I recently had my mind blown by a number of things in John Lakos's talks about C++17 PMR allocators
https://www.youtube.com/watch?v=CFzuFNSpycI . It seems like OpenVDB could really benefit from them (and just from using C++11-style templated allocators). Has this been considered?
Between using scaleable allocators, the fact that OpenVDB tends to allocate only a few block sizes, and that cache coherency can help spatial algorithms, it feels like there are some big potential wins. Additionally Trees seem like ideal candidates for the instant "winking out" destruction that arena allocators make possible.
Cheers,
Ben