Project 154: Logic Synthesis for Large-Scale Approximate Circuits and Its Application in Deep Neural Network Accelerators

Contact Information:

Assoc. Prof. Weikang Qian

Email: qianwk@sjtu.edu.cn

Project Description and Objectives:

Designing approximate circuits is considered as an effective way to further improve circuit performance and energy efficiency in the post-Moore era. Its key idea is to exploit the error tolerance of applications so that we can design circuits with reduced area, delay, and power consumption by deliberately sacrificing a small amount of the accuracy. In order to design an approximate circuit automatically under any given specification, research on approximate logic synthesis (ALS) is attracting more and more attention recently. This project will develop a new ALS algorithm for large-scale digital circuits. Furthermore, with the new ALS tool available, we will also apply it to build hardware accelerators for an important error-tolerant application, the deep neural network.

Eligibility Requirements:

Be self-motivated and hard working.

Always pursue high quality work.

Good at programming.

Main Tasks:

Read the related papers; develop new ideas; implement the new idea; improve the proposed idea and the implementation; write reports.

Website:

Lab: https://umji.sjtu.edu.cn/~wkqian/ 

School: https://www.ji.sjtu.edu.cn/