Research program aims at parallelizing real-time software

October 21, 2011 // By Christoph Hammerschmidt
Research program aims at parallelizing real-time software
More powerful software for time-critical systems through parallel processing is the goal of the parMERASA research project funded by the EU. Examples for products that would benefit from the project are real-time systems in airplanes and cars.

The parMERASA project gets granular on the question how to improve the performance of industrial real-time software by means of parallel processing. It tries to find an answer to the question of how to parallelize sequential programs which have to meet hard real-time requirements - even though the different threads of the program will be executed concurrently on a multi-core processor? A precursor of the project took a look to available processors with up to eight cores while the current project focuses on processors with up to 64 cores and innovative inter-core communication mechanisms. In order to find answers to these questions, the researchers need to develop a suitable software design process, find appropriate parallel software structures, provide analysis tools and systems software and inquire into the appropriateness of hardware architectures.

The project is coordinated by Prof. Theo Ungerer from the Augsburg University. Participating are the Barceoloa Supercomputing Centre, the University of Dortmund, Honeywell International s.r.o in Brno (Czech Republic) and construction machine manufacturer Bauer Maschinen GmbH. The automotive camp is represented by Denso Deutschland. In addition, the Paul Sabatier University (Toulouse, France) and British software company Rapita Systems are providing real-time tools which allow calculating worst-case timing behavior estimates.

Benefitting from the results of the project will be real-time critical automotive, aerospace and industrial applications in general. In particular, the scientists hope to find ways to improve performance and safety in vehicles. Examples for software potentially making use of massive parallelism are advanced driver assistant systems and combustion engine injection controls - with extremely fast computers and adequate sensors and actuators, the engines could become more fuel-efficient.




Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.