Upcoming Talks
Professor Trevor Mudge
Bredt Family Professor of Engineering
Advanced Computer Architecture Laboratory
Electrical Engineering and Computer Science Department
The University of Michigan, Ann Arbor
When: June 19, 2009 | 2:00pm-3:30pm
Where: Bldg 904, 6th floor, 6D-000
"Two Projects in Low Power Computing"
Abstract:
This talk will cover two very different projects currently being undertaken in my group. Their common denominator is low power operation.
1. Baseband Processors for Smart Phones - A High Performance Embedded Challenge
Wireless communication protocols today have a computationally demanding workload that has to be supported by mobile terminals. The need for these terminals to be small and portable dictates a highly constrained energy budget. Traditionally these goals were satisfied by ASIC solutions. However, reusable hardware platforms and the need to interoperate between a wide range of protocols has led to research into programmable hardware platforms that can support these protocols in software. We will overview our work in this area.
2. Femtoserver - Building a Compact Energy Efficient Multiprocessor
With power and cooling becoming an increasingly costly part of the operating budget of a server, the old trend of striving for higher performance with little regard for power is over. Emerging semiconductor process technologies, multicore architectures, and new interconnect technology provide an avenue for future servers to become low power, compact, and possibly mobile. In talk we examine three techniques for achieving low power:
-Near threshold operation;
-3D die stacking; and
-replacing DRAM with Flash memory
In this talk, we overview the case for a Femtoserver, a compact energy efficient multicore architecture
Upcoming Talks
Professor Arvind
Johnson Professor of Computer Science and Engineering
Massachusetts Institute of Technology
When: June 24, 2009 | 10:30am-12:00pm
Loc: Bldg 904, 6th floor, 6D-000
"Cycle-Accurate Modeling of Processors on FPGAs"
Abstract:
Cycle-accurate models are required for the development of new microarchitectures, often way before the RTL for implementation is available. Such models must be flexible enough so that many variants can be studied and fast enough so that appropriate software stack can be run on it for the desired studies. Models should also be easier to develop than the actual machine. Currently cycle-accurate models are built primarily in software and often run 5 to 6 orders of magnitude slower than the target architectures. We, along with groups at Intel and UT Austin, are developing cycle-accurate models to be run on FPGAs, with the goal of a 1000-fold speed up over equivalent software models. In this talk we will describe our approach to cycle-accurate modeling. Our approach is based on latency-insensitive bounded dataflow networks (LI-BDNs) and facilitates the ability to refine parts of a model for efficient implementation on FPGAs without compromising the cycle-accuracy of the model.
An MIT-IBM project, partially sponsored by Nancy Greco at IBM, to develop a cycle-accurate model of an in-order multithreaded, multicore PowerPC is currently underway. The team consists of Jessica Tseng and K. Ekanadham from IBM Research, and Asif Khan, Murali Vijayaraghavan and others in my group at MIT. This talk will cover two very different projects currently being undertaken in my group.
Bio:
Arvind is the Johnson Professor of Computer Science and Engineering at MIT where in the late eighties his group, in collaboration with Motorola, built the Monsoon dataflow machines and its associated software. In 2000, Arvind started Sandburst which was sold to Broadcom in 2006. In 2003, Arvind co-founded Bluespec Inc., an EDA company to produce a set of tools for high-level synthesis. In 2001, Dr. R. S. Nikhil and Arvind published the book "Implicit parallel programming in pH". Arvind's current research focus is on enabling rapid development of embedded systems. Arvind is a Fellow of IEEE and ACM, and a member of the National Academy of Engineering. www.csg.csail.mit.edu/Users/arvind
