Interactive Software Parallelization Based on Hybrid Analysis and Software Architecture Reconstruction
Andreas Wilhelm · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2019
In this thesis we propose an interactive approach to support identification of parallelization opportunities in legacy software. Thus, we implemented Parceive, a tool that collects behavior information during run-time and combines it with architectural information to provide useful visualizations and analyses for parallelization. We additionally propose a rule-based language to reconstruct various views on the implemented software architecture. This enables multi-granular dependency analysis for parallelization. We demonstrate the usefulness of our approach on two case studies.