Analysing the Variability of OpenMP Programs Performances on Multicore Architectures
Résumé
In [8], we demonstrated that contrary to sequential applications, parallel OpenMP appli- cations su er from a severe instability in performances. That is, running the same parallel OpenMP application with the same data input multiple times may exhibit a high variability of execution times. In this article, we continue our research e ort to analyse the reason of such performance variability. With the architectural complexity of the new state of the art hardware designs, comes a need to better understand the interactions between the operating system layers, the applications and the underlying hardware platforms. The ability to characterise and to quantify those interactions can be useful in the processes of performance evaluation and analysis, compiler optimisations and operating system job scheduling allowing to achieve better performance stability, reproducibility and predictability. Under- standing the performance instability in current multicore architectures is even more complicated by the variety of factors and sources in uencing the applications performances. This article focus on the e ects of thread binding, co-running processes, L2 cache sharing, automatic hardware prefetcher and memory page sizes.
Domaines
Autre [cs.OH]
Origine : Fichiers produits par l'(les) auteur(s)
Loading...