Implementing a portable Multi-threaded Graph Library: The MTGL on Qthreads

Authors:
Brian W. Barrett;Jonathan W. Berry;Richard C. Murphy;Kyle B. Wheeler
Affiliations:
Sandia National Laboratories, Albuquerque, NM, USA;Sandia National Laboratories, Albuquerque, NM, USA;Sandia National Laboratories, Albuquerque, NM, USA;University of Notre Dame and Sandia National Laboraroties, IN, USA
Venue:
IPDPS '09 Proceedings of the 2009 IEEE International Symposium on Parallel&Distributed Processing
Year:
2009

Citing 0
Cited 4

Multithreaded Asynchronous Graph Traversal for In-Memory and Semi-External Memory

Proceedings of the 2010 ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis
The Combinatorial BLAS: design, implementation, and applications

International Journal of High Performance Computing Applications
Introducing ScaleGraph: an X10 library for billion scale graph analytics

Proceedings of the 2012 ACM SIGPLAN X10 Workshop
Load balanced clustering coefficients

Proceedings of the first workshop on Parallel programming for analytics applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

Developing multi-threaded graph algorithms, even when using the MTGL infrastructure, provides a number of challenges, including discovering appropriate levels of parallelism, preventing memory hot spotting, and eliminating accidental synchronization. In this paper, we have demonstrated that using the combination of Qthreads and MTGL with commodity processors enables the development and testing of algorithms without the expense and complexity of a Cray XMT. While achievable performance is lower for both the Opteron and Niagara platform, performance issues are similar. While we believe it is possible to port Qthreads to the Cray XMT, this work is still on-going. Therefore, porting work still must be done to move algorithm implementations between commodity processors and the XMT. Although it is likely that the Qthreads-version of an algorithm will not be as optimized as a natively implemented version of the algorithm, such a performance impact may be an acceptable trade-off for ease of implementation.