Toward Efficient Logic-in-Memory Computing With Magnetic Reconfigurable Logic Circuits

Abstract

Logic-in-memory (LIM) structures are promising candidates to obviate limitations of the conventional von Neumann architecture, especially in big data applications, such as image processing. In this paradigm, simple logic operations are embedded in memory to perform basic processes and consequently decrease the workload of the main processor. This letter presents an efficient hybrid fin field-effect transistor and magnetic tunnel junction (MTJ) logic structure compatible with all kinds of memory and which performs nor/or and nand/and operations. The design utilizes MTJs to obtain different voltage levels and two sense amplifiers to generate the outputs. Simulation results assert that the design improves the delay and power by 33% and 20%, respectively, compared to its state-of-the-art counterparts. Moreover, the magnetic LIM structure is appropriately utilized in image processing applications, such as minimum and maximum image filters, for preparing intermediate data. In the case study, high-level simulations indicate that the design reduces the delay and power by 31% and 21%, respectively.