Welcome to the ENPOWER (Elastic and Nonstationary POWER) project website
ENPOWER (Elastic and Nonstationary POWER) investigates energy proportional computing techniques for reconfigurable devices that include an FPGA fabric and embedded microprocessors. This three and a half year project sponsored by EPSRC is a collaboration among the University of Bristol, Queens University of Belfast, Xilinx and Qioptiq. The main idea is to deploy automatic fine grain control of the FPGA parameters that impact power and energy requirements. The design paradigm includes full scalability (i.e. capacitance, voltage and frequency) in a variation-aware, closed-loop configuration which is exposed to, and can be changed by the application software.
The application software is written in OpenCL extended to allow the programmer to request different levels of performance or power. The OpenCL run-time decides during execution which is the best configuration point to achieve these requirements based on energy availability, thermal management, resource availability, etc. Each configuration point is defined by a triplet of voltage, frequency and capacitance. Modern FPGAs enable, thanks to partial reconfiguration, changes to the logic at run-time that can be used to implement virtual power gating techniques.
Additionally, ENPOWER deploys a technology created at Bristol called Elongate to achieve fine grain control of voltage and frequency with embedded in-situ detectors forming a closed-loop control system. Initial results with 28-nm Zynq chips from Xilinx have shown that the same silicon can deliver elastic energy and performance points ranging from 100% better performance to 70% lower energy without changes to the device itself. These concepts could be potentially applied to other heterogenous processing platforms (e.g. CPU-GPU, big.LITTLE processing) extended with adequate hardware support. ENPOWER started in November 2013 and in 2014, July our work on energy scaling received the best paper award in the NASA/ESA conference on adaptive hardware systems (check the publications section to get the paper) . You can click the navigation links on the left to learn more or contact:
Dr Jose Luis Nunez-Yanez (email: firstname.lastname@example.org) (phone: +44 117 3315128)