Reflecting RTOS Model During WCET Timing Analysis: MSP430/Freertos Case Study

Josef Strnadel,Peter Rajnoha

doi:10.2478/v10198-012-0041-3

Abstract

The determination of the execution time upper bound, commonly called Worst-Case Execution Time (WCET), is a necessary step in the development and validation process for real-time systems. The WCET analysis techniques can be classified as static or dynamic. While a high-level language code suffices for the static techniques, for a precise WCET analysis a target architecture or its authentic simulator able to run the final machine-level code of an analyzed application is needed by the dynamic techniques. In the paper, we have decided not only to present a novel hybrid timing analysis technique, but also to show its practical applicability in the area of WCET analysis over particular embedded architecture (MSP430) and real-time operating system (FreeRTOS). Novelty of the presented method can be seen in the fact the operating system model is reflected during the analysis in order to facilitate the process of derivating schedulability test formulas, create detail task/stack analysis etc. Applicability of the method was tested using the MSPsim simulator of the MSP430 architecture.

Highlights

The information can be send to an OSModel object – implementing model of particular real-time operating system (RTOS) consisting of a task model and a scheduling-policy model; OSCommand is used to implement a reaction to an event – and/or to a RangeProfiler object implementing module for collecting statistics about previously specified piece of a code5
There are many functions implemented in FreeRTOS interface, but presentation of results related to all of them would occupy a lot of space of the paper
Time-complexity of the operations is introduced as a function of n and m

Summary

Introduction

The paper is organized as follows: First, approaches to modeling RT systems are introduced (section 1.1) with special attention payed to the implementation of RT systems by means of tasks running over the real-time operating system (RTOS) kernel. Lacks of the actual methods can be seen in the fact they do not reflect target RTOS during the analysis As a result, they make the derivation of mechanisms such as schedulability test more difficult and hard to automate. The method reflecting the fact is presented and at the end of the paper it is shown how the results can be utilized to derive the test for given scheduling mechanism and target resources (RTOS, platform and its simulator enriched about RTOS model).

Methods

Results

Conclusion