AI 中文总结
针对大语言模型推理能耗高的问题，提出CIMERA可重构精度推理加速器，集成计算与内存，缓解内存墙，实现精度感知执行，相比英伟达H100，处理1B和13B模型时能源效率分别提升25倍和10倍。
AI 中文摘要
大语言模型对计算和内存要求很高，给从数据中心到功率受限的边缘设备等跨平台的节能推理带来挑战。权重精度在平衡推理准确性、吞吐量和能耗方面起着关键作用，而现代大语言模型工作负载具有明显的异构性和容错性，有利于自适应精度执行。本文提出了CIMERA，一种可重构精度的大语言模型推理加速器，它集成了互连和内存计算，以缓解内存墙并实现精度感知执行。与英伟达H100相比，CIMERA在处理1B和13B模型时，能源效率分别提高了25倍和10倍。
英文摘要
LLM impose significant computational and memory demands, creating challenges for energy-efficient inference across platforms ranging from data centers to power-constrained edge devices. Weight precision plays a critical role in balancing inference accuracy, throughput, and energy consumption, while modern LLM workloads exhibit pronounced heterogeneity and tolerance that favors adaptive precision execution. This paper presents CIMERA, a reconfigurable-precision LLM inference accelerator that integrates compute-in-interconnect and memory to mitigate the memory wall and enable precision-aware execution. Compared to Nvidia H100, CIMERA delivers up to $25\times$ and $10\times$ higher energy efficiency for 1B and 13B models, respectively.
CommentsAccepted to 2026 IEEE 8th International Conference on Artificial Intelligence Circuits and Systems (AICAS'26)
