Why's NanoToOpen that slow ?

[ZephirFXEC]

Hey folks,
In my project I'm running computations on NanoVDB grids on the GPU, however I have to convert them back to OpenVDB in order to actually building the Grid in my software ( Houdini ).
I was wondering why the convertion from Nano to Open was sooo slow. We're talking 40ms for a float grid with 400K active voxel and very little sparsity, and of course it increase with the sparsity.

Is there an alternative to this, or a possibility to multithread this by splitting the nanovdb grid ?