CAREER: Beyond Approximate Computing: Enabling Full-System Energy-Quality Scalability in Embedded Systems

职业：超越近似计算：在嵌入式系统中实现全系统能源质量可扩展性

• 批准号: 1845469
• 负责人: Younghyun Kim
• 金额: $ 51.28万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-06-01 至 2025-05-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1845469&HistoricalAwards=false
• 关键词: CAREER; Beyond; Approximate; Computing; Enabling

Energy efficiency is a daunting challenge in embedded systems that run on limited energy budgets. Better performance, longer battery life, and smaller environmental footprints - improving energy efficiency will be the key enabler of applications and services that have not been possible in the past. Approximate computing has recently emerged as a promising approach to the energy-efficient computing of intrinsically error-tolerant applications like image processing, where small deviations from the exact results in the underlying computations do not substantially degrade the resulting application-level quality. For such applications, approximate computing can produce "just good enough" results to save energy at the cost of only minor or no quality loss.This project will develop design methodologies for taking advantage of approximate computing in embedded systems, where the contribution of non-computing subsystems (e.g., sensors, actuators, user interfaces, and network interfaces) to energy consumption is at least as significant as that of computing subsystems (e.g., microcontrollers and memory). In embedded systems, both computing and non-computing subsystems must be holistically considered to take full advantage of approximate computing and accomplish full-system energy quality and scalability. The project scope includes characterization, optimization, and design toolchain development, with the focus on embedded systems design.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

在能源预算有限的嵌入式系统中，能源效率是一项艰巨的挑战。更好的性能、更长的电池寿命和更小的环境足迹——提高能源效率将成为过去不可能实现的应用和服务的关键推动者。近似计算最近作为一种很有前途的方法出现，用于本质上容错的应用程序（如图像处理）的节能计算，在这些应用程序中，底层计算中与精确结果的小偏差不会大大降低最终的应用级质量。对于这样的应用，近似计算可以产生“刚刚好”的结果，以很少或没有质量损失为代价来节省能源。该项目将开发设计方法，以利用嵌入式系统中的近似计算，其中非计算子系统（例如，传感器，执行器，用户界面和网络接口）对能源消耗的贡献至少与计算子系统（例如，微控制器和存储器）一样重要。在嵌入式系统中，为了充分利用近似计算的优势，实现全系统的能量质量和可扩展性，必须从整体上考虑计算子系统和非计算子系统。项目范围包括表征、优化和设计工具链开发，重点是嵌入式系统设计。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

UNO: Virtualizing and Unifying Nonlinear Operations for Emerging Neural Networks

• DOI: 10.1109/islped52811.2021.9502473
• 发表时间: 2021-07
• 期刊: 2021 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED)
• 作者: Di Wu;Jingjie Li;Setareh Behroozi;Younghyun Kim

A Domain-Specific System-On-Chip Design for Energy Efficient Wearable Edge AI Applications

适用于节能可穿戴边缘人工智能应用的特定领域片上系统设计

• DOI: 10.1145/3531437.3539711
• 发表时间: 2022
• 期刊: Proceedings of the ACM/IEEE International Symposium on Low Power Electronics and Design
• 作者: Tuncel, Yigit;Krishnakumar, Anish;Chithra, Aishwarya Lekshmi;Kim, Younghyun;Ogras, Umit
• 通讯作者: Ogras, Umit

SynthNet: A High-throughput yet Energy-efficient Combinational Logic Neural Network

SynthNet：高吞吐量且节能的组合逻辑神经网络

• DOI: 10.1109/asp-dac52403.2022.9712554
• 发表时间: 2022
• 期刊: 2022 27th Asia and South Pacific Design Automation Conference (ASP-DAC
• 作者: Chen, Tianen;Kemp, Taylor;Kim, Younghyun
• 通讯作者: Kim, Younghyun

uBrain: a unary brain computer interface

• DOI: 10.1145/3470496.3527401
• 发表时间: 2022-06
• 期刊: Proceedings of the 49th Annual International Symposium on Computer Architecture
• 作者: Di Wu-;Jingjie Li;Zhewen Pan;Younghyun Kim

MIPAC: Dynamic Input-Aware Accuracy Control for Dynamic Auto-Tuning of Iterative Approximate Computing

• DOI: 10.1145/3394885.3431551
• 发表时间: 2021-01
• 期刊: 2021 26th Asia and South Pacific Design Automation Conference (ASP-DAC)
• 作者: Taylor Kemp;Yao Yao-Yao;Younghyun Kim