October 28, 2016
Congratulations to Tae Jun Ham (Princeton University), Lisa Wu (University of California, Berkeley), Narayanan Sundaram (Intel), Nadathur Satish (Intel), and Margaret Martonosi (Princeton University) who received the Best Paper Award at the 9th Annual ACM/IEEE International Symposium on Microarchitecture for their paper Graphicionado: A High-Performance and Energy-Efficient Accelerator for Graph Analytics.
The paper describes a specialized-yet-flexible approach to designing hardware accelerators for graph analytics problems. Graph analytics pertains to analyzing graphs which come up in analyzing social networks, analyzing financial flows, and many other settings. Their approach allows specialization of the vertex processing kernels in a graph problem, and it provides techniques for graph algorithms to more efficiently access memory at high performance. The paper’s results show that Graphicionado achieves a 1.76−6.54× speedup while consuming 50−100× less energy compared to a state-of-the-art software graph analytics processing framework executing 32 threads on a 16-core Haswell Xeon processor.
Professor Martonosi gave the keynote address tiled “Internet of Things: History and Hype, Technology and Policy". It was the first time in the 49 year history of the conference that a female had been given that honor.