Interactive Parallelization of Embedded Real-Time Applications Starting from Open-Source Scilab & Xcos

Abstract

In this paper, we introduce the workflow of interactive parallelization for optimizing embedded real-time applications for multicore architectures. In our approach, the real-time applications are written in the Scilab high-level mathematical & scientific programming language or with a Scilab Xcos block-diagram ap-proach. By using code generation and code parallelization technol-ogy combined with an interactive GUI, the end user can map appli-cations to the multicore processor iteratively. The approach is eval-uated on two use cases: (1) an image processing application written in Scilab and (2) an avionic system modeled in Xcos. Using the workflow, an end-to-end model-based approach targeting multicore processors is enabled resulting in a significant reduction in devel-opment effort and high application speedup. The workflow de-scribed in this paper is developed and tested within the EU-funded ARGO project focused on WCET-Aware Parallelization of Model-Based Applications for Heterogeneous Parallel Systems.