Section: New Results

Formal translation validation of multi-processor real-time schedules

Participants : Keryan Didier, Dumitru Potop-Butucaru.

This research direction is mainly represented by the PhD thesis of Keryan Didier, and takes place in the framework of the ITEA3 ASSUME project. The technical focus of the ASSUME project is on formal compiler verification and on correct real-time implementation for parallel applications. The objective of this PhD thesis is to formally prove the correctness of (part of) the automatic code generation technology of Lopht, considering the respect of non-functional requirements, and in particular real-time requirements such as release dates, deadlines and periods.

During this first year of work we have:

1. Simplified the allocation and scheduling algorithms of Lopht to facilitate proof while still being able to handle the industrial use case. The resulting algorithms consider all the aspects pertaining to functional specification and non-functional requirements, but make simplifying assumptions on the execution platform (by not taking into account memory access interferences during parallel execution).

2. Developed a formally proved translation validation tool to determine the correctness of schedules produced by the algorithms at point (1). The tool is developed and proved in Coq. Coq code extraction is used to produce OCaML code that integrates in the allocation and scheduling flow.

3. Evaluated the tool on a large-scale industrial use case from Airbus (6000 Scade nodes). We demonstrated the tool to our project partners and during the ASSUME project evaluation. This evaluation showed that our scheduling and formally proved validation tools scale up to the size of large applications.

The main limitation of the current work is that it does not take into account the interferences due to concurrent memory accesses. This gives the main research direction for the next year.

We are currently writing a paper on this subject.