2013 Poster Sessions : Reevaluating Memory Systems for Energy Efficiency: Bringing Computation and Data Closer

Student Name : Grant Ayers, Mingyu Gao
Advisor : Christos Kozyrakis
Research Areas: Computer Systems
During the last four decades, the exponential growth of computing capabilities has been possible primarily because of semiconductor technology. However, while the world’s demands for better and faster computation are only increasing, voltage scaling in semiconductors is no longer possible, making today’s systems energy-bound. While much attention has been given to improving the energy efficiency of processors, memory technologies have largely remained the same, accounting for up to 45% of the total power consumption of many systems. In order to continue improving computer performance, future memory systems must be more energy-efficient.

Our work is focused on reevaluating the memory system at an architectural level. Traditional cache hierarchies, address translation, and general data movement policies are not suitable for many workloads, including those with low locality or certain access and computation patterns. We plan to overcome this by bringing computation and data together in a smarter way. This involves adding computational resources close to the memory where possible, and improving the efficiency of data movement when memory-resident computation is infeasible. We expect that our work will improve performance and energy efficiency for a wide range of workloads in ways not previously possible.

Mingyu Gao is a first year Ph.D. student in the Department of Electrical Engineering, Stanford University. His research interests are Computer Architecture and System. Now he is working with Professor Christos Kozyrakis about energy efficient memory systems for heterogeneous multi-core chips. Also, he is interested in and has experience about high-performance computing, including FPGA, General-Purpose GPU, and specific platforms for ray-tracing. Before coming to Stanford, Mr. Gao received his Bachelor degree in Microelectronics in Tsinghua University, Beijing, China, in 2012.

Grant Ayers is a first-year Ph.D. student working with Professor Christos Kozyrakis as part of the Pervasive Parallelism Lab. He received a B.S. in Computer Engineering as well as an M.S. in Electrical and Computer Engineering from the University of Utah. His research interests include computer architecture and security.