Massively Multithreaded Architectures and 
                     Algorithms for Graph Problems

                      Dr. Mahantesh Halappanavar, 
              Pacific Northwest National Laboratory (PNNL)
                          http://www.pnl.gov/



Abstract:

A wide range of problems in diverse fields of science and engineering
can be efficiently solved by formulating them as graph problems. While
graph problems have been extensively studied for decades, they are
challenging to implement on traditional high performance computing
architectures. In this talk, we will discuss some of these challenges
that limit performance and show how to address them. In particular, we
will focus on a non-traditional platform Cray XMT and highlight the
features that can influence the design of emerging multicore and
manycore architectures. The confluence of architecture, algorithm
design, and input characteristics will be discussed using graph coloring
as an example. We will conclude the talk with an introduction to the
massively multithreaded work currently being undertaken at the Center
for Adaptive Super Computing Software (CASS-MT http://cass-mt.pnl.gov)
at the Pacific Northwest National Laboratory.

Mahantesh Halappanavar is a Research Scientist with the Computational
Sciences and Mathematics Division at the Pacific Northwest National
Laboratory. He received his MS and PhD degrees in Computer Science from
the Old Dominion University in 2003 and 2009 respectively.  His research
interests are in graph algorithms and parallel computing.