Advanced Processor Architectures

We investigate high performance and low power architectures for data centers, architectures that may enable future big-data machine learning, and processing-in-storage architectures. High performance low power is studied in the context of manycore systems, where many small and slow processors can collectively achieve the same performance as a single state-of-the-art processor but at significantly lower power consumption. At the extreme end, an implantable manycore processor was studied with power dissipation of less than one micro-Watt, appropriate for battery-less cardiac implants. The investigation of big-data machine learning has led to the study of associative parallel processing. New technology for non-volatile memory integrated with the processor has triggered research at the center focused on creating arrays of small processors and small memories, and applying them to bio-informatics and to deep learning.