Raffinement temporel et exécution parallèle dans un langage synchrone fonctionnel

The generation of efficient sequential code for synchronous data-flow languages raises twointertwined issues: control and memory optimization. While the former has been extensivelystudied, for instance in the compilation of LUSTRE and SIGNAL, the latter has only been addressedin a restricted manner. Yet, memory optimization becomes a pressing issue when arrays are addedto such languages.This article presents a two-level solution to the memory optimization problem. It combines acompile-time optimization algorithm, reminiscent of register allocation, paired with languageannotations on the source given by the designer. Annotations express in-place modificationsand control where allocation is performed. Moreover, they allow external functions performingin-place modifications to be safely imported. Soundness of annotations is guaranteed by asemilinear type system and additional scheduling constraints. A key feature is that annotationsfor well-typed programs do not change the semantics of the language: removing them may leadto less efficient code but will not alter the semantics.The method has been implemented in a new compiler for a LUSTRE-like synchronous languageextended with hierarchical automata and arrays. Experiments show that the proposed approachremoves most of the unnecessary array copies, resulting in faster code that uses less memory.