A 28-nm 135.19 TOPS/W Bootstrapped-SRAM Compute-in-Memory Accelerator With Layer-Wise Precision and Sparsity
IEEE Transactions on Circuits and Systems I: Regular Papers · 2025-07 · 72 (7) · 3236-3246
DOI: 10.1109/tcsi.2024.3486322 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems I: Regular Papers, 2025-07. Verfasst von Wei Mao und 9 weiteren Autor:innen. Die Arbeit wurde bisher 13-mal zitiert (Crossref) und nennt 45 Literaturangaben. Gefördert von: Shenzhen Science and Technology Program, Science and Technology Innovation Committee Foundation of Shenzhen.
Abstract (Originalsprache)
Artificial intelligence (AI) edge devices demand high energy efficiency as well as inference accuracy. SRAM-based compute-in-memory (CIM) accelerators have great potential for power reduction but still need to exploit higher throughput and better linearity performance. To meet edge-AI computing demands by CIM works, it is crucial to optimize algorithms and parameters for specific circuit systems to achieve hardware acceleration. This work firstly employs neural network search (NAS) method to find out the layer-wise optimized precisions and sparsities for convolutional neural networks (CNNs). Then, a 144-Kb charge-domain signed mixed-precision (2/4/8-bit) CIM accelerator employing bootstrapped SRAM cells with 9-transistors and 1-capacitor (9T1C) structure is proposed that incorporates a bit-level sparsity-aware analog-to-digital converter (ADC). This work not only achieves highly linear parallel accumulation operations to meet AI computing demands but also implements a hardware and software co-optimization system tailored to specific data characteristics. The design is verified on NAS-optimized networks VGG-16 and ResNet-18 using Cifar-10 dataset, which could achieve an equivalent accuracy at 4-bit of 68.68% while maintaining a high energy efficiency at 2-bit of 135.19TOPS/W by measurements.
Förderung
- Shenzhen Science and Technology Program
- Science and Technology Innovation Committee Foundation of Shenzhen
Zitieren
APA
Mao, W., Liu, D., Zhou, H., Li, F., Li, K., Wu, Q., Yang, J., Cheng, Q., Zhang, L., & Yu, H. (2025). A 28-nm 135.19 TOPS/W Bootstrapped-SRAM Compute-in-Memory Accelerator With Layer-Wise Precision and Sparsity. IEEE Transactions on Circuits and Systems I: Regular Papers, 72(7), 3236-3246. https://doi.org/10.1109/tcsi.2024.3486322
Vancouver
Mao W, Liu D, Zhou H, Li F, Li K, Wu Q, et al. A 28-nm 135.19 TOPS/W Bootstrapped-SRAM Compute-in-Memory Accelerator With Layer-Wise Precision and Sparsity. IEEE Transactions on Circuits and Systems I: Regular Papers. 2025;72(7):3236-3246. doi:10.1109/tcsi.2024.3486322
BibTeX
@article{Mao2025,
author = {Mao, Wei and Liu, Dingbang and Zhou, Haoxiang and Li, Fuyi and Li, Kai and Wu, Qiuping and Yang, Jiaqi and Cheng, Quan and Zhang, Liuyang and Yu, Hao},
title = {A 28-nm 135.19 TOPS/W Bootstrapped-SRAM Compute-in-Memory Accelerator With Layer-Wise Precision and Sparsity},
journal = {IEEE Transactions on Circuits and Systems I: Regular Papers},
year = {2025},
volume = {72},
number = {7},
pages = {3236--3246},
doi = {10.1109/tcsi.2024.3486322}
}
EUVIMED