Code generation for multi-phase tasks on multicore with distributed memory

Frédéric Fort (1) , Julien Forget (2)

1 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
2 ONERA - The French Aerospace Lab [Toulouse]

Predicting the temporal behavior of a real-time system embedded on a multicore is a challenging task. One of the main reasons is that cores share access to the main memory, and contentions on the memory bus cause execution delays. Multi-phase task models, where computations phases and communication phases are separated (such as PREM and AER), have been proposed to both mitigate these delays and make them easier to analyze. In this paper we present a compilation process, part of the Prelude compiler, that automatically translates a high-level system specification, consisting in a set of periodic tasks with data-dependencies, into a C program that operates according to the AER multi-phase model. The generated C program is targeted for a multicore platform with distributed memory and includes code to perform communications between the different memories of the platform.

Mots clés: Code generation; multi-phase tasks; distributed memory

Domaines: Systèmes embarqués; Langage de programmation [cs.PL]

Frédéric Fort, Julien Forget. Code generation for multi-phase tasks on multicore with distributed memory. Junior Researcher Workshop on Real-Time Computing, Oct 2018, Poitiers, France. ⟨hal-02014418⟩