Berkeley Dwarfs

Berkeley Dwarfs are a set of 13 fundamental computational patterns, or "building blocks," identified by researchers at UC Berkeley to categorize high-performance computing workloads. These motifs represent common types of tasks and data operations across diverse applications, such as matrix computations, graph analysis, and machine learning. Recognizing these patterns helps in designing hardware and software optimized for real-world workloads, improving efficiency and performance. Instead of focusing on specific programs, Dwarfs generalize the core computational activities that underpin many complex tasks, guiding innovations in computing architecture and system design.