TraceViz: une plateforme de visualisation pour l'analyse interactive de traces d'exécution
In Actes de la 28ème conférence francophone sur l'Interaction Homme-Machine (IHM 2016). pages 115-125. 2016.
Hardware platforms of embedded systems are more powerful at each new generation thank to the integration of System-on-Chip (SoC). Developing streaming multimedia applications on embedded systems becomes an increasingly complex process. Modern applications are highly multi-threaded and have to decode the multimedia stream in real time to prevent the apparition of audio and video artifacts. Debugging this kind of issue cannot be done with traditional debuggers that interrupt the decoding and perturb the synchronization of the different threads. The solution is to record all the events that occurred during the decoding in a trace and perform the analysis post-mortem. There exists many visualization tools to analyze execution traces but they have reached their limits with the amount of data generated by modern applications. They either provide a too generalized representation to be useful, or they show too much details leading to a fastidious data exploration. We propose a novel interaction visualization framework to address these problems. In particular, our contribution is in two parts: (a) we present a new fast backend suitable for the interactive browsing of huge traces and (b) a new visualization tool to explore the trace at different level of details.