Parallel Computing for Time and Safety Critical Applications

Cliquer ici pour une version française de ce document

CAPACITES is a LEOC (Logiciel Embarqué et Objets Connectés) project, gathering french academic and industrial partners in the field of critical real-time embedded software on many-core architectures.

Executive Summary

The objective CAPACITES is to build a hardware and software platform based on the exploitation of integrated many-core processors and to demonstrate the relevance of these many-core architectures and the implementation in Kalray's MPPA for several industrial applications. This family of processors built on the MPPA manycore architecture is currently the only one able to meet the needs of embedded systems simultaneously requiring high power computing, lower power consumption, and the ability to meet the requirements of critical systems (low latency input-deterministic processing times, and dependability).

This project aims to extend the development environment Kalray MPPA ACCESSCORE (optimized tools and domain-specific libraries) and align software technologies and partner offers the MPPA manycore architecture to adapt to critical parallel application requirements on time response and dependability. Application areas considered in the project are representative of critical embedded systems implementing significant computing power, and whose deployment is currently limited or impossible because of technological choices that have been made on the traditional many-core platforms. Indeed the architectural choices and production techniques used for these many-core platforms do not control the temporal effects of interference between cores, which poses a serious obstacle to their use in time-critical applications.

Some developments planned in this project will require specific changes in the missing hardware architecture of current processors MPPA. An FPGA-based emulator provided by Kalray will validate these hardware developments and the associated low-level software layers. The motivation of these architectural studies validated by FPGA emulation is to immediately be integrated into a variant of the MPPA processor for a particular market (avionics, automotive, medical, ...).

The application areas identified in this project are very important for the target markets Kalray society. The substantive work done in the field of avionics as part of this project is in itself a market. It will also be a powerful trigger to conquer the automotive market especially for ADAS-type applications ("Advanced Driver Assistance System") or automatic behavior that will claim considerable computing power again with strong time constraints and certification requirements. The work done by Probayes and MBDA employ technologies that will they be as relevant to the world of the automobile. Similarly the market for medical imaging is undergoing a mutation that leads to manufacturers of computing platforms capable of adapting quickly to new market needs without using dedicated circuits costly and time consuming to implement .


Application domain: Industrial software; embedded systems; avionics; transports; critical systems; robust perception; image processing; signal processing.

Technologies: many-core architectures; network-on-chip; network algebra; deterministic Ethernet; real-time scheduling; communication and execution time evaluation; parallel programming; virtualization; operating system; electric consumption; safety; observability; guaranteed real-time; embedded high-performance computing.



A list of publications written as part of CAPACITES is available here: publications.html


This project is supported by the French DGE and BPI. It is part of the "Logiciel embarqué et objets connectés" (Embedded software and connected objects) program of "Investissements d'Avenir - Développement de l'Economie Numérique".

It has been approved by the competitiveness clusters Aerospace Valley, Minalogic, Pegase and Systematic.


Renaud Stevens (Kalray)