\section{Conclusion}
\label{conclusion}


This paper presented Cheddar ADL, an Architecture 
Design Language for the scheduling of real-time systems.

The particularity of this ADL, as compared to other ADLs, is that it allows to 
capture all required aspects for the schedulability analysis of real-time systems. 

After the presentation of the requirements of this ADL, we 
classified the elements into two categories: software part, which 
contain \addressspace, \task, \buffer, \resource, \messages and \dependency, 
and hardware part, which contain \core, \cache, 
\processor and \network.

Another category, deployment, is in fact a combination of these two categories.

We then presented in detail each of theses elements, by giving for each, the definitions, 
the standard attributes, the legality rules, an implementation and an example.

Subsequently, we have shown how the language is used as an entry point 
to the Cheddar tool, a real-time scheduling simulator. 

In order to extend the usability of the language, 
we have also been interested in interoperability with other tools for analysing 
real-time systems that are more and more numerous.

For that, we proposed an approach which allow to use our Cheddar tool, in order 
to perform schedulability analysis of real-time systems described in 
other languages.
This approach consists in fact in transforming the entry ADL, into Cheddar ADL.

The language is meant to evolve. Therefore, our future works will concern 
the extension, in order to consider new hardware (e.g. heterogeneous multiprocessors), 
that can provide schedulability results closest to reality; and then, new feasibilities 
tests.

We also plan to focus on evaluation of our language, by comparing with other 
ADLs. The paper of \cite{Kamal07} will be the starting point. 
Indeed, \cite{Kamal07} compare different ADLs under certain aspects: 
syntax; visualization, which concerns graphical representation; 
variability and extensibility, the capability to model new patterns. 
The paper of \cite{Medvidovic00} provides also a good basis for this 
study, because it provides a framework which is used to classify 
and compare several existing ADLs.