Parallel accelerated isocontouring for out-of-core visualization
Chandrajit Bajaj, Valerio Pascucci, David Thompson, Xijin Zhang · 1999
In this paper weintroduce a scheme for static analysis that allows us to partition large geometric datasets at multiple levels of granularitytoachieve both load balancing in parallel computations and minimal access to secondary memory in out-of-core computations. The idea is illustrated and fully exploited for the case of isosurface extraction, but extendible to a class of algorithms based on a small set of algorithm parameters and for which an appropriate static analysis can be performed. 1 Introduction and Related Work Given a scalar #eld, w#x#, de#ned over a d-dimensional bounded volume mesh #x 2 R d #, we often visualize the data by rendering a #d , 1#-dimensional surface satisfying w#x#=const. This visualization technique is popularly known as isocontouring. In order to obtain a good understanding of the volume data, isosurfaces of multiple representative isovalues need to be computed. one needs to visit all cells of the input if you only query for a single isosurface. However, ...