Increasing Efficiency and Scalability with Tunable Error Protection and Memory Access Granularity

Thursday March 18, 2010
12:00 noon

Mattan Erez
UT Austin ECE


I will present my group's vision and initial work towards enabling more efficient and scalable systems through dynamic and flexible cooperative error protection and variable granularity memory systems. Flexible protection recognizes that different computations and data require different degrees of protection, and hence different amounts of resources for acceptable execution outcomes. Variable granularity memory systems relax the trend of ever-increasing access granularities and improve utilization of scarce memory bandwidth and power resources. While the two topics may seem unrelated, they are in fact closely tied to one another and share many of the same enabling mechanisms.

I will present motivating examples for when to apply cooperative protection and variable granularity, as well as their overall potential given current system design trends. I will then discuss our initial set of hardware mechanisms, and explain why dynamic protection and access granularity are intimately related. These mechanisms involve a range of tradeoffs with respect to redundancy, storage capacity, error-tolerance complexity, and expected performance. I will conclude by discussing how to best expose the hardware flexibility to the programmer and software and a framework that will enable exploring the tradeoff space.


Mattan Erez is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Texas at Austin. Mattan received a B.Sc. in Electrical Engineering and a B.A. in Physics from the Technion, Israel Institute of Technology in 1999. He subsequently received his M.S and Ph.D. in Electrical Engineering from Stanford University in 2002 and 2007 respectively. His experience includes working as a computer architect in the Israeli Processor Architecture Research team, Intel Corporation. As a Ph.D. candidate at Stanford University he was the student leader of the Merrimac Stanford Streaming Supercomputer project, where his work spanned the entire system from microarchitecture to the Brook and Sequoia programming models. Mattan is continuing this whole-system approach at UT Austin, where his research focuses on improving performance and efficiency through advances in processor architecture and programming models and techniques.

Back to the seminar page