ADASS 2017: Extracting Insights from Astrophysics Simulations

Simulations inform all aspects of modern astrophysical research, ranging in scale from 1D and 2D test problems that can run in seconds on an astronomer's laptop all the way to large-scale 3D calculations that run on the largest supercomputers, with a spectrum of data sizes and shapes filling the landscape between these two extremes. In this talk I will review the diversity of astrophysics simulation data formats commonly in use by researchers, providing an overview of the most common simulation techniques, including pure N-body dynamics, smoothed particle hydrodynamics (SPH), adaptive mesh refinement (AMR), spectral methods, and unstructured meshes. Additionally, I will highlight methods for incorporating physical phenomena that are important for astrophysics, including chemistry, magnetic fields, radiative transport, and "subgrid" recipes for important physics that cannot be directly resolved in a simulation. In addition to the numerical techniques, I will also discuss the communities that have developed around these simulation codes and argue that increasing use and availability of open community codes has dramatically lowered the barrier to entry for novice simulators. Extracting scientific results from astrophysical simulation data requires detailed knowledge of the underlying data structures and data formats, as well as the semantic meaning of the data in relation to the physics problem posed by the simulation. As a solution to this problem, I will present yt, a community-developed python library for analyzing and visualizing simulation data. With support for most of the common astrophysics simulation research data formats, yt endeavors to provide a universal language for asking physically motivated questions of simulation data, regardless of the underlying data format. I will highlight the community of yt contributors and users, showcase scientific results where yt was used to facilitate the analysis.