Enabling efficient ReRAM-based neural network computing via crossbar structure adaptive optimization
Enabling efficient ReRAM-based neural network computing via crossbar structure adaptive optimization
复制标题
DOI：
10.1145/3370748.3406581
复制
发表时间：
2020-08
期刊：
影响因子：
--
通讯作者：
Chenchen Liu;Fuxun Yu;Zhuwei Qin;Xiang Chen
中科院分区：
文献类型：
--
作者：
Chenchen Liu;Fuxun Yu;Zhuwei Qin;Xiang Chen
基于电阻式随机存取存储器（ReRAM）的加速器已经被广泛研究以在速度和能量上实现高效的神经网络计算。开发了诸如稀疏性的神经网络优化算法，以在诸如CPU和GPU的传统计算机架构上实现高效的神经网络计算。然而，当在基于ReRAM的加速器上部署这些算法时，由于其独特的交叉结构计算，这种计算效率的提高受到阻碍。而针对基于ReRAM架构的具体算法和硬件协同优化还处于缺乏之中。在这项工作中，我们提出了一个高效的神经网络计算框架，专门用于基于ReRAM的加速器上的交叉结构计算。所提出的框架包括交叉开关特定的特征映射修剪和自适应神经网络部署。实验结果表明，我们的设计可以提高9.1%的计算精度相比，最先进的稀疏神经网络。基于著名的基于ReRAM的DNN加速器，该框架展示了高达1.4倍的加速比，4.3倍的功率效率和4.4倍的面积节省。
Resistive random-access memory (ReRAM) based accelerators have been widely studied to achieve efficient neural network computing in speed and energy. Neural network optimization algorithms such as sparsity are developed to achieve efficient neural network computing on traditional computer architectures such as CPU and GPU. However, such computing efficiency improvement is hindered when deploying these algorithms on the ReRAM-based accelerator because of its unique crossbar-structural computations. And a specific algorithm and hardware co-optimization for the ReRAM-based architecture is still in a lack. In this work, we propose an efficient neural network computing framework that is specialized for the crossbar-structural computations on the ReRAM-based accelerators. The proposed framework includes a crossbar specific feature map pruning and an adaptive neural network deployment. Experimental results show our design can improve the computing accuracy by 9.1% compared with the state-of-the-art sparse neural networks. Based on a famous ReRAM-based DNN accelerator, the proposed framework demonstrates up to 1.4× speedup, 4.3× power efficiency, and 4.4× area saving.