GVDB_VOX macro and aprons

[nvidia]

GVDB_VOX macro and aprons

[ramakarl]

Hi,
This is an interesting suggestion, but changing GVDB_VOX directly would be a fairly big change since this is used by many different kernels. In order to accept such a big change, you would need to create a pull request that also included testing of all the existing samples.

Instead, I would suggest that you consider a pull request where you simply provide a new macro, say GVDB_VOX_ONLY, or GVDB_VOX_COVER, that only touches the actual voxels. Keep the original GVDB_VOX as is. Note that your solution is not yet complete, because while you have resolved the indexing to just touch actual voxels, you would also need to modify the host code to reduce the number of threads being launch. Otherwise you still have the same GPU grid size, and will not get much benefit, just cause some threads to be idle. The grid launch itself would need to only launch 1 thread per real voxel (/w apron), and then the indexing maps the logical thread to a real voxel, as you have done.

[alvingitlord]

Thanks for the comment, I will close this issue and may be create a pull request in the future