Dev Builds C++ Benchmarking Suite, Finds Big-O Theory Often Misleads in Practice

A software developer spent four months building 'hashbrowns,' a C++17 benchmarking suite that compares arrays, linked lists, and hash maps implemented from scratch. The project aimed to measure insert, search, and remove operations and identify crossover points where one data structure outperforms another. Along the way, the developer discovered that virtual function call overhead, CPU cache behavior, and frequency scaling can significantly distort benchmark results in ways that theoretical complexity analysis does not predict. To address measurement noise, the project incorporated warm-up runs, outlier removal, bootstrap confidence intervals, and CPU affinity pinning. The exercise highlighted that real-world performance depends heavily on hardware and implementation details, making empirical benchmarking essential before any optimization work.