A framework to experiment optimizations for real-time and embedded software

TL;DR

This paper presents a flexible framework for experimenting with and evaluating various software transformations to optimize embedded systems under multiple constraints like code size, energy, and deadlines.

Contribution

It introduces a unified framework that allows systematic testing and impact analysis of different code transformations on embedded software performance.

Findings

Framework enables multi-criteria evaluation of transformations

Supports experimentation with compiler optimizations and memory mapping

Facilitates analysis of side effects on software constraints

Abstract

Typical constraints on embedded systems include code size limits, upper bounds on energy consumption and hard or soft deadlines. To meet these requirements, it may be necessary to improve the software by applying various kinds of transformations like compiler optimizations, specific mapping of code and data in the available memories, code compression, etc. However, a transformation that aims at improving the software with respect to a given criterion might engender side effects on other criteria and these effects must be carefully analyzed. For this purpose, we have developed a common framework that makes it possible to experiment various code transfor-mations and to evaluate their impact of various criteria. This work has been carried out within the French ANR MORE project.