Post-processing large datasets efficiently in parallel requires good load balancing of geometry supplied to the visualization pipeline. When datasets are not pre-partitioned or cannot be read back from simulation output ...