SMART: A Heterogeneous Scratchpad Memory Architecture for Superconductor SFQ-based Systolic CNN Accelerators
SMART: A Heterogeneous Scratchpad Memory Architecture for Superconductor SFQ-based Systolic CNN Accelerators
复制标题
DOI：
10.1145/3466752.3480041
复制
发表时间：
2021-09
期刊：
影响因子：
--
通讯作者：
Farzaneh Zokaee;Lei Jiang
中科院分区：
文献类型：
--
作者：
Farzaneh Zokaee;Lei Jiang
基于超快、低功耗超导单通量量子 (SFQ) 的 CNN 脉动加速器旨在增强 CNN 推理吞吐量。然而，由于缺乏随机访问功能，基于移位寄存器 (SHIFT) 的暂存存储器 (SPM) 阵列会阻止 SFQ CNN 加速器超过其峰值吞吐量的 40%。本文首先记录了我们对多种低温存储器技术的研究，包括涡旋过渡存储器（VTM）、约瑟夫森-CMOS SRAM、MRAM和超导纳米线存储器，期间我们发现上述技术都没有使SFQ CNN加速器同时实现高吞吐量、小面积和低功耗。其次，我们提出了一种异构 SPM 架构 SMART，由 SHIFT 阵列和随机存取阵列组成，以提高 SFQ CNN 脉动加速器的推理吞吐量。第三，我们通过构建连接 CMOS 子组的基于 SFQ 无源传输线的 H 树，提出了一种快速、低功耗和密集流水线随机访问 CMOS-SFQ 阵列。最后，我们创建一个基于 ILP 的编译器来在 SMART 上部署 CNN 模型。实验结果表明，在相同的芯片面积开销下，相比最新的基于SHIFT的SFQ CNN加速器，SMART在推断单张图像（一批图像）时，推理吞吐量提高了3.9×（2.2×），推理能量降低了86%（71%）。
Ultra-fast & low-power superconductor single-flux-quantum (SFQ)-based CNN systolic accelerators are built to enhance the CNN inference throughput. However, shift-register (SHIFT)-based scratchpad memory (SPM) arrays prevent a SFQ CNN accelerator from exceeding 40% of its peak throughput, due to the lack of random access capability. This paper first documents our study of a variety of cryogenic memory technologies, including Vortex Transition Memory (VTM), Josephson-CMOS SRAM, MRAM, and Superconducting Nanowire Memory, during which we found that none of the aforementioned technologies made a SFQ CNN accelerator achieve high throughput, small area, and low power simultaneously. Second, we present a heterogeneous SPM architecture, SMART, composed of SHIFT arrays and a random access array to improve the inference throughput of a SFQ CNN systolic accelerator. Third, we propose a fast, low-power and dense pipelined random access CMOS-SFQ array by building SFQ passive-transmission-line-based H-Trees that connect CMOS sub-banks. Finally, we create an ILP-based compiler to deploy CNN models on SMART. Experimental results show that, with the same chip area overhead, compared to the latest SHIFT-based SFQ CNN accelerator, SMART improves the inference throughput by 3.9 × (2.2 ×), and reduces the inference energy by 86% (71%) when inferring a single image (a batch of images).