Events
Seminar @ Cornell Tech/ECE Colloquium: Karu Sankaralingam
Role of Silicon Innovation in Advancing Deep Learning
GPUs have been relentlessly advancing the state-of-art in hardware and software for deep-learning. The SIMT design and Tensorcores seem well suited for deep-learning. In this talk, I will explore what else might be possible and whether substantial improvements can be made. Dataflow seems quite promising while having its own challenges. I will describe some of my research group’s efforts in understanding the limits of dataflow computation from a compiler, architecture, and workload perspective as applied to deep learning. In particular, I will delve into some depth on our constraint theory-based compiler, our stream dataflow architecture, and workload analysis on end-to-end deep-learning applications. I will touch upon some of the lessons learned from our commercial production silicon implementation and software efforts.
Speaker Bio
Karu Sankaralingam is a Professor at UW-Madison. Based on 5 years of fundamental research, he founded SimpleMachines in 2017 which developed chip designs applying dataflow computing to push the limits of AI generality in hardware. In his career, he has led three chip projects: Mozart (16nm, HBM2 based design), MIAOW open source GPU on FPGA, and the TRIPS chip as a student during his PhD. In his research he has pioneered the principles of dataflow computing, focusing on the role of architecture, microarchitecture and the compiler. His research breakthroughs include constraint-theory based compilation for spatial architectures, specialized datapaths that can be dynamically configured, hybrid dataflow von-Neumann execution, modularizing specialization principles to allow programmability while retaining specialization benefits, new dataflow execution models that combine streaming and dataflow, and sampling theory applied to reliable computing. His work has been featured in industry forums of Mentor and Synopsys, and has been covered by the New York Times, Wired, and IEEE Spectrum. He has published over 100 research papers, has graduated 9 PhD students, is an inventor on 21 patents, and 9 award papers. He is a recipient of the Vilas Faculty Early Career Investigator Award in 2018, IEEE TCCA Young Computer Architecture Award in 2012, an NSF CAREER award in 2009, the Emil H Steiger Distinguished Teaching award in 2014, the Letters and Science Philip R. Certain – Gary Sandefur Distinguished Faculty Award in 2013 which recognizes outstanding teaching by a member of the College of The Letters and Science.