Pro-Cache-CIM: A 28nm 69.4TOPS/W Product-Cache-based Digital-Compute-in-Memory Macro Leveraging Data Locality Pattern in Vision AI Tasks
Digital-based compute-in-memory macros (DCIM) and digital accelerators that leverage lookup tables (LUT) [1, 2, 3] have demonstrated remarkable advantages in achieving high energy efficiency while maintaining full precision. By pre-computing summation results, LUT-based DCIMs eliminate the energy consumption of one or two stages of addertrees. However, scaling these LUT-based DCIMs to higher orders-which requires using wider bit-widths for indexing LUT content-poses significant challenges. This is primarily due to the exponential growth in the combinations of pre-computed results, which consequently leads to proportional increases in area and power consumption, ultimately degrading both area and energy efficiency.
Authors
- Wenxun Wang (ORCID: https://orcid.org/0009-0007-1999-6441)
- Huazhong Yang (ORCID: https://orcid.org/0000-0003-2421-353X)
- Wenbin Jia
- Yongpan Liu (ORCID: https://orcid.org/0000-0002-4892-2309)
- Yaolei Li (ORCID: https://orcid.org/0000-0002-6640-1275)
- Jinshan Yue (ORCID: https://orcid.org/0000-0001-8234-7400)
- B.-S. Chen
- Hongyang Jia (ORCID: https://orcid.org/0000-0001-8692-1860)
- Xueqing Li (ORCID: https://orcid.org/0000-0002-8051-3345)
- Yifan He (ORCID: https://orcid.org/0009-0001-8664-037X)
- Xiang Li (ORCID: https://orcid.org/0000-0003-4215-6153)
- Y. H. Xie (ORCID: https://orcid.org/0000-0002-6031-3426)
- Zongle Huanq
Institutions
- Chinese Academy of Sciences (CN)
- Tsinghua–Berkeley Shenzhen Institute (CN)
- Institute of Microelectronics (CN)
- Tsinghua University (CN)
Publication Details
- Published
- 2025-04-13
- DOI
- https://doi.org/10.1109/cicc63670.2025.10983358
- Primary Topic
- Image Retrieval and Classification Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00