Solving Systems of Linear Equations Using In-Memory Computing

This paper presents a novel hardware approach for solving systems of linear equations by leveraging in-memory computing (IMC) with memristive crossbar arrays. Unlike conventional architectures that suffer from memory-compute bottlenecks and high computational complexity, the proposed method performs the required matrix-vector multiplications directly within memory by exploiting the intrinsic physical properties of memristive devices. We propose a hardware architecture for solver kernel that integrates programmable crossbars to enable energy-efficient, parallel computation of linear systems. A comparative analysis with traditional digital architectures demonstrates the advantages of the IMC-based approach in terms of computational efficiency, scalability, and reduced hardware overhead. As a result, the proposed method shows strong potential for signal processing applications in next-generation computing systems.