In-Memory Computing

(3 Credit Hours) · Undergraduate

This course is intended for senior undergraduate and graduate students who are interested in (i) designing efficient algorithms and software running on platforms beyond traditional multicore CPUs, and/or (ii) developing systems to accelerate popular applications such as machine learning and autonomous systems. It introduces recent and emerging hardware platforms that are typically based on nontraditional computing models, architectures, and technologies. The course specifically focuses on how demanding applications such as data analytics, artificial intelligence and graph processing can best benefit from such hardware platforms. Some example hardware platforms to be discussed include graphics processing units (GPUs), domain-specific accelerators based on near- and in-memory computing, neuromorphic computing engines based on CMOS as well as ferroelectric, magnetic and photonic devices, etc. The course emphasizes the cross-layer design practice in which unique properties of applications, algorithms, architectures, circuits and devices are identified and exploited.