An Eight-Core 1.44GHz RISC-V Vector Machine in 16nm FinFET

📋 论文概要

本文设计了一款八核RISC-V向量处理器，运行频率1.44GHz，采用台积电16nm FinFET工艺制造。该芯片面向深度神经网络等密集算术计算场景，支持可变精度计算，提升了能效。

💡 主要创新点

• 提出了八核RISC-V向量处理器架构，支持高频率1.44GHz运行。
• 实现了可变精度计算支持，适应不同层和训练/推理阶段的精度需求。
• 集成了系统管理核和高速串行链路，构成完整SoC系统。

🏷 关键词

📄 原文摘要

such as deep neural networks (DNNs), increasingly rely on dense arithmetic compute patterns that are ill-suited for general-purpose processors, leading to a rise in domain-specific compute accelerators [1]. Many of these workloads can benefit from varying precision during computation, e.g. different precisions among layers and between training and inference for DNNs has been shown to improve energy efficiency [2]. The SoC is manufactured in TSMC 16nm FinFET, measuring 4.9×4.9mm2 (Fig. 4.3.3). The system contains eight application cores, one system-management core, eight serial links, and three levels of cache (Fig. 4.3.1). The application cores comprise one scalar RISC-V processor and a decoupled vector accelerator. The application cores measure

👥 作者与机构

Colin Schmidt*, John Wright*, Zhongkai Wang, Eric Chang, Albert Ou,

Woorham Bae, Sean Huang, Anita Flynn, Brian Richards, Krste Asanović, Elad Alon, Borivoje Nikolić University of California, Berkeley, CA *Equally Credited Authors (ECAs) Modern workloads,