*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Breaking the Abstractions for Productivity and Performance in the Era of Specialization
Jongse Park
Ph.D. Candidate
School of Computer Science
College of Computing
Georgia Institute of Technology
Date: Monday, July 16, 2018
Time: 10:00AM - 12:00PM (EDT)
Location: Klaus 2100
Committee:
Dr. Hadi Esmaeilzadeh (Advisor, School of Computer Science, Georgia Institute of Technology)
Dr. Hyesoon Kim (School of Computer Science, Georgia Institute of Technology)
Dr. Tushar Krishna (School of Electrical and Computer Engineering and School of Computer Science, Georgia Institute of Technology)
Dr. Milos Prvulovic (School of Computer Science, Georgia Institute of Technology)
Dr. Nam Sung Kim (Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign)
Abstract:
Due to the ever-increasing number of connected devices, data is growing in an unprecedented and exponential rate. Emerging applications bring opportunities to extract insights from this explosion of data. However, this trend has coincided with the diminishing benefits from conventional transistor and microarchitecture performance scaling. We have entered the era of specialization where the hardware is being tailored and curated for a specific domain of application. However, designing hardware without considering the rest of the stack will not lead to the disruptive yet adoptable solutions that we need. As such, in this thesis, we explore breaking the traditional abstractions in the computing stack and define new programming models, system software layers, and hardware platforms that deliver orders of magnitude performance while providing low programming effort and ensuring productivity. We specifically consider two complementary specialization techniques, approximation and acceleration.
