search-icon
Workshop
:
Mix-and-Match: A Model-Driven Runtime Optimization Strategy for BFS on GPUs
Event Type
Workshop
Registration Categories
W
Tags
Architectures
Data Analytics
Graph Algorithms
TimeMonday, November 12th3:30pm - 3:55pm
LocationD172
DescriptionIt is universally accepted that the performance of graph algorithms is heavily dependent on the algorithm, the execution platform, and the structure of the input graph. This variability remains difficult to predict and hinders the choice of the right algorithm for a given problem.

In this work, we focus on a case study: breadth-first search (BFS), a level-based graph traversal algorithm, running on GPUs. We first demonstrate the severity of this variability by presenting 32 variations of 5 implementation strategies for GPU-enabled BFS, and showing how selecting one single algorithm for the entire traversal can significantly limit performance. To alleviate these performance losses, we propose to mix-and-match, at runtime, different algorithms to compose the best performing BFS traversal. Our approach is based on two novel elements: a predictive model, based on a decision tree, which is able to dynamically select the best performing algorithm for each BFS level, and a quick context switch between algorithms, which limits the overhead of the combined BFS.

We demonstrate empirically that our dynamic switching BFS achieves better performance, outperforming our non-switching implementations by 2x and existing state-of-the-art GPU BFS implementations by 3x. We conclude that mix-and-match BFS is a competitive approach for doing fast graph traversal, while being easily extended to include more BFS implementations and easily adaptable to other types of processors or specific types of graphs.
Archive
Back To Top Button