update comments

This commit is contained in:
Daan Leijen 2023-03-28 09:14:17 -07:00
parent a21ddd03fe
commit 560e32b2e1

View file

@ -11,12 +11,10 @@ large blocks (>= MI_ARENA_BLOCK_SIZE, 32MiB).
In contrast to the rest of mimalloc, the arenas are shared between In contrast to the rest of mimalloc, the arenas are shared between
threads and need to be accessed using atomic operations. threads and need to be accessed using atomic operations.
Currently arenas are only used to for huge OS page (1GiB) reservations, Arenas are used to for huge OS page (1GiB) reservations or for reserving
otherwise it delegates to direct allocation from the OS. OS memory upfront which can be improve performance or is sometimes needed
In the future, we can expose an API to manually add more kinds of arenas on embedded devices. We can also employ this with WASI or `sbrk` systems
which is sometimes needed for embedded devices or shared memory for example. to reserve large arenas upfront and be able to reuse the memory more effectively.
(We can also employ this with WASI or `sbrk` systems to reserve large arenas
on demand and be able to reuse them efficiently).
The arena allocation needs to be thread safe and we use an atomic bitmap to allocate. The arena allocation needs to be thread safe and we use an atomic bitmap to allocate.
-----------------------------------------------------------------------------*/ -----------------------------------------------------------------------------*/