Publication Date



Computer Systems Organization, Computing Methodologies


Programs depend on the entities. Their hierarchy varies with the complexity and as their utilization increases the complexity involved in hierarchy also increases. Visualization is a traditional method of representing complex entities to enhance the human understanding. Visualization of parallel programs is not a new concept. But, the approaches followed to obtain this goal of visualizing the Message Passing Interface (MPI) programs are unique in their own ways. There are no tools, which are capable of implementing all aspects of MPI programs in an efficient way. The main reason is the complexity of parallel programs. This thesis involves using Flatland [1], a virtual reality application and its implementation through a concept called Continuous Semantic Zooming (CSZ) [2]. It is hypothesized that this method of visualization can help naïve programmers, that is, people with little or no programming experience to understand the concept of MPI programming and also can serve the purpose of the experts in the fields of MPI programming to look deeper into the issues involving debugging and analyzing the MPI program behavior. The output generated as a trace file from a set of nine MPI programs is taken as an input for the visual representation. We focus on the design, data format, the results and experiences obtained from this project.