|Title:||Embracing Errors in Computing Systems - from Timing Speculation to Approximate Computing|
|Date:||November 3, 2017 (Friday)|
|Time:||4:00 p.m. - 5:00 p.m.|
|Venue:||Room 121, 1/F, Ho Sin-hang Engineering Building,
The Chinese University of Hong Kong,
|Speaker:||Prof. Qiang XU
Department of Computer Science and Engineering
The Chinese University of Hong Kong
Conventional integrated circuit (IC) designs try all means to achieve error-free computation, even under worst-case combinations of process, voltage, and temperature (PVT) variations and wearout effects. As the above circuit non-idealities inevitably worsen with technology scaling, more design guardband has to be incorporated to ensure IC timing correctness. Consequently, such worst-case design methodology results in pessimistic designs with considerable power and performance overheads, lessening the benefits provided by technology scaling. On the other hand, emerging Recognition, Mining and Synthesis (RMS) applications demonstrate good intrinsic error-resilience property. They process noisy and redundant data sets obtained from non-traditional input sources such as various types of sensors (inexact inputs) and the associated algorithms are often stochastic in nature (e.g., iterative algorithms). Moreover, these applications usually do not require computing a unique or golden numerical result ("acceptable" instead of precise outputs). Consider two different classifiers that produce similar classification results on a set of example objects. It is very difficult, if not impossible, to tell which one is "better" for the classification of new objects.
With the above, in this talk, we present timing speculation techniques for better-than-worst-case (BTWC) design at circuit level and approximate computing techniques that relax the numerical equivalence between the specification and implementation of error-tolerant applications. By embracing errors in computing systems, we are able to achieve significant savings in energy and/or improvements in performance.
Qiang Xu is an associate professor of Computer Science & Engineering at the Chinese University of Hong Kong. He leads the CUhk REliable laboratory (CURE Lab.), and his research interests include electronic design automation, fault-tolerant computing and trusted computing. Dr. Xu has published 150+ papers at referred journals and conference proceedings, and received two Best Paper Awards and five Best Paper Award Nominations. He is currently serving as an associate editor for IEEE Transaction on Computer-Aided Design of Integrated Circuits and Systems and Integration, the VLSI Journal.
Enquiries: Mr. Calvin Tsang at tel. 3943 8440
For more information, please refer to http://www.cse.cuhk.edu.hk/seminar.