Assessing Reproducibility: An Astrophysical Example of Computational Uncertainty in the HPC Context
Abstract: We present an experiment using the Enzo simulation code on NCSA's Blue Waters system to highlight the importance of computational and numerical uncertainty in scientific computing on HPC systems. We quantify the (surprising) variability of outputs from 200 identical simulation runs. We make two recommendations to improve the assessment of reproducibility of computational results: the inclusion of computational variability measures in standard reporting practices; and modular software designs that permit localized assessments of the source of computational uncertainty.
Back to ResCuE-HPC: 1st Workshop on Reproducible, Customizable, and Portable Workflows for HPC Archive Listing
Back to Full Workshop Archive Listing