Who’s Xiaoming Chen
March 1st, 2022

Xiaoming Chen
Associate Professor
Institute of Computing Technology,
Chinese Academy of Sciences
Email:
chenxiaoming@ict.ac.cn
Personal webpage
http://people.ucas.edu.cn/~chenxm
Research interests
EDA and computer architecture
Short bio
Xiaoming Chen received the B.S. and Ph.D. degrees in Electronic Engineering from Tsinghua University, in 2009 and 2014, respectively. Since 2017, he has been an associate professor at Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS). Before joining ICT, CAS, he was a postdoctoral research associate in Electrical and Computer Engineering, Carnegie Mellon University from 2014 to 2016, and a visiting assistant professor in Computer Science and Engineering, University of Notre Dame from 2016 to 2017.
His research interests are mainly focused on EDA and computer architecture. He has published about 100 papers in top journals and conference proceedings, including DAC, ICCAD, DATE, HPCA, IEEE TCAD, IEEE TVLSI, IEEE TPDS, etc. He has served as a technical program committee member for DAC, ICCAD, ASP-DAC, GLSVLSI, AsianHOST, VLSI Design, etc. He was awarded the Excellent Young Scientists Fund of National Natural Science Foundation of China in 2021. He received the 2015 EDAA Outstanding Dissertation Award and the 2018 Alibaba DAMO Academy Young Fellow Award. He received one of the two best paper awards in ASP-DAC 2022 and several best paper nominations in ASP-DAC and ISLPED.
Reasearch highlights
Prof. Xiaoming Chen has spent more than 10 years in the EDA trarea. Specifically, he has developed a parallel sparse direct solver named NICSLU that is well suited for SPICE-based circuit simulators. He proposed a series of novel techniques to elevate the performance of solving highly sparse linear systems from circuit simulation applications, including a new matrix ordering method to minimize fill-ins, a hybrid dynamic scheduling method for parallel matrix factorization, a numerically stable pivoting reduction technique, and an adaptive numerical kernel selection method. NICSLU achieves much higher performance than other sparse solvers in circuit simulation applications, and is also generally faster than state-of-the-art GPU-based solvers which are specially designed for circuit matrices. NICSLU has been used in a number of academic studies, EDA tools and power system simulators. Some techniques have been adopted in commercial SPICE tools of a leading EDA company in China. NICSLU is available at https://github.com/chenxm1986/nicslu.
Prof. Chen has also made important contributions in computing-in-memory (CiM) architecture design. He exposed how to utilize the device-level CiM feature of resistive random-access memories (RRAMs) and ferroelectric field-effect transistors (FeFETs) which can act as both storage elements and switch units, to unify the computing and storage resources at the circuit level, to realize interchangeable computing and storage functionalities at the architecture level. In addition, he has investigated the solutions to some fundamental problems in CiM systems, including data coherence, data contention, simulation methodologies, task assignment, etc. He has also explored the CiM feature of RRAMs and FeFETs in various applications, and designed energy-efficient and high-performance accelerators for neural networks, graph processing, linear algebra, and robots.