Faculty Candidate Talk
(Live-streamed Event) Smarter Hardware Prediction Mechanisms
Thursday, March 12, 2020, 10:30am
****IMPORTANT NOTICE: Due to the measures determined by the university in regards to the COVID-19 virus, this Faculty Candidate Talk has been transitioned to a virtual, live-streamed talk.
You can view the event via Webex using the following information:
Meeting Link: https://rutgers.webex.com/rutgers/j.php?MTID=m842af9baec112413a7c476bb38dd638f
Meeting Number: 797 361 728
Meeting Password: tSC2x63VynB
Speaker: Akanksha Jain, University of Texas at Austin
Akanksha Jain is a Research Associate at the University of Texas
at Austin. She received her PhD in Computer Science from The
University of Texas in December 2016. In 2009, she received the
B.Tech and M. Tech degrees in Computer Science and Engineering
from the Indian Institute of Technology Madras. Her research
interests are in computer architecture, with a particular focus
on the memory system and on using machine learning techniques
to improve the design of memory system optimizations.
Location : Via Webex
Event Type: Faculty Candidate Talk
Abstract: The performance of many applications is limited by large memory access latencies. In this talk, I will present significant advances to two foundational techniques---caching and prefetching---that aim to mitigate this bottleneck. A common theme is the use of idealized design points, which at first glance seem impossible or infeasible, as a key building block of highly effective and practical solutions. I will start by presenting a cache replacement policy that leverages Belady's optimal but clairvoyant algorithm in the design of a practical solution. This policy won the second Cache Replacement Championship, and it inspired us to create new optimal algorithms for complex scenarios. I will then briefly present a series of work in irregular data prefetching that makes a prohibitively expensive prefetching technique practical via new data representations. Finally, I will discuss the role that machine learning can play in improving hardware systems, and I will present a novel approach that uses machine learning as a white-box tool to build practical hardware predictors. I will conclude by discussing my vision for intelligent, high-performant memory systems for future hardware systems.
Contact Faculty Host: Santosh Nagarakatte