htg file format

I have a few questions about the HyperTreeGrid .htg format (as of ver 5.7):

  1. There is an option called TransposedRootIndexing which supposedly changes some indexing from ijk to kij or vice versa. To make the “description” array in the software i would like to use a
    for z{ for y{ for x nested loop order. Is this what the TransposedRootIndexing should do? (changing it from 0 to 1 does not change anything)

  2. I would like to use the ‘Mask’ array to write truncated domains from each separate process in an mpi simulation. However, it seems that every file needs to define a full cuboid domain, and then mask the unwanted cells, which is difficult with mpi, since each process only knows abouts its own grid and subdivisions. How would you use htg files in an mpi setting?

BR, Arne

The format is still evolving but hopefully should become fully stable for ParaView 5.8.

Some of the fields are getting renamed for consistency.

For the distribution, we always have full HTs on one process, so the distribution happen at the root grid level. PV 5.7 actually start handling distribution for HTG.

Also I think (not fully sure) that the transpose is more to align the grid along different axis especially for 2D dataset.

Thanks! I see. The program I’m working with splits the octrees between processes also at the higher levels (I think it simply snips the z-order curve into equal segments), so it will probably be challenging to write code that exports the trees correctly. Anyhow, thanks a bunch for adding proper support for octrees!