Graduate student Robbie Watling, Computer Science M.S., will present his final oral examination (master’s defense) on Wednesday, April 12, 2023, at 10 a.m. in Dillman Hall, Room 208, and virtually.
Watling is advised Zhenlin Wang and Junqiao Qiu.
The title of Watling’s defense is, “Analysis of GPU-Based Graph Processing Algorithms.”
Abstract: Parallel graph processing is central to analytical applications of computer science, and GPUs have proven to be an ideal platform for parallel graph processing. Existing GPU graph processing frameworks present performance improvements, but often neglect two fundamental issues: the unpredictability of a given input graph and the energy consumption of the graph processing. Our prototype software, EE- Graph (Energy Efficiency of Graph processing ) is a flexible system consisting of several graph processing algorithms with configurable parameters for vertex update synchronization, vertex activation, and memory resources along with a lightweight software-based GPU energy measurement scheme. The relationships between different configurations of our software, performance, and GPU energy, are observed for the processing of in-memory graphs and out-of-memory graphs. Through the analysis of the observed relationships, the ideal parameters are discovered for specific input graphs. We also motivate the utility of subgraph generation as a way to predict the performance and energy consumption of full graph configurations. EEGraph improves upon state-of-the-art GPU-based graph processing software by 2.08 times for performance and 1.60 times for GPU energy for in-memory graph datasets. Additionally, EEGraph improves GPU-based graph processing by 3.30 times for performance and 1.63 times for GPU energy on large out-of-memory graph datasets.