Single-Thread Performance in the Multi-core Era
Skidmore College
October 24, 2013
12:50-1:50
Olin 107
Lunch will be served at noon.

Abstract

The era of multi-core processors has begun and simultaneously represents a significant shift in processor design. This shift is a change in the design focus from reducing individual program (thread) latency to improving overall thread throughput. For over three decades, programs automatically ran faster on each new generation of processor because of improvements to processor performance. However, in this last decade, many of the techniques for improving processor performance reached their end. As a result, individual core performance has become stagnant, causing diminished performance gains for programs which are single-threaded.

Although the ability to improve a single core has decreased, designers now have the ability to produce multiple cores per processor. In fact, multi-core processors offer more thread throughput than ever before, and represent a large share of desktop processors in the market today. As such, multi-threaded programs written to exploit the increasingly parallel hardware will continue to achieve performance gains. In this talk, Professor Porter will explain why this shift in processor design occurred, how this shift impacts software engineers, and will outline some of the steps moving forward.

Bio: Leo Porter has been an Assistant Professor of Computer Science at Skidmore College since the Fall 2011. Before that he taught as a summer graduate research fellow at UC San Diego and was adjunct faculty at the University of San Diego. He his PhD on computer architecture at UC San Diego. In addition to computer architecture his research interests include computer science education.


Please email Kristina Striegnitz (striegnk@union.edu) if you have any questions concerning the seminar series or if you would like to receive the seminar announcements by email.