D2CIM: A 28nm 53.3 TFLOPS/W Decoding Digital CIM Macro for Efficient FlashMLA-based LLM Inference
| Title: | D2CIM: A 28nm 53.3 TFLOPS/W Decoding Digital CIM Macro for Efficient FlashMLA-based LLM Inference |
|---|---|
| Authors: | Chen, Jia; Xuan, Zihao; Li, Yewen; Yan, Zhuoya; Li, Zimu; Yung, Yat-Fong; Deng, Yi; Huo, Xiao; Tsui, Chi-Ying; Cheng, Kwang-Ting; Tu, Fengbin |
| Source: | 2026 IEEE Custom Integrated Circuits Conference (CICC) Custom Integrated Circuits Conference (CICC), 2026 IEEE. :1-4 Apr, 2026 |
| Relation: | 2026 IEEE Custom Integrated Circuits Conference (CICC) |
| Database: | IEEE Xplore Digital Library |