Processing-In-Memory Circuits and Systems
Circuit Design, Domain-Specific/General Architecture
chips
Assistant Professor Institute for Artificial Intelligence · Peking University
燕博南 Ph.D. · Compute-in-Memory · AGI Processor Silicon
Assistant Professor, Institute for Artificial Intelligence, Peking University.
Prof. Bonan Yan (燕博南), Ph.D. is an Assistant Professor at Institute for Artificial Intelligence, Peking University. He received his Ph.D. degree in Electrical and Computer Engineering from Duke University in 2020. His research focuses on artificial general intelligence processor chip design and integrated circuits design for emerging memory and computing-in-memory—primarily targeting at SRAM, RRAM, MRAM, PCM, flexible electronics, and monolithic 3D integration.
He has published papers in international journals and conferences, including Nature, Nature Electronics, Nature Communications, ISSCC, IEDM and DAC. He was awarded the ACM SIGDA Outstanding New Faculty Award in 2024 and the CCF-Huawei Populus Euphratica Forest Fund Outstanding Proposal Award in Memory Systems from China Computer Federation (CCF). He serves as a TPC member for DAC, DATE, EDTM, etc..
His team developed a digital SRAM-based computing-in-memory AI processor chip. He is a co-inventor of a WASM hardware accelerator processor and a fully-flexible compute-in-memory LSI chip.
Circuit Design, Domain-Specific/General Architecture
chips
Perception, Reasoning, Cognition, Massive Multi-Agents Reinforcement Learning
chips
Duke University
University of Pittsburgh · Pittsburgh, PA, USA
Beihang University · Beijing, China
Peking University
Association for Computing Machinery (ACM)
Design, Automation and Test in Europe Conference (DATE)
Peking University Education Foundation
Peking University
China Computer Federation (CCF)
Peking University
Peking University
Peking University
International Symposium of Integrated Circuits (ISIC)
Peking University · Science and Technology Track
Peking University
11 / 11 items · verbatim CV order
Undergraduate
Graduate
Venue wordlines
End of array · 61 cells · titles & full author lists all present in the DOM
open-source PIM macro Verilog/Chisel behavioral model library
bonany.gitlab.io/pis/vanilla self-attention block for large language model accelerators
bonany.cc/attentionlego/probabilistic circuits — Beihang University, Beijing, China (collaboration with Dr. Anji Liu)
liuanji.github.io「Let them help you!」
notes — bonany.cc/asset/phdguide.mdGreat talk of Richard W. Hamming (Hamming code's inventor)
here — bonany.cc/asset/youandyourresearch.mdUseful tool :D
here — bonany.cc/asset/verilatorverilog.mdCurrent Group Members · 9
memristor-based compute-in-memory for graph processing
SRAM compute-in-memory microarchitecture
RRAM compute-in-memory IC design
probabilistic computing IC Design
cognitive architecture processor
WASM processor
intelligent sensing & processing
reinforcement learning processor
3D integrated circuit design & architecture
The enrollment quota for recruit in 2027 is full
请认真上课。
想体验research的二年级同学,建议在大二下(春季学期)五一假之前联系我,可以立项「本研训练项目」。
欢迎想保研的同学来陶瓷,陶瓷信发这里bonanyan[at]pku[dot]edu[dot]cn,告诉我你有多强+想研究的方向。据往年经验竞争会比较激烈,请务必在三年级尽早联系,越早越好。我的考核方式是会分配一个小的project一起研究一下,这个过程来双向选择一下。
招生标准
花大量时间在研究(学习、适应、思考、动手)上
以隔周为单位,有显著进展;肯下手一点一点实现(所谓的「keep your hands dirty」),而不是只想不做
遇到极小困难自己google,遇到小到中等困难自己解决,遇到重大困难第一时间跟我说一起寻求解决方案;我平常杂事比较多,需要每周催我排时间和你一对一meet
遇到困难,从最naive到最花里胡哨可以提出各种可能的解决方案,实在手足无措的时候才算是「重大困难」(重大困难一定要尽早约我一对一meet讨论解决方案)
本科四年级是工作或者读研第0年,一定要珍惜可以加油的机会!去跟着未来老板或导师学习吧
在导师明确同意的情况下,欢迎研究生同学来找我一起research,邮件我你的情况bonanyan[at]pku[dot]edu[dot]cn。
我的研究的重要性总体逻辑:存算一体通用人工智能芯片。先读一下这两篇文章,看看是不是喜欢:我写的一篇综述、一篇经典文章。再看一下我这里的两个demo,看看是不是有自己想做一套的冲动:RRAM CIM Chip Demo、SRAM CIM Chip Demo
Compute-in-memory readout IDLE · operands in cells
Email operands (obfuscated at rest)
Result register— — — — —
Postal / affiliation
Institute for Artificial Intelligence
Peking University
Prospective Ph.D. and undergraduate applicants: read U11 · Advice first — the 2027 enrollment quota is full.