Tuesday, November 29, 2022
Memory-Centric Computing

Abstract:
Computing is bottlenecked by data. Large amounts of application data overwhelm storage capability, communication capability, and computation capability of the modern machines we design today. As a result, many key applications' performance, efficiency, and scalability are bottlenecked by data movement. In this lecture, we describe three major shortcomings of modern architectures in terms of 1) dealing with data, 2) taking advantage of the vast amounts of data, and 3)  exploiting different semantic properties of application data. We argue  that an intelligent architecture should be designed to handle data  well. We show that handling data well requires designing architectures  based on three key principles: 1) data-centric, 2) data-driven, 3)  data-aware. We give several examples for how to exploit each of these  principles to design a much more efficient and high performance computing system. We especially discuss recent research that aims to  fundamentally reduce memory latency and energy, and practically enable  computation close to data, with at least two promising novels  directions: 1) processing using memory, which exploits analog  operational properties of memory chips to perform massively-parallel operations in memory, with low-cost changes, 2) processing near  memory, which integrates sophisticated additional processing capability in memory controllers, the logic layer of 3D-stacked memory  technologies, or memory chips to enable high memory bandwidth and low  memory latency to near-memory logic. We show both types of  architectures can enable orders of magnitude improvements in  performance and energy consumption of many important workloads, such  as graph analytics, database systems, machine learning, video  processing. We discuss how to enable adoption of such fundamentally more intelligent architectures, which we believe are key to  efficiency, performance, and sustainability. We conclude with some  guiding principles for future computing architecture and system  designs. A short accompanying paper, which appeared in DATE 2021, can be found  here and serves as recommended reading:  https://people.inf.ethz.ch/omutlu/pub/intelligent-architectures-for-intelligent-computingsystems-invited_paper_DATE21.pdf A longer overview & survey of modern memory-centric computing can be found  here and also serves as recommended reading: "A Modern Primer on Processing in Memory"  https://people.inf.ethz.ch/omutlu/pub/ModernPrimerOnPIM_springer-emerging-computing-bookchapter21.pdf


Biography:
Onur Mutlu is a Professor of Computer Science at ETH Zurich. He is also a faculty member at Carnegie Mellon University, where he  previously held the Strecker Early Career Professorship.  His current  broader research interests are in computer architecture, systems,  hardware security, and bioinformatics. A variety of techniques he,  along with his group and collaborators, has invented over the years  have influenced industry and have been employed in commercial  microprocessors and memory/storage systems. He obtained his PhD and MS  in ECE from the University of Texas at Austin and BS degrees in  Computer Engineering and Psychology from the University of Michigan,  Ann Arbor. He started the Computer Architecture Group at Microsoft  Research (2006-2009), and held various product and research positions  at Intel Corporation, Advanced Micro Devices, VMware, and Google.  He  received the Intel Outstanding Researcher Award, IEEE High Performance  Computer Architecture Test of Time Award, NVMW Persistent Impact Prize,  the IEEE Computer Society Edward J. McCluskey Technical Achievement  Award, ACM SIGARCH Maurice Wilkes Award, the inaugural  IEEE Computer Society Young Computer Architect Award, the inaugural  Intel Early Career Faculty Award, US National Science Foundation  CAREER Award, Carnegie Mellon University Ladd Research Award, faculty  partnership awards from various companies, and a healthy number of  best paper or "Top Pick" paper recognitions at various computer  systems, architecture, and security venues. He is an ACM Fellow "for  contributions to computer architecture research, especially in memory  systems", IEEE Fellow for "contributions to computer architecture  research and practice", and an elected member of the Academy of Europe  (Academia Europaea). His computer architecture and digital logic  design course lectures and materials are freely available on YouTube  (https://www.youtube.com/OnurMutluLectures ), and his research group  makes a wide variety of software and hardware artifacts freely  available online (https://safari.ethz.ch/). For more information,  please see his webpage at https://people.inf.ethz.ch/omutlu/.  

Mis à jour le 25 November 2022