Deadlock detection in a distributed implementation of a visualization system for medical measurement signals

We apply the methods proposed by Heiner (1992) and Heiner et al. (1994) for detecting anomalies in existing executable INMOS C code for a parallel program using static and dynamic analyses of Petri nets. The architecture used is a transputer network. The algorithmic translation of the INMOS C program into hierarchical place/transition nets which preserves both control and message flow is performed by a Petri net generator. Every implemented process is converted into an equivalent net. After this, the net parts generated are visualized and linked, and supplements are added, using a powerful graphical Petri net editor. Hierarchical methods are furnished and communication interface objects are highlighted. Every net object in the graphical representation is referenced with its counterpart in the source code so that error localization is possible. In the next processing step, the output capabilities of the Petri net editor are used to produce several input files for analyzing tools. Finally, the process nets investigated separately are brought together by merging the communication interfaces into one whole top level net. Here, a renewed analysis is performed.