Loop Aggregation for Approximate Scientific Computing by June Sallou, Alexandre Gauvain, Johann Bourcier, Benoit Combemale, Jean-Raynald de Dreuzy.
Trading off some accuracy for better performances in scientific computing is an appealing approach to ease the exploration of various alternatives on complex simulation models. Existing approaches involve application either time-consuming model reduction techniques or resource-demanding statistical approaches. Such requirements prevent any opportunistic exploration, e.g., exploring scenarios environmental This limits ability analyse new models scientists, support trade-off analysis decision-makers and empower general public towards informed intelligence. In this paper, we present a approximate technique, aka. loop aggregation, which consists automatically reducing main by aggregating corresponding spatial temporal data. We apply geophysical hydraulic with input The experimentation demonstrates drastically decrease time while preserving acceptable results minimal set-up. obtain median speed-up 95.13% up 99.78% across all 23 case studies.
Read Loop Aggregation for Approximate Scientific Computing on OA.mg