arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.04797cs.AR
收藏

面向3D-DRAM的LLM服务加速器的软硬件协同设计

Hardware-Software Co-design for 3D-DRAM-based LLM Serving Accelerator

Peking University(北京大学);Xi’an UniIC Semiconductors(西安联芯集成电路);Houmo AI(厚墨智能);HKUST(香港科技大学)

Cong Li, Yihan Yin, Chenhao Xue, Zhao Wang, Fujun Bai, Yixin Guo, Xiping Jiang, Qiang Wu, Yuan Xie, Guangyu Sun

AI总结:

Helios通过软硬件协同设计,优化LLM服务加速器的动态KV缓存管理和分布式内存抽象,实现3.25倍加速和3.36倍能效提升。

摘要:

大型语言模型(LLMs)已被广泛部署用于在线生成服务,其中多个LLM实例共同处理具有波动请求到达率和变化请求长度的工作负载。为高效执行共存的计算密集型和内存密集型运算符,基于近内存处理(NMP)的计算范式已被广泛提出。然而,现有NMP设计采用粗粒度的KV缓存管理和刚性的注意力执行流程。这些限制阻碍了这些方案有效处理高度动态的LLM服务工作负载的能力,限制了其加速LLM服务的能力。为解决这些问题,我们提出了Helios,一种基于混合键合的LLM服务加速器。Helios旨在弥合LLM服务中KV缓存管理的动态性质与NMP处理引擎(PEs)之间分布式、非均匀内存抽象之间的根本差距。为此,我们设计了PE内部的执行流程和PE之间的通信原语,用于分布式 tiled 注意力执行。我们进一步提出了空间感知的KV缓存分配机制,以在平衡注意力工作负载分布的同时,最小化PE之间数据传输的开销。与现有GPU/NMP设计相比,Helios实现了3.25倍(几何均值)的加速率和3.36倍(几何均值)更好的能效,同时在P50/P99时间间隔之间达到高达72%/76%的退化。

英文摘要:

Large language models (LLMs) have been widely deployed for online generative services, where numerous LLM instances jointly handle workloads with fluctuating request arrival rates and variable request lengths. To efficiently execute coexisting compute-intensive and memory-intensive operators, near-memory processing (NMP) based computing paradigm has been extensively proposed. However, existing NMP designs adopt coarse-grained KV cache management and inflexible attention execution flow. Such limitations hinder these proposals from efficiently handling \textit{highly dynamic} LLM serving workloads, limiting their ability to accelerate LLM serving. To tackle these problems, we propose Helios, a Hybrid-bonding-based \uline{L}LM \uline{S}erving accelerator. Helios aims to bridge the fundamental gap between the dynamic nature of KV cache management in LLM serving and the distributed, non-uniform memory abstraction among NMP processing engines (PEs). To this end, we design both the intra-PE execution flow and the inter-PE communication primitives for distributed tiled attention execution. We further propose \textit{spatially-aware} KV cache allocation mechanism to balance the attention workload distribution while minimizing the inter-PE data transfer overhead. Compared with existing GPU/NMP designs, Helios achieves 3.25 times (geomean) speedup and 3.36 times (geomean) better energy efficiency, along with up to 72%/76% P50/P99 time-between-tokens degradation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑