HPC Graph Analysis

Graph Algorithms Building Blocks (GABB’2017)

Buena Vista Palace Hotel
Orlando, Florida, USA

29 May 2017

Scope and Goals:

This workshop series started with the narrow goal of exploring the definition of a set of basic building blocks for graph algorithms (http://graphblas.org), analogous to the Basic Linear Algebra Subprograms (BLAS) in numerical linear algebra. Over the years, our scope has expanded.  We’ve welcomed a wide range of papers into GABB covering graph computations with an emphasis on high-performance and parallel computing. We expect this trend to continue in 2017.

In particular, while we retain our focus on "building blocks" for graph computations (including those not based on linear algebra), we also welcome work that focuses on libraries and frameworks to support applications that use graph algorithms.

Our target audience is developers of graph algorithms, researchers in high-performance computing working on irregular applications, applied mathematicians working on fundamental algorithmic building blocks for graph computations, and application scientists using graphs in their computational work and data analysis.

Location:

IPDPS 2016 logoThis workshop is co-located with IPDPS 2017, held 29 May - 2 June 2017, at the Buena Vista Hotel, in Orlando, Florida, USA. Registration information for IPDPS2017 can be found at here.

 

Details and Dates

To submit a paper, upload a PDF copy here. Submitted manuscripts may not exceed ten (10) single-spaced double-column pages using 10-point size font on 8.5x11 inch pages (IEEE conference style), including figures, tables, and references (see IPDPS Call for Papers for more details). All papers will be reviewed. Proceedings of the workshops are distributed at the conference and are submitted for inclusion in the IEEE Xplore Digital Library after the conference.


GABB 2017 Keynote Speaker

Ümit V. Çatalyürek, Georgia Institute of Technology

GABB 2017 Agenda

GABB 2017 Final Program (PDF)

Session

Title

Authors

1

(8:30am-10am)

Chair: Aydin Buluc

Breadth-first search with a multi-core computer

Maryia Belova and Ming Ouyang

Order or Shuffle: Empirically Evaluating Vertex Order Impact on Parallel Graph Computations

George M Slota, Siva Rajamanickam and Kamesh Madduri

A Study of Graph Decomposition Algorithms for Parallel Symmetry Breaking

Sayyad Nayyaroddeen, Mahak Gambhir and Kishore Kothapalli

 

Morning Break (10-10:30am)

2

(10:30am-12)

Chair: Chris Long

Constructing Adjacency Arrays from Incidence Arrays

Hayden Jananthan, Karia Dibert and Jeremy Kepner

Mini-Gunrock: A Lightweight Graph Analytics Framework on the GPU

Yangzihao Wang, Sean Baxter and John Owens   

Algebraic Multigrid for Least Squares Problems on Graphs with Applications to HodgeRank

Charles Colley, Junyuan Lin, Xiaozhe Hu and Shuchin Aeron

 

Lunch (12-1:30pm)

3

(1:30pm-3pm)

Chair: Tim Mattson

Keynote talk:

HPC Graph Analytics: Trends and Fallacies

In the era of data and ubiquitous computing, solutions to many of the social, scientific, and engineering problems necessitate the analysis and integration of very large data captured in multiple scales. Often, the data we capture are irregular and modeled as graphs. In this talk, I will first present some examples of algorithm reengineering and system support for efficient graph analytics. Then, I will discuss fallacies and shortcomings of the current HPC Graph Analytics trends

Ümit V. Çatalyürek is Professor and Associate Chair of the School of Computational Science and Engineering in the College of Computing at the Georgia Institute of Technology. Dr. Çatalyürek is a recipient of an NSF CAREER award and is the primary investigator of several awards from DOE, NIH, and NSF. Dr. Çatalyürek currently serves as an Associate Editor for Parallel Computing, and as an editorial board member for IEEE TPDS, and the JPDC. He is a Fellow of IEEE, member of ACM and SIAM, and the Chair for IEEE TCPP for 2016-2017. His main research areas are in parallel computing, combinatorial scientific computing and biomedical informatics.

Deriving Streaming Graph Algorithms from Static Definitions

David Ediger and James Fairbanks

 

Afternoon Break (3-3:30pm)

4

(3:30pm-5:15pm)

Chair: Henning Meyerhenke

Design of the GraphBLAS API for C

Aydin Buluc, Tim Mattson, Scott McMillan, Jose Moreira and Carl Yang

A Linear Algebra-based Programming Interface for Graph Computations in Scala and Spark

William Horn, Gabriel Tanase, Hao Yu and Pratap Pattnaik

Panel: Computational primitives across application domains: converging or diverging?

 

We will prime our discussion by asking whether building blocks used in graph computations have any commonalities with those used in other domains such as machine learning, sparse linear algebra, computational biology and computational chemistry.

 

Panelists: Srinivas Aluru (Georgia Tech), John Feo (PNNL), Esmond Ng (LBNL), Tim Mattson (Intel)

 

Moderator: Aydin Buluc



Workshop Organizers:

Co-chairs:

Program committee members (in addition to the co-chairs):

Steering committee: