The STAPL Parallel Graph Library

Harshvardhan,; Fidel, Adam; Amato, Nancy M.; Rauchwerger, Lawrence

doi:10.1007/978-3-642-37658-0_4

Cited by 25 publications

(25 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We implemented DCSC, MULTIPIVOT, and SCCMULTI using C++ and the STAPL library, a parallel library which provides a distributed graph [2]. We did not implement NSCC because it uses a specific RQ algorithm that maintains comAlgorithm 1 SCCMULTI 1: for k from 0 to log n do 2: for all i ← 1 to 2 k |V | n do 3: vi ← random node(V ) 4: for all u ∈ SU CC(vi) do 5: u.marks ← u.marks ∪ si 6: for all u ∈ P RED(vi) do 7:…”

Section: Resultsmentioning

confidence: 99%

SCCMulti

Tomkins

Smith

Amato

et al. 2014

Proceedings of the 19th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

Self Cite

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

SCCMulti

Tomkins

Smith

Amato

et al. 2014

Proceedings of the 19th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

Self Cite

View full text Add to dashboard Cite

“…We implemented the KLA paradigm and some important graph algorithms in the stapl Graph Library (sgl) [14] to evaluate its performance. sgl consists of a generic parallel graph container (pGraph), graph pViews [9], and a collection of parallel graph algorithms to allow users to easily process graphs at scale.…”

Section: Sgl Overviewmentioning

confidence: 99%

“…We implement the KLA paradigm using the stapl Graph Library (sgl) [14] and evaluate its performance and scalability up to 98,000 cores on a Cray XE6 machine for important classes of graph algorithms including traversals (e.g., BFS, single-source shortest paths, connected components), random walks (e.g., PageRank) and k-core decomposition. Our experimental studies show that KLA improves the performance of these algorithms on some graphs by a factor of 10 or more, at scale.…”

Section: Introductionmentioning

confidence: 99%

Kla

Harshvardhan

Fidel

Amato

et al. 2014

Proceedings of the 23rd International Conference on Parallel Architectures and Compilation

Self Cite

View full text Add to dashboard Cite

This paper proposes a new algorithmic paradigm -k-level asynchronous (KLA) -that bridges level-synchronous and asynchronous paradigms for processing graphs. The KLA paradigm enables the level of asynchrony in parallel graph algorithms to be parametrically varied from none (levelsynchronous) to full (asynchronous). The motivation is to improve execution times through an appropriate trade-off between the use of fewer, but more expensive global synchronizations, as in level-synchronous algorithms, and more, but less expensive local synchronizations (and perhaps also redundant work), as in asynchronous algorithms. We show how common patterns in graph algorithms can be expressed in the KLA pardigm and provide techniques for determining k, the number of asynchronous steps allowed between global synchronizations. Results of an implementation of KLA in the stapl Graph Library show excellent scalability on up to 96K cores and improvements of 10x or more over levelsynchronous and asynchronous versions for graph algorithms such as breadth-first search, PageRank, k-core decomposition and others on certain classes of real-world graphs.

show abstract

“…Parallel algorithms are expressed as arbitrary task dependence graphs in STAPL. Load imbalance in parallel computations is dealt with in various ways in STAPL. For repartitioning, this is realized through redistribution of the two pGraphs [16] (i.e., the region graph and the roadmap or RRT graph) in the parallel motion planning algorithms. Alternatively, load balancing can be addressed by using a custom work stealing scheduler for parallel motion planning algorithms.…”

Section: A Implementation In Staplmentioning

confidence: 99%

Using Load Balancing to Scalably Parallelize Sampling-Based Motion Planning Algorithms

Fidel

Jacobs

Sharma

et al. 2014

2014 IEEE 28th International Parallel and Distributed Processing Symposium

Self Cite

View full text Add to dashboard Cite

Abstract-Motion planning, which is the problem of computing feasible paths in an environment for a movable object, has applications in many domains ranging from robotics, to intelligent CAD, to protein folding. The best methods for solving this PSPACE-hard problem are so-called sampling-based planners. Recent work introduced uniform spatial subdivision techniques for parallelizing sampling-based motion planning algorithms that scaled well. However, such methods are prone to load imbalance, as planning time depends on region characteristics and, for most problems, the heterogeneity of the subproblems increases as the number of processors increases. In this work, we introduce two techniques to address load imbalance in the parallelization of sampling-based motion planning algorithms: an adaptive work stealing approach and bulk-synchronous redistribution. We show that applying these techniques to representatives of the two major classes of parallel samplingbased motion planning algorithms, probabilistic roadmaps and rapidly-exploring random trees, results in a more scalable and load-balanced computation on more than 3,000 cores.

show abstract

The STAPL Parallel Graph Library

Cited by 25 publications

References 13 publications

SCCMulti

SCCMulti

Kla

Using Load Balancing to Scalably Parallelize Sampling-Based Motion Planning Algorithms

Contact Info

Product

Resources

About