Speaker
Description
Computational acoustics deals with a variety of data, mathematical models and numerical algorithms along with a disparate set of workflows to achieve insights into the physics of sound.The diversity of its approaches, while valuable, presents a significant challenge: a lack of systematic means to develop, compare, validate, and evaluate their performance against another under consistent reproducible conditions.The concept of benchmarking emerged precisely to address this very need - yet its definition remains inconsistent across the broader community.Existing benchmarking efforts in computational science remain largely ad hoc, domain-specific, and rarely reusable across studies.The absence of standardized benchmarks obstructs transparent, reproducible evaluation of algorithms or feasibility of modelling approaches.To address this, we present the MaRDIMark benchmarking specification that aims to unify these different perspectives - developed within the MaRDI (Mathematical Research Data Initiative) consortium in Germany.The proposed framework enables a transparent means in which data, algorithms, metrics and workflows of varying complexity and formulations be systematically compared against one another for user-defined metrics.We describe its extensible design and demonstrate its application to representative test cases in computational acoustics.With this, MaRDIMark aims for a community-driven FAIR (Findable, Accessible, Interoperable, Reusable) benchmark ecosystem promoting broader accessibility and reusability.