PARALLEL SESSION
A1L-G
Digital Compute-in-Memory Session
SALA BACH
TUESDAY - 8 September | 11:00 - 13:00
CHAIRS
Riccardo Zurla, Takashi Maeda
SESSION PROGRAM
11:00 - 11:20
DM-CIM: A 28nm SRAM Compute-in-Memory Macro with Dynamic Microscaling and Flexible Mixed-Precision MAC
Xiaofeng Hu, Hangyeol Mun, Yuan Liao, Yifan He, Arnab Raha, Amit Agarwal, Mark Anders, Jae-Sun Seo
11:20 - 11:40
A 12nm 80.6TFLOPS/W BF16 Floating-Point Compute-in-Memory Macro with In-Memory Exponent Search and Shift-Free Mantissa Alignment for LLM Decoding
Siqi He, Jiapei Zheng, Kaijun Zhang, Zexing Chen, Yunzhengmao Wang, Qi Liu, Chixiao Chen, Haozhe Zhu
11:40 - 12:00
PDE-CIM: A 12nm FinFET 9.3TFLOPS/W Post-Aligned FP32-Lossless Digital CIM Macro with Input-Partial-Product-LUT and Fused Compression for PDE Acceleration
Ming Li, Rui Zhang, Yutianshi Li, Hao Wu, Zichen Fan, Shang Shi, Bingxin Zhang, Yiyang Yuan, Yuanwei Song, Qianyi Zha...
12:00 - 12:20
A 28nm 105.4TOPS/W Via-Programmed Digital Compute-in-ROM Accelerator with In-Cell Multiplication and Trimmed Adder Tree
Minghan Jiang, Xinran Li, Jianan Pan, Shuaiting Li, Haibin Shen, Kejie Huang
12:20 - 12:40
A 77.1 TOPS/W Simultaneous Refresh-Compute eDRAM CIM Macro with Input-Aware ADC and Effective-Bit Reconfigurable Accumulator
Hoichang Jeong, Seungbin Kim, Gyumin Choi, Minsoo Kim, Kyuho Lee
12:40 - 13:00
From Macro to System: Quantifying SoC Overheads in a 22nm 165 TOPS/W Innovative Digital CIM Accelerator
Asmae El Arrassi, Lucas Huijbregts, Georgi Gaydadjiev, Mottaqiallah Taouil, Rajiv Joshi, Said Hamdioui