EUVIMEDEUVIMED entdecken
Crossref

A 28-nm Digital Transpose SRAM Compute-in-Memory Macro With Accurate/Approximate Dual Mode for Floating-Point Edge Training and Inference

Yiyang Yuan, Bingxin Zhang, Yiming Yang, Yishan Luo, Qirui Chen, Haitao Wang, Qihao Liu, Zhiming Chen, Hao Wu, Jinshan Yue, Shidong Lv, Xinghua Wang und 3 weitere

IEEE Journal of Solid-State Circuits · 2026-09 · 61 (9) · 5098-5110

DOI: 10.1109/jssc.2026.3679560 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Journal of Solid-State Circuits, 2026-09. Verfasst von Yiyang Yuan und 14 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 33 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Beijing Natural Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Static random-access memory (SRAM)-based computing-in-memory (CIM) macros have been widely studied to improve the energy efficiency of edge artificial intelligence (AI) inference tasks. However, less attention has been given to AI training, which requires CIM macros to not only perform matrix multiply-accumulate (MAC) operations but also support matrix transposition. To address the limitations of previous analog transpose and digital non-transpose SRAM CIM macros, this work features: 1) a cyclic-weight-mapping SRAM array that enables matrix transposition and reuse of MAC circuits during both feed-forward (FF) and back-propagation (BP) phases; 2) a digital CIM architecture employing signed fixed-point mantissa encode and a vector-wise pre-alignment (VWPA) scheme, supporting multiple data formats including INT4/8, FP8, and BF16; and 3) an accurate/approximate dual-mode bit-parallel MAC circuit (DMBP-MAC) designed to provide a tradeoff between computational accuracy and energy efficiency. A fabricated 28-nm 32-kB transpose SRAM CIM macro achieved average energy efficiency of 70.2–285.4 TOPS/W in INT4, 17.5–71.4 TOPS/W in INT8, 51.1–192.3 TFLOPS/W in FP8, and 12.8–48 TFLOPS/W in BF16.

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Beijing Natural Science Foundation

Zitieren

APA

Yuan, Y., Zhang, B., Yang, Y., Luo, Y., Chen, Q., Wang, H., Liu, Q., Chen, Z., Wu, H., Yue, J., Lv, S., Wang, X., Mak, P., Li, X., & Zhang, F. (2026). A 28-nm Digital Transpose SRAM Compute-in-Memory Macro With Accurate/Approximate Dual Mode for Floating-Point Edge Training and Inference. IEEE Journal of Solid-State Circuits, 61(9), 5098-5110. https://doi.org/10.1109/jssc.2026.3679560

Vancouver

Yuan Y, Zhang B, Yang Y, Luo Y, Chen Q, Wang H, et al. A 28-nm Digital Transpose SRAM Compute-in-Memory Macro With Accurate/Approximate Dual Mode for Floating-Point Edge Training and Inference. IEEE Journal of Solid-State Circuits. 2026;61(9):5098-5110. doi:10.1109/jssc.2026.3679560

BibTeX
@article{Yuan2026,
  author = {Yuan, Yiyang and Zhang, Bingxin and Yang, Yiming and Luo, Yishan and Chen, Qirui and Wang, Haitao and Liu, Qihao and Chen, Zhiming and Wu, Hao and Yue, Jinshan and Lv, Shidong and Wang, Xinghua and Mak, Pui-In and Li, Xiaoran and Zhang, Feng},
  title = {A 28-nm Digital Transpose SRAM Compute-in-Memory Macro With Accurate/Approximate Dual Mode for Floating-Point Edge Training and Inference},
  journal = {IEEE Journal of Solid-State Circuits},
  year = {2026},
  volume = {61},
  number = {9},
  pages = {5098--5110},
  doi = {10.1109/jssc.2026.3679560}
}
Kostenlos starten →