XOR-CiM: An Efficient Computing-in-SOT-MRAM Design for Binary Neural Network Acceleration
XOR-CiM: An Efficient Computing-in-SOT-MRAM Design for Binary Neural Network Acceleration
复制标题
DOI：
10.1109/isqed57927.2023.10129322
复制
发表时间：
2023-04
期刊：
影响因子：
--
通讯作者：
Mehrdad Morsali;Ranyang Zhou;Sepehr Tabrizchi;A. Roohi;Shaahin Angizi
中科院分区：
文献类型：
--
作者：
Mehrdad Morsali;Ranyang Zhou;Sepehr Tabrizchi;A. Roohi;Shaahin Angizi
在这项工作中，我们利用自旋轨道扭矩磁随机存取存储器(SOT-MRAM)的单极开关特性，开发了一个高效的数字内存计算(CIM)平台XOR-CIM。XOR-CIM将典型的MRAM子阵转换为具有超高带宽的大规模并行计算核，大大降低了处理卷积层的能量消耗，并加速了X(N)OR密集型二进制神经网络(BNN)的推理。与目前基于∼的CIM平台相比，XOR-CIM的推理精度与数字CIMS相似，能效和加速分别提高了4.5倍和1.8倍。
In this work, we leverage the uni-polar switching behavior of Spin-Orbit Torque Magnetic Random Access Memory (SOT-MRAM) to develop an efficient digital Computing-in-Memory (CiM) platform named XOR-CiM. XOR-CiM converts typical MRAM sub-arrays to massively parallel computational cores with ultra-high bandwidth, greatly reducing energy consumption dealing with convolutional layers and accelerating X(N)OR-intensive Binary Neural Networks (BNNs) inference. With a similar inference accuracy to digital CiMs, XOR-CiM achieves ∼4.5× and 1.8× higher energy-efficiency and speed-up compared to the recent MRAM-based CiM platforms.