[Amath-seminars] Reminder Seminar today, 4pm-5pm, Dempsey 104

Loyce M. Adams lma3 at uw.edu
Thu Mar 2 11:15:13 PST 2017

Hi All,

Today's seminar will be given by Prof. Andrew Lumsdaine from 4 to 5
pm in Dempsey 104. Please join us.

Loyce Adams

Andrew Lumsdaine
UW-PNNL Distinguished Faculty Fellow
Chief Scientist, Northest Institute for Advanced Computing
Affiliate Professor of Computer Science, University of Washington
Laboratory Fellow, ACMD, Pacific Northwest National Lab

Title: Challenges and Opportunities in Large Scale Graph Computation

Abstract: Graphs are fundamental abstractions that model relationships
among abstract data entities. Graph algorithms are becoming
increasingly important for solving many problems in computational
science, data-driven science, and the science of data itself. And now
in the Age of Big Data, graph problems are of unprecedented -- and
growing -- scale, suggesting the use of high-performance computing
(HPC) resources. Unfortunately, the algorithms, software, and hardware
that have worked well for developing parallel scientific applications
are not necessarily effective for large-scale graph problems. In this
talk, we illustrate the inter-dependencies among graph problems,
software, and parallel hardware in the current state of the art and
discuss how those issues present inherent challenges in solving
large-scale graph problems. The range of these challenges suggests a
research agenda for the development of scalable high-performance
software for graph problems. To this end, we will present recent work
at the UW-PNNL Northwest Institute for Advanced Computing (NIAC). Our
systematic approach phrases graph algorithm families as collections of
asynchronous, concurrently executing, concise code fragments which may
be invoked both locally and in remote address spaces. A critical
aspect of mapping these fragments to the machine is the runtime system
layer, which performs a number of dynamic optimizations, including
message coalescing, message combining, and software routing. We
identify a number of common patterns in these algorithms, and explore
how this programming model can express those patterns. Algorithmic
transformations are discussed that expose asynchrony, which can be
leveraged by the runtime to improve performance and reduce resource
utilization. Practical implementations and performance results are
provided for a number of representative algorithms. Finally, we
discuss the prospects for hardware/software co-design to build HPC
systems that would better and more naturally support large-scale graph

More information about the Amath-seminars mailing list