Skip to content Skip to navigation

Qualifying Exam: An Effective Graph Edge Partition Model for GPU Computing


Graph edge partition models have become an appealing alternative to graph vertex partition models for parallel computing, largely due to their flexibility in balancing workloads and their efficiency in reducing communication cost. We propose a graph edge partition model with high partition quality while maintaining low partition overhead. Our edge partition model places an emphasis on data; computation is modeled as interaction between data. We demonstrate the benefits of the graph edge partition model for GPU computing on an important class of applications processing sparse matrices and graphs. Our model can balance work, and also data, which enhances locality over space and time respectively.

Yanhao Chen
CoRE B (305)
Event Date: 
09/30/2019 - 11:00am
Prof. Zheng Zhang (Chair), Prof. Desheng Zhang, Prof. Ulrich Kremer , Prof. Martin Farach-Colton
Event Type: 
Qualifying Exam
Dept. of Computer Science