Visually enriching and comparing runtime performance of visualization pipelines

Abstract In many graphics and visualization frameworks, directed acyclic graphs are a popular way to model visualization pipelines. Pipeline editors provide a visual interface to modify the underlying graph and use, e.g., a node-link diagram to visualize the data and program flow. This paper proposes an interactive tool for post-mortem performance analysis and comparison of pipeline variants. We extend the node-link representation of a visualization pipeline and enrich it with fine-grained runtime performance metrics. Annotating this static structure with dynamic performance information lets the developer evaluate performance characteristics in depth. Our approach further supports a visual comparison of two user-selected states of the graph. The comparison allows us to identify the impact of (topological) changes on the graph’s performance. We demonstrate the utility of our approach with different scientific visualization use cases and report expert feedback.