First steps towards more numerical reproducibility*
1 Sorbonne Universités, UPMC Univ Paris
06, CNRS, UMR 7606, LIP6, F-75005, Paris,
France and Université Paris
2 Université de Perpignan Via Domitia, DALI, LIRMM (UMR5506 CNRS-Université Montpellier 2), France
3 INRIA, U. de Lyon, LIP (UMR5668 CNRS-ENS de Lyon-INRIA-UCBL), ENS de Lyon, France
Questions whether numerical simulation is reproducible or not have been reported in several sensitive applications. Numerical reproducibility failure mainly comes from the finite precision of computer arithmetic. Results of floating-point computation depends on the computer arithmetic precision and on the order of arithmetic operations. Massive parallel HPC which merges, for instance, many-core CPU and GPU, clearly modifies these two parameters even from run to run on a given computing platform. How to trust such computed results? This paper presents how three classic approaches in computer arithmetic may provide some first steps towards more numerical reproducibility.
© EDP Sciences, SMAI 2014