The exponential growth in processor performance and storage capacity following Moore's law has been the fundamental driver of computing for the past three decades. However, with the stop of Dennard's voltage scaling, the power consumption of the additional transistors in new technology generations can no longer be mitigated through circuit-level techniques. The post-Dennard power limitations, coupled with the limited and nonscalable off-chip bandwidth, highlights the need for novel power- and bandwidth-efficient architectures for next generation processors.
Motivated by these emerging semiconductor technology challenges, our research is focused primarily on the exploration of high-performance and low-power computer architectures with a current emphasis on highly parallel computer architectures, interconnection networks, and in-memory processing. Our solutions target emerging applications from various compute-intensive fields, including scientific computing, embedded healthcare and medical applications, big data analytics, multimedia, and deep learning.
Active research projects in the group, and areas in which we have made significant accomplishments include the following topics:
-- Neural Network Acceleration
-- Near Data Processing