Project Details
Description
Greybox Computing: An Associative Computing Methodology with Instruction Directed Power and Clock Management
Project Summary:
Energy consumption has become the bottleneck for several fastest growing applications such as wearable electronics, biomedical devices. Through more than a decade of improvement, the conventional low power design approaches have reached its maximum achievement bounded by the fundamental tradeoff of performance and power. As a result, new design paradigm needs to be established to continue the journey of energy saving. Conventionally, the hardware and software design are performed separately and treats each other as blackboxes. Different from traditional design approaches, this project aims at creating a “greybox” methodology where the information from hardware design of a microprocessor are visible to the higher level software compilers and vice versa. The additional information from different layers of design space can be used to establish a real-time power optimization operation. Our preliminary analysis shows that by breaking the boundary of traditional design space, significant energy saving that is not obtainable from conventional scheme can be achieved. To efficiently create an associative operation across the boundary of hardware and software, this project will jointly develop several key techniques such as cross mapping between instruction and hardware performance, enhanced power and clock management circuitry for instruction driven operation, calibration based software and hardware optimization. The associative computing methodology proposed in this project can significantly extend the power saving limitation of existing low power technology and provide a new design paradigm on hardware and software optimization.
Intellectual Merit Summary:
The proposed research creates a new systematic cross-layer design methodology that breaks the boundary of conventional design space. The proposed project will yield severa
Status | Finished |
---|---|
Effective start/end date | 7/1/16 → 6/30/21 |
Funding
- National Science Foundation (CCF-1618065-001)
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.