Ideally, we'd always return buffers to the pool, but ensuring this in the code is tricky & error prone. Instead, we'll treat allocators as best effort & measure them. This should probabilistically create the reuse we want in lieu of a more perfect solution and allow us to continue using & repopulating pools when we drop pointers.
Ideally, we'd always return buffers to the pool, but ensuring this in the code is tricky & error prone. Instead, we'll treat allocators as best effort & measure them. This should probabilistically create the reuse we want in lieu of a more perfect solution and allow us to continue using & repopulating pools when we drop pointers.