Biswas, AyanStrelitz, RichardWoodring, JonathanChen, Chun-MingShen, Han-WeiEnrico Gobbetti and Wes Bethel2016-06-092016-06-092016978-3-03868-006-21727-348Xhttps://doi.org/10.2312/pgv.20161183https://diglib.eg.org:443/handle/10Streamlines are commonly used for visualizing flow fields, but particle-tracing based streamline computation usually does not scale well as the data size and complexity increase. Large flow simulations like global ocean or climate models can obtain near perfect load balancing and the resulting data sets are generally analyzed in two dimensional slices. To match the computational properties of these simulations, we propose the use of flux- based stream functions for generating streamlines in parallel. In our method, local stream functions are efficiently generated per block based on flux conservation property, followed by low-cost communication of flux offsets among neighboring blocks. A scalar field is thus generated where streamlines can be extracted through parallel iso- contouring. Experimental results show that our system offers higher streamline computation performance with higher scalability than traditional particle-tracing based method.10.2312/pgv.2016118369-78