Computer Engineering Cider Seminars

Past Seminar

Computer Architecture Research in the CMU Impetus Group

Babak Falsafi
Carnegie Melon University
July 15, 2003
10AM-11AM, Room GB248

Cider Seminar HomePage

Abstract

Advances in semiconductor fabrication have given rise to tremendous levels of integration. These advances have also been accompanied by a number of key design challenges requiring: (1) architectural innovation to to exploit the abundance of chip transistors to reduce single (sequential) program execution time, to bridge the ever-growing performance gap between the processor and memory, and to enhance performance in the presence of deep-submicron CMOS circuit bottlenecks; and (2) tools that allow for accurate and rapid evaluation of novel designs in the presence of these key challenges.

In this talk, I will first give an overview of the projects in the Impetus Group that tackle the above design challenges. I will then discuss in detail two recent contributions that were presented at the flagship architecture conference (ISCA) this year: (1) Implicitly-Multithreaded processors that extract instruction sequences dynamically from a single (sequential) program and speculatively execute them on a modified Simultaneous Multithreading (e.g., Intel Pentium 4) processor to reduce execution time, and (2) a Sampling Microarchitecture Simulation framework that applies rigorous statistical sampling for fast and accurate simulation of modern microarchitecture.

Biography

Babak Falsafi is an Associate Professor in the Electrical and Computer Engineering Department at Carnegie Mellon University. Prior to joining Carnegie Mellon, he held a position as an Assistant Professor in the School of Electrical and Computer Engineering at Purdue University. His research interests include prediction and speculation in high-performance memory systems, power-aware processor and memory architectures, single-chip multiprocessor/multi-threaded architectures, and analytic and simulation tools for computer system performance evaluation. He has made several contributions in the design of distributed shared-memory multiprocessors and memory systems, including a recent result indicating that hardware speculation can bridge the performance gap among memory consistency models, and an adaptive and scalable caching architecture, Reactive NUMA, that lays the foundation for a family of multiprocessors built by Sun Microsystems code-named WildFire. He is a recipient of an NSF CAREER award in 2000 and IBM Faculty Partnership Awards in 2001 and 2003, and a University of Virginia Top Gun distinguished lecturer in 2002. You may contact him at babak@ece.cmu.edu (http://www.ece.cmu.edu/~babak).