| Course Code: CS5290 Course Name: Memory-Centric Computing Architectures Syllabus: This course introduces architectural approaches that reduce data movement by bringing computation closer to memory. It begins with a brief review of the memory hierarchy and the memory wall problem, followed by an overview of computation patterns in AI workloads. The course then covers near-memory and in-memory computing architectures. In-memory computing is discussed using ReRAM crossbar arrays as a case study for logic operations and neural network inference. The course also examines reliability issues, fault models, and possible security vulnerabilities in such systems. It concludes with an introduction to neuromorphic computing architectures inspired by biological neural systems. Content: Computer Architecture Foundations: Processor performance metrics, instruction-level parallelism, pipelining, memory hierarchy organization, cache design principles, memory latency and bandwidth. Memory Wall and Data Movement: Processor?memory performance gap, bandwidth and energy bottlenecks, roofline performance model, workload characterization for data-intensive applications. AI Workloads and Accelerator Architectures: Neural network computation patterns, matrix multiplication and convolution dataflows, data reuse strategies, memory traffic analysis, design principles of neural network accelerators. Near-Memory Architectures: Data-centric computing paradigms, processing-in-memory concepts, 3D-stacked memory integration, high-bandwidth memory architectures, near-memory accelerator design, system-level composition. In-Memory Computing Architectures: Analog and digital compute-in-memory models, crossbar array organization, in-memory matrix?vector multiplication, precision and scalability constraints; Case study-ReRAM crossbar arrays for logic operations and neural network inference. Reliability, Faults and Security: Fault models in memory-centric systems, process variability and noise effects, fault propagation in neural computations, redundancy techniques, reliability-aware design, fault injection attacks and data integrity vulnerabilities. Neuromorphic Computing Architectures: Biological foundations of neural computation, spiking neuron models, spiking neural network architectures, event-driven processing, neuromorphic processor design. Texts: 1. John L. Hennessy & David A. Patterson (2025). Computer Architecture: A Quantitative Approach (7 ed.). Elsevier 2. Mohamed M. Aly & Anupam Chattopadhyay (Ed.) (2022). Emerging Computing: From Devices to Systems: Looking Beyond Moore and Von Neumann. Springer Nature References: 1. Rasit O. Topaloglu (Ed.) (2015). More than Moore Technologies for Next Generation Computer Design. Springer 2. Bruce Jacob, David Wang & Spencer Ng (2010). Memory Systems: Cache, DRAM, Disk. Morgan Kaufmann |