讲座编号：jz-yjsb-2017-y001

讲座问题：Automatic Systems Optimisation – How Computers Are Too Complex to Understand

主 讲 人：John Thomson  University of Edinburgh

讲座时间：2017年02月22日（周三）下昼14:00

讲座所在：阜成路东校区耕作楼809室

加入工具：计信学院西席、研究生
，接待其他感兴趣的西席或者研究生加入

主理单位：研究生院

承办单位：盘算机与信息工程学院

主讲人简介：

Dr. John Thomson got his PhD from University of Edinburgh and currently is a lecturer (assistant professor) and the director of postgraduate teaching in School of Computer Science, University of St Andrews, UK. His previous works include applying machine learning approaches on compiler optimisation and code generation, heterogeneous systems, video compress which published on top conferences including ACM ICML, ACM/IEEE CGO, Euro-Par and ACM MM. He has been an reviewer for top conferences including IEEE IPDPS, HiPEAC and journals including ACM TACO.

主讲内容：

Automatic Systems Optimisation – How Computers Are Too Complex to Understand.

   Since the invention of the microprocessor, systems problems have been becoming more complex. We design layer upon layer of abstraction on bare hardware, from compilers, to runtime systems to operating systems. Each of these layers continues to become more complicated independently, and the demands for faster processors push hardware towards its own complexity spiral.

This talk discusses how to exploit modern, intractable hardware and software toolchains by using machine learning to empirically model program behaviour in a system, rather than using analytical mathematical models, or heuristics.