*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Dani Denton
Summary Sentence: The School of Computer Science and the School of Electrical and Computer Engineering Seminar - Scott Mahlke, University of Michigan
Full Summary: No summary paragraph submitted.
MahlkeTitle: Energy Efficient Computing is Easy if You Don't Care about Programmability
Speaker: Scott Mahlke, University of Michigan
Abstract:
In recent years, advances in semiconductor technologies have pushed the instrumentation of our world to unprecedented levels. Sensors are now all around us: cell phones contain GPS receivers, video cameras, and audio recorders, watches sense heart rates, and unmanned aerial vehicles flying overhead are constantly imaging the Earth. However, computing in this abundant data world has not kept pace in large part due to energy constraints. From data centers, where electricity costs dominate operating expenses, to mobile phones, where battery life is precious to all users, fundamental increases in energy efficiency are critical to drive abundant data computing. To scale energy efficiency, engineers have traditionally employed hardwired accelerators. Hardwired accelerators provide the most efficient hardware implementations of specific functionality at orders of magnitude lower power or higher performance compared to traditional processors, but at the cost of programmability. Such accelerators cannot execute a spectrum of applications, some of which have not even been written yet. Throughput processors provide an opportunity to bridge this gap by scaling efficiency while still providing a programmable substrate for data-parallel workloads that dominate abundant data processing including image/signal processing, natural language processing, computer vision, and deep learning. But throughput processors are not a panacea, commercial realizations in the form of graphics processing units (GPUs) face their own challenges of high peak power consumption and high software development costs where even the best experts struggle to extract only a small fraction of the peak performance. In this talk, I will examine the hardware/software challenges of designing and exploiting energy-efficient throughput processors. By rethinking the conventional GPU architecture with an energy-first approach and creating a source-to-source compiler toolchain to tame the programming complexity, a new generation of throughput processors can be created.
Bio:
Scott Mahlke is a Professor in the Electrical Engineering and Computer Science Department at the University of Michigan where he leads the Compilers Creating Custom Processors group (http://cccp.eecs.umich.edu). The CCCP group delivers technologies in the areas of customized processors for energy-efficient computing, reliable system design, and compiler code generation for heterogeneous systems. Prior to joining academia, Scott was a Senior Researcher at Hewlett-Packard Laboratories. He was one of the original contributors to both the OpenImpact and Trimaran compilers for VLIW processors. Scott's achievements were recognized by several awards, including the Morris Wellman Professorship in 2004, ACM SIGARCH/IEEE-CS TCCA Most Influential Paper Award in 2006, Young Alumni Achievement Award in 2007, Ted Kennedy Family Team Excellence Award in 2009, the EECS Outstanding Achievement Award in 2011, and the 2014 Monroe-Brown Foundation Education Excellence Award. Scott received the Ph.D. degree in Electrical Engineering from the University of Illinois at Urbana-Champaign in 1997 and is an IEEE Fellow.
