A methodology of parallelization for continuous verified global optimization

Nathalie Revol 1 Yves Denneulin 2 Jean-François Méhaut 2 Benoît Planquelle 3
1 ARENAIRE - Computer arithmetic
Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme
2 APACHE - Parallel algorithms and load sharing
ID-IMAG - Informatique et Distribution, Inria Grenoble - Rhône-Alpes, UJF - Université Joseph Fourier - Grenoble 1
Abstract : Hansen's algorithm for verified continuous global optimization is based on interval arithmetic. It is of the Branch&Bound type. Our proposal for the parallelization of Hansen's algorithm consists in creating a lightweight process or thread to explore each subinterval and in beginning its execution immediately after its creation. This strategy provides a high level of speculation and enables to virtualize the architecture. The implementation of this parallel algorithm is based on the parallel execution support PM2 for the handling of threads at a low cost and on the PROFIL/BIAS interval library. Preliminary experimental results exhibit superlinear speed-ups.
Document type :
Conference papers
Complete list of metadatas

https://hal.inria.fr/inria-00545078
Contributor : Nathalie Revol <>
Submitted on : Thursday, December 9, 2010 - 3:23:29 PM
Last modification on : Monday, July 8, 2019 - 3:11:37 PM

Links full text

Identifiers

Citation

Nathalie Revol, Yves Denneulin, Jean-François Méhaut, Benoît Planquelle. A methodology of parallelization for continuous verified global optimization. PPAM 2001: Parallel Processing and Applied Mathematics, Roman Wyrzykowski, Sep 2001, Naleczow, Poland. pp.803-810, ⟨10.1007/3-540-48086-2_90⟩. ⟨inria-00545078⟩

Share

Metrics

Record views

297