Graph Processing - Efficient Acceleration of Graphs at Low Energy

Abstract

Graph processing has been a recent topic of interest in the high performance computing and systems community in the context of data analytics. Graph processing presents difficulties in data-level parallelism and data-layout, which depend on the graph structure. In this talk, I will describe our current process of designing an energy efficient graph processor, with the purpose of enhancing graph analytics. I will describe insights discovered during this design process, the hardware challenges of graph processing, and what role do high level frameworks play in this design.

Alon Amid

Hardware/Software Engineer