Communication Dans Un Congrès Année : 2026

Convolutions Predictable Offloading to an Accelerator: Formalization and Optimization

Résumé

Convolutional neural networks (CNNs) require a large number of multiply-accumulate (MAC) operations. To meet real-time constraints, they often need to be executed on specialized accelerators composed of an on-chip memory and a processing unit. However, the on-chip memory is often insufficient to store all the data required to compute a CNN layer. Thus, the computation must be performed in several offloading steps. We formalise such sequences of steps and apply our formalism to a state of the art decomposition of convolutions. In order to find optimal strategies in terms of duration, we encode the problem with a set of constraints. A Python-based simulator allows to analyse in-depth computed strategies.

Fichier principal
Vignette du fichier
main.pdf (625.66 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
licence

Dates et versions

hal-05513761 , version 1 (16-02-2026)
hal-05513761 , version 2 (20-03-2026)

Licence

Identifiants

Citer

Benjamin Husson, Mohammed Belcaïd, Thomas Carle, Claire Pagetti. Convolutions Predictable Offloading to an Accelerator: Formalization and Optimization. 13th European Congress of Embedded Real Time Systems (ERTS), Feb 2026, Toulouse, France. ⟨10.82331/ERTS.2026.25⟩. ⟨hal-05513761v2⟩
1619 Consultations
222 Téléchargements

Altmetric

Partager

  • More