Slides available here!


Prof. Tohru Ishihara, Kyoto University, Japan


Near-Threshold Computing on Heterogeneous Multicore Architectures


Near-threshold computing is one of the most promising approaches for high-performance and energy efficient design of microprocessors. However, it is not well studied how the optimal transistor size in standard cells and the optimal pipeline structure in CPU cores change when the supply voltage of the microprocessors gets closer to the threshold voltage. In this talk, we first present an experimental observation that the optimal transistor size for near-threshold operation is different from the optimal transistor size for the nominal supply voltage in our target process technology. Based on this fact, the talk will next present our cell optimization flow which finds the optimal transistor sizes of individual standard cells operating at the near-threshold voltage. The experimental results show that, when operating at a 0.6V condition, the energy consumptions of several benchmark circuits synthesized with our standard cells optimized for the 0.6V condition can be reduced by 31% at the best case and by 23% on average compared with those of the same circuits synthesized with the cells optimized for the nominal supply voltage.


Tohru Ishihara received his Dr.Eng. degree in computer science from Kyushu University in 2000. For the next three years, he was a Research Associate in the VLSI Design and Education Center, the University of Tokyo. From 2003 to 2005, he was with Fujitsu Laboratories of America as a Research Staff of an Advanced CAD Technology Group. From 2005 to 2011, he was with Kyushu University as an Associate Professor. In April 2011 he joined Kyoto University, where he is currently with the Dept. of Communications and Computer Engineering. His research interests include low-power design methodologies and power management techniques for embedded systems. Dr. Ishihara is a member of the IEEE, ACM, IPSJ and IEICE. He has served on program committee of numerous conferences such as CODES+ISSS, ISLPED and so on.

* If you wish to modify any information or update your photo, please contact the web