Parallel expression template for large vectors

Laurent Plagne (1) , Frank Hülsemann (2) , Denis Barthou (3) , Julien Jaeger (3)

1 EDF - EDF
2 EDF R&D - EDF R&D
3 PRISM - Parallélisme, Réseaux, Systèmes, Modélisation

This paper describes a short and simple way of improving the performance of vector operations (e.g. X = aY +bZ+::) applied to large vectors. In a previous paper [1] we described how to take advantage of high performance vector copy operation provided by the ATLAS library [2] in the context of C++ Expression Template (ET) mechanism. Here we present a multi-threaded implementation of this approach. The proposed ET implementation that involves a parallel blocking technique, leads to signi cant performance increase compared to existing implementations (up to 2:7) on dual socket x86 64 targets.

Mots clés: C++ template; OpenMP; expression template; intel TBB; multi-core processors; parallel computing

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Laurent Plagne, Frank Hülsemann, Denis Barthou, Julien Jaeger. Parallel expression template for large vectors. Workshop on Parallel/High-Performance Object-Oriented Scientific, Jul 2009, Genova, Italy. p8:1-8:8. ⟨hal-00551682⟩