Magnetic Racetrack Memories: Design and Optimizations (Synthesis Lectures on Computer Architecture)

This book provides an interdisciplinary overview of spin-orbitronics-based magnetic racetrack memory (RTM). The authors provide an overview of RTM device physics, its evolution, modeling tools, strengths and challenges, and its application in the memory subsystem. They describe leading-edge developments in RTM physics and materials, data sensing elements, controlled and energy-efficient movement of the magnetic elements within the nanowire, and custom shifts-aware data and instruction placement solutions using optimal, near-optimal, and evolutionary algorithms. The book describes automatic software optimization methods that take an input program, analyze its memory access behavior, and, if possible, optimize it by reordering the memory access without any user intervention. Further, it also discusses state-of-the-art reliability schemes that can accurately detect and correct position errors in RTMs. Finally, this book demonstrates RTM-based computation-in-memory and its potential to break the memory and bandwidth walls. The authors explain how the RTM devices can morph into compute elements and perform in-situ computations using their properties. As a concrete use case, the book presents an entire hyperdimensional computing framework and explains the implementation of various operations. Read more