Analysing gCSP Models Using Runtime and Model Analysis Algorithms

Abstract

Nowadays the amount of embedded systems is getting bigger and bigger. The systems themselves also become more complex. To aid developers of embedded systems a formal language called Communicating Sequential Processes (CSP) was designed. At the Control Engineering (CE) group a graphical tool called gCSP is developed to create CSP models. This tool allows for code generation. The generated code can be compiled with the Communicating Threads (CT) library into an executable. Users of gCSP tend to create a separate process for each task. From the model point of view this is excellent, but from a software point of view this consumes too many resources. First of all a new gCSP file format is created, such that an algorithm is able to read model information without too much effort. Next algorithms are designed and implemented, which perform analysis on models using different techniques. This assignment results into two analysis tools, which are able to perform an analysis of gCSP models. The first tool is a runtime analyser and uses the compiled executable to get information of the order in which the processes are running. The second tool is a model analyser which directly analyses a gCSP model. It creates a dependency graph and will try to group processes which are related to each other, resulting in a schedule for multi-core target systems. For single-core target systems the results can be used to recreate the model with a few bigger processes. Both tools are tested to see whether they both are functional and usable. Functional tests are executed using a simple model to be able to manually check the analysis results with the expected outcome. The usability tests perform tests using a series of models and interpret the outcome of the analysers. From these tests it is concluded that both tools are complementing each other and are suitable for analysing gCSP models in order to be able to create optimised models. The main recommendations are to implement the automatic creation of a new gCSP model using the results of the analysers and to improve the set of rules of both analysers even more to get better results.