Overview The main contribution of the paper is a novel approach to ease the performance modeling process by introducing a pre-processing step that automatically determines performance-irrelevant configuration options and removes them from the remaining modeling process. We show that our identification process can accurately identify performance-irrelevant options using Pace3D, a multi-physics simulation, as case study. As Pace3D is proprietary software to which we are granted access to, we cannot provide an executable. Therefore, it is not possible to repeat measurements of the software. We do, however, provide our collected measurement data, including logs of our conducted experiments and inputs used, as well as performance models created. The measurement data can be used to reproduce the learning of models using DECART. Furthermore, the evaluation data shows the runtime difference between configurations. DECART is available for download here: https://github.com/jmguo/DECART/ Structure - training-data: Input configurations and measurement data used for training models.- models: DECART models.- evaluation-data: Input configurations and measurement data for evaluating the models.- questionnaire: Filled questionnare by the Pace3D developer.