GPU Computing

This preprint describes the scientific vision, user interface, governing equations, and numerical methods behind Oceananigans. It brings together relatively simple finite-volume methods, GPU-oriented implementation, and an expressive Julia scripting interface. Examples span classroom-scale fluid problems, small-scale turbulence, and realistic global ocean circulation. The discussion connects software accessibility and extensibility to advances in parameterizations, numerical methods, and computational performance. Readers get a broad technical guide to the model and a rationale for how its software design supports experiments across very different ocean scales.

This software paper introduces Oceananigans as a Julia framework for geophysical fluid simulations on CPUs and GPUs. It combines a high-level scripting interface with numerical methods for incompressible, stratified, rotating flow. Users can specify model configurations and diagnostics while sharing a common code base across hardware architectures. The paper describes performance benchmarks, verification practices, and applications such as ocean boundary-layer turbulence. Readers get a concise introduction to the software’s original capabilities and design, helping them assess it as a platform for numerical experiments.