In-memory computing using phase change memory

Phase change Memory is considered one of the promising memory elements to be used as a non-volatile memory device, especially in non-von Neumann computing. With data intensive applications, in-memory computing (IMC) is considered as an attractive solution to the memory bottleneck. In this computing scheme, both logical and computational tasks are performed in the same physical location. A PCM device can store data by switching between a high-conductive (crystalline) phase and a low-conductive (amorphous) phase. The wide resistance range between these two states allows for a sufficient noise margin during READ operation and provides the opportunity to store multiple resistive levels allowing higher bit density. Moreover, scalability, high endurance and retention are additional great that makes PCM devices suitable for IMC-based applications. This chapter discusses several applications utilizing PCM-based IMC designs to accelerate computations efficiently. The applications cover various areas and schemes, including Compressed Sensing, Spiking Neural Networks (SNNs), Convolutional Neural Networks (CNNs), and Hyper-Dimensional Computing (HDC).