The current mechanism for thresholding data reads is to use .quickinfo from PDAL to estimate a point count for a window, and then split and do the same thing again recursively. This has a lot of unavoidable fixed overhead.
It would be better to slurp up the entire hierarchy, load it into memory, and then use that to define all of the work units at once.
The current mechanism for thresholding data reads is to use
.quickinfofrom PDAL to estimate a point count for a window, and then split and do the same thing again recursively. This has a lot of unavoidable fixed overhead.It would be better to slurp up the entire hierarchy, load it into memory, and then use that to define all of the work units at once.