Brown University School of Engineering

Electrical and Computer Engineering Seminar: Game-Like Design Tools for EDA: New Architectures and Algorithms through Crowdsourcing

Add event to my Google calendar Add event to my Google calendar Share this event on facebook E-mail this event
Monday, November 11, 2013 10:30am - 11:30am

Gayatri Mehta, University of North Texas, Denton. Chip architectures that are customized to an application domain allow designs to achieve optimal efficiency and flexibility in a tiny low power package. However, it is extremely difficult to find optimal tradeoffs in designing a new architecture, or even to fully understand the design space. My vision is to develop a highly visual environment for exploring custom domain-specific architectures. Furthermore, the environment should be so accessible that even non-engineers and non-scientists can contribute to exploring architectural designs and uncovering efficient processes for mapping algorithms onto novel architectures. I will show that human reasoning and problem solving skills can be brought to bear in solving real problems in EDA. In this talk, I will present our interactive game UNTANGLED, which focuses on the problem of mapping algorithms onto coarse-grained reconfigurable architectures. I will show that our players can outperform and improve upon state of the art algorithms. I will also show that the strategies they use can inspire even better algorithms. In particular, I will discuss our pattern-based Anytime A* algorithm, which obtains an order of magnitude speedup over conventional A* by exploring placements of connected collections of nodes in commonly observed patterns. Gayatri Mehta is an Assistant Professor in the department of Electrical Engineering at the University of North Texas, Denton, TX. She received her Ph.D. in Electrical and Computer Engineering from the University of Pittsburgh in 2009 and is a recipient of a Junior Faculty Summer Research Fellowship at the University of North Texas in 2010. Her research interests are broadly in the areas of Electronic Design Automation, Reconfigurable Computing, Low-Power VLSI Design, System-on-Chip Design, Embedded Computing, and Portable/Wearable Computing.