Final Report Summary - LARGEFLOWVIS (Large Scale Flow Visualization and Analysis) Scientific visualization, i.e. the generation of images from scientific and engineering data that aid in interpreting and understanding this data, has recently achieved an important role in the scientific process. The increasing providence of scientific computing that enables computational experiments at unprecedented fidelity using large-scale computer simulations has resulted in a deluge of data that can only be reliably made sense of using advanced scientific visualization methods. In many scientific and engineering problems – for example the flow of air in a combustion chamber, fluid flow around embedded objects, or the study of magnetic fields of supernovae – vector field variables are used to model and describe transport. A large majority of visualization approaches devised to extract meaningful structures from vector fields are based on the study of their integral curves. Naturally understood as trajectories of massless particles, such curves are ideal tools to study transport, mixing and other similar processes. Scientific visualization of large-scale vector fields with corresponding integration-based methods that rely on the analysis of particle trajectories was not practicable with existing methods, since these were unable to make efficient use of parallel architectures such as clusters and supercomputers, leaving researchers in science and industry unable to visualize, analyze and understand the processes described by large vector field data from simulation or measurement. This project has developed a novel methodological framework for integration-based visualization that enables visualization of vector fields arising from large-scale simulation in current scientific applications. This methodology provides algorithms that are efficient and scale well to supercomputing architectures while offering conceptual simplicity and extensibility, advancing the state of the art. In particular, the following main results are obtained from the project:- A novel class of algorithms for the computation of integral-curves over large-scale datasets was developed, based on the work-stealing parallelization paradigm. These algorithms are well suited to a wide range of computational architectures, and provide increased flexibility over previous algorithms when used as building blocks for advanced visualization techniques.- A framework for pre-computation, filtering, and interpolation of integral curves was obtained. Through this, integral curves of particular relevance for a specific application can be quickly and interactively identified and leveraged to provide insightful visualization.- A novel approach to data reduction specifically aimed at visualization was devised that facilitates the visualization of integral curves in large-scale simulations on commodity hardware. Strong impact on fundamental scientific research in an interdisciplinary setting of scientific and industrial application areas that rely on vector field visualization has been demonstrated. Furthermore, the increased ability gained in this project to study large vector fields permits a renewed focus on fundamental vector field visualization research in the future. The developed approaches combine techniques from scientific visualization, parallel algorithms, applied mathematics, and software design. Within the project, novel visualization techniques have been developed that address specific problems in fluid flow, aerodynamics, and oceanography. In this regard, the research project and its applications include research on technologies related to timely problems such as combustion, fusion, and aerodynamics. Increased understanding of vector field processes will benefit scientific research and improvements to the “Resource Efficient Europe” flagship initiative of the Europe 2020 strategy.The ability to conduct the project has significantly aided the fellow in building up an independent research group at the University of Kaiserslautern, where he currently holds an assistant faculty position with a tenure track option. The successful completion of this project has enabled him to integrate deeply into the research community both within and outside the University of Kaiserslautern by fostering active collaborations with other researchers from computer science and other scientific disciplines. The results obtained within the project have aided the fellow in significantly broadening and deepening his independent research profile, and strongly benefited a successful tenure evaluation and thus re-integration into the European research area. Powiązane dokumenty final1-supplementary-figures.pdf
