable, and software engineering techniques that make the combinatorial BLAS generic and extendible. I will conclude by describing an application to social network analysis in detail. I will also briefly mention our pilot studies on emerging architectures.

CSRI POC: Mark D. Rintoul, (505) 844-9592