Optimal Mapping of Task-based Computation Models over Heterogeneous Hardware using Placer

Optimal Mapping of Task-based Computation Models over Heterogeneous Hardware using Placer

Renaud De Landtsheer, Jean-Christophe Deprez, Christophe Ponsard, Optimal Mapping of Task-based Computation Models over Heterogeneous Hardware using Placer, Tool demonstration at the 21st International Conference on Model Driven Engineering Languages and Systems (MODELS), Copenhagen, Denmark, 14-19 October 2018.

Date: 14 octobre 2018

Publication: Publications scientifiques 

Expertises:

Ingénierie des systèmes IT complexes 

Algorithmique et Optimisation Combinatoire 

Evolutivité des systèmes embarqués et réseaux IoT 

A propos du projet: TANGO 

Placer is a model-based tool that, given a model of heterogeneous (or at least multi-core) hardware and a task-based complex software, finds a mapping of the software tasks on the various processing elements of the hardware, together with a routing of the transmissions on the available busses, and provides a schedule for the tasks and transmissions. The mapping can minimize either the run time, energy consumption, or both in a multi-objective fashion. The tool combines an Eclipse front-end and a web-service placement backend based on the OscaR constraint programming library. The tool demonstration will illustrate how to build the task model (either as JSON for third party or using the Sirius interface), submit it to the web-service, analyse the result back in the front-end and possibly iterate to improve the placement.

Conference web-site