A Survey on System-Level Design of Neural Network Accelerators
A Survey on System-Level Design of Neural Network Accelerators
复制标题
DOI：
10.29292/jics.v16i2.505
复制
发表时间：
2021-08
影响因子：
--
通讯作者：
Kenshu Seto
中科院分区：
文献类型：
--
作者：
Kenshu Seto
在本文中，我们简要介绍了用于卷积神经网络（CNN）推理加速器的系统级优化。对于卷积层的嵌套循环（CONV），我们讨论了循环交换、平铺、展开和融合等循环优化对CNN加速器的影响。我们还解释了循环优化中有效的内存优化。此外，我们还讨论了CNN加速器中常用的流架构和单计算引擎架构。CNN模型的优化进行了简要说明，其次是CNN加速器设计的最新趋势和未来方向。
In this paper, we present a brief survey on the system-level optimizations used for convolutional neural network (CNN) inference accelerators. For the nested loop of convolutional (CONV) layers, we discuss the effects of loop optimizations such as loop interchange, tiling, unrolling and fusion on CNN accelerators. We also explain memory optimizations that are effective with the loop optimizations. In addition, we discuss streaming architectures and single computation engine architectures that are commonly used in CNN accelerators. Optimizations for CNN models are briefly explained, followed by the recent trends and future directions of the CNN accelerator design.