Auto-tuning for floating-point precision with Discrete Stochastic Arithmetic
| Autor: | Bruno Lathuilière, François Févotte, Romain Picot, Fabienne Jézéquel, Stef Graillat |
|---|---|
| Přispěvatelé: | Performance et Qualité des Algorithmes Numériques (PEQUAN), LIP6, Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS), Université Panthéon-Assas (UP2), Simulation Neutronique, Technologies de l’Information et Calcul Scientifique (EDF R&D SINETICS), EDF R&D (EDF R&D), EDF (EDF)-EDF (EDF) |
| Rok vydání: | 2019 |
| Předmět: |
Discrete Stochastic Arithmetic
Floating point floating-point arithmetic General Computer Science Computer simulation Computer science round-off errors Linear system auto-tuning 02 engineering and technology 01 natural sciences 010305 fluids & plasmas Theoretical Computer Science Reduction (complexity) Modeling and Simulation 0103 physical sciences Arbitrary-precision arithmetic 0202 electrical engineering electronic engineering information engineering Code (cryptography) Delta Debugging numerical validation [INFO]Computer Science [cs] 020201 artificial intelligence & image processing SIMD Algorithm |
| Zdroj: | Journal of computational science Journal of computational science, Elsevier, 2019, 36, pp.101017. ⟨10.1016/j.jocs.2019.07.004⟩ |
| ISSN: | 1877-7503 1877-7511 |
| Popis: | International audience; The type length chosen for floating-point numbers (e.g. 32 bits or 64 bits) may have an impact on the execution time, especially on SIMD (Single Instruction Multiple Data) units. Furthermore optimizing the types used in a numerical simulation causes a reduction of the data volume that is possibly transferred. In this paper we present PROMISE, a tool that makes it possible to optimize the numerical types in a program by taking into account the requested accuracy on the computed results. With PROMISE the numerical quality of results is verified using DSA (Discrete Stochastic Arithmetic) that enables one to estimate round-off errors. The search for a suitable type configuration is performed with a reasonable complexity thanks to the delta debugging algorithm. The PROMISE tool has been successfully tested on programs implementing several numerical algorithms including linear system solving and also on an industrial code that solves the neutron transport equations. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect