A Proposal for a Distributed Performance Analysis Tool for a Task-based Parallelism Framework

We propose CharmVZ, a distributed Python-based performance analysis tool for applications developed with Charm++. Inspired by StarVZ, CharmVZ addresses the limitations of current tools by offering a distributed option to analyze large-scale traces of widely used applications such as ChaNGa and NAMD. Additionally, by integrating with the Python ecosystem, CharmVZ will enable new ways of analyzing data, for example, through machine learning, making it a novel tool for optimizing HPC applications.