Title:
COMPUTING-IN-MEMORY EDRAM ACCELERATOR FOR CONVOLUTIONAL NEURAL NETWORK
Document Type and Number:
WIPO Patent Application WO/2023/056779
Kind Code:
A1
Abstract:
A computing-in-memory eDRAM accelerator for a convolutional neural network, which is characterized in that same comprises four P2ARAM blocks; each P2ARAM block comprises a 5T1C ping-pong eDRAM bit cell array consisting of 64x16 5T1C ping-pong eDRAM bit cells; in each P2ARAM block, 64X2 digital time converters convert 4-bit activation values into different pulse widths in a row direction, and inputs same to the 5T1C ping-pong eDRAM bit cell array for calculation; and a total of 16X2 convolution result outputs are obtained in a column direction of the 5T1C ping-pong eDRAM bit cell array. The convolutional neural accelerator provided in the present invention performs parallel multi-bit storage and convolution by using the 5T1C ping-pong eDRAM bit cells; and an input sampling capacitance of a cumulative bit line is shared to symbol-numerical SAR ADC cells of a CDAC array without additional area overhead, and an S2M-ADC solution is provided. By using the described means, a peak computational density of a computing-in-memory eDRAM accelerator based on the convolutional neural network is 59.1 TOPS/mm2.
More Like This:
| JP2000232485 | SIGNAL DELAY CIRCUIT |
| JPS629430 | DATA BUFFER CONTROL SYSTEM |
| JP2001092778 | DIGITAL SIGNAL PROCESSOR |
Inventors:
ZHANG HONGTU (CN)
SHU YUHAO (CN)
HA YAJUN (CN)
SHU YUHAO (CN)
HA YAJUN (CN)
Application Number:
PCT/CN2022/110410
Publication Date:
April 13, 2023
Filing Date:
August 05, 2022
Export Citation:
Assignee:
UNIV SHANGHAI TECH (CN)
International Classes:
G06F5/16; G06N3/063
Foreign References:
| CN113946310A | 2022-01-18 | |||
| CN114281301A | 2022-04-05 | |||
| US20200310758A1 | 2020-10-01 | |||
| US20200117986A1 | 2020-04-16 |
Other References:
CHEN ZHENGYU, CHEN XI, GU JIE: "15.3 A 65nm 3T Dynamic Analog RAM-Based Computing-in-Memory Macro and CNN Accelerator with Retention Enhancement, Adaptive Analog Sparsity and 44TOPS/W System Energy Efficiency", 2021 IEEE INTERNATIONAL SOLID- STATE CIRCUITS CONFERENCE (ISSCC), IEEE, 13 February 2021 (2021-02-13) - 22 February 2021 (2021-02-22), pages 240 - 242, XP055980894, ISBN: 978-1-7281-9549-0, DOI: 10.1109/ISSCC42613.2021.9366045
Attorney, Agent or Firm:
SHANGHAI SHENHUI PATENT AGENT CO., LTD. (CN)
Download PDF: