A New Memory Layout for Self-Rebalancing Trees

Paul Iannetta (1, 2, 3)

1 ENS de Lyon - École normale supérieure de Lyon
2 LIP - Laboratoire de l'Informatique du Parallélisme
3 UCBL - Université Claude Bernard Lyon 1

In this paper, we show that trees implemented as a collection of pointers suffer from a lack of parallelism opportunities. We propose an alternative implementation based on arrays. Both implementations appear to be equivalently efficient time-wise. However, this new layout exposes new parallelism opportunities which can be then exploited by an optimizing compiler.

Domaines: Performance et fiabilité [cs.PF]; Algorithme et structure de données [cs.DS]; Calcul parallèle, distribué et partagé [cs.DC]

Paul Iannetta. A New Memory Layout for Self-Rebalancing Trees. CGO'21, Feb 2021, Séoul, South Korea. ⟨hal-03123491⟩