NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
crates.io · #918 most downloaded on crates.io
Scalable lock-free delayed memory reclaimer
Last release 22 days ago
16 Sep 2026
Release timing varies
gaps range from 8 days to 3 months
Nearly every release is documented
notes for 46 of 49 stable releases
26 versions withdrawn
withdrawn after publishing
2 years old
75 releases · first in 2024
Nothing published for this version
Minor memory fence optimization.
Minor memory fence optimization.
One column per month.
Fix a data race on process termination: `PR15` by `Ollie-Pearce`.
PR15 by Ollie-Pearce.Fix `#14`: unsound Send implementation of AtomicOwned and Owned.
#14: unsound Send implementation of AtomicOwned and Owned.Fix `#14`: unsound Send implementation of AtomicShared and Shared.
#14: unsound Send implementation of AtomicShared and Shared.* Optimize PrivateCollector.
PrivateCollector.* Minor doc updates.
API update: add AtomicRaw::update_tag_if.
AtomicRaw::update_tag_if.* Minor doc updates.
* More robust OOM handling.
Fix loom issues with the scc crate.
loom issues with the scc crate.API update: PrivateCollector enables forced bulk-reclamation of memory chunks: `#6`.
PrivateCollector enables forced bulk-reclamation of memory chunks: #6.{Owned, Shared}::{new_checked, new_with_checked}.suspend; the function is unsound since this allows instances of !Send types to move to another thread.AtomicOwned and AtomicShared when storing Owned or Shared with Relaxed: #12.ebr when a thread is spawned in drop.Fix performance regression in 4.6.6.
* Minor doc updates.
Fix MSRV violation, affected versions = [4.7.0, 4.7.2]: `#8`.
MSRV violation, affected versions = [4.7.0, 4.7.2]: #8.API update: AtomicRaw::store no longer requires the caller to pass a reference to a Guard.
AtomicRaw::store no longer requires the caller to pass a reference to a Guard.Nothing published for this version
Add AtomicRaw and RawPtr for atomic raw pointer operations.
AtomicRaw and RawPtr for atomic raw pointer operations.Nothing published for this version
The size of Collector has been reduced by 50%: 128B -> 64B.
Collector has been reduced by 50%: 128B -> 64B.Fix a spurious error in Queue::is_empty.
Queue::is_empty.Optimize thread-local Collector traversal by using an arena allocator.
Collector traversal by using an arena allocator.Accelerate memory reclamation on {Bag, Queue, Stack}::drop.
{Bag, Queue, Stack}::drop.Nothing published for this version
Add support for MIRIFLAGS="-Zmiri-strict-provenance".
MIRIFLAGS="-Zmiri-strict-provenance".Add {Shared, Unique}::{new_with, new_with_unchecked}.
{Shared, Unique}::{new_with, new_with_unchecked}.Minor optimization of scanning remote thread-local variables.
Adjust epoch countdown parameters.
Remove LinkedEntry::take_inner: the method is dangerous to use.
codeberg.LinkedEntry::take_inner: the method is dangerous to use.* Add Guard::set_has_garbage.
Guard::set_has_garbage.Add Ptr::as_{ptr|ref}_unchecked.
Ptr::as_{ptr|ref}_unchecked.{Owned|Shared}::as_non_null_ptr.Prepare for an upcoming Rust breaking change: `Rust#136702`.
Rust#136702.* Add Ptr::as_ref_unchecked.
Ptr::as_ref_unchecked.* Minor code cleanup.
* Add Bag::try_push.
Bag::try_push.Add lock-free concurrent data structures: Bag, LinkedList, Queue, and Stack.
Bag, LinkedList, Queue, and Stack.Nothing published for this version
* Add Guard::has_garbage.
Guard::has_garbage.* Minor optimization.
Nothing published for this version
Guard::accelerate now only accelerates garbage collection of the current thread without affecting other threads.
Guard::accelerate now only accelerates garbage collection of the current thread without affecting other threads.* u8 can be converted to Epoch.
u8 can be converted to Epoch.Epoch uses a range of [0, 63] u8 values instead of rotating four values.
Epoch uses a range of [0, 63] u8 values instead of rotating four values.* More const functions.
Let miri not execute Intel-specific code paths.
miri not execute Intel-specific code paths.The size of Option is now that of Guard.
Option<Guard> is now that of Guard.Minor improvements to documentation.
Bump MSRV to 1.85.0 / Edition 2024.
Minor epoch update policy optimization.
NonNull optimization on Owned and Shared.Fix unsound Sync implementations of AtomicShared and Shared; previously, the Sync implementation allowed an arbitrary thread to own or drop the contai
Sync implementations of AtomicShared and Shared; previously, the Sync implementation allowed an arbitrary thread to own or drop the contained instance.* Minor const optimization.
const optimization.Fix a use-after-free issue when thread-local storage is dropped.
Nothing published for this version
* Fix minor linting errors.
Adjust tests to be more Miri friendly.
Miri friendly.Fix a rare memory ordering issue when dropping thread-local storage.
Make SDD much more Miri-friendly.
SDD much more Miri-friendly.Compatible with the `Miri` memory leak checker.
Miri memory leak checker.Collectible private since it is unsafe.Guard::defer which depends on Collectible.prepare.Nothing published for this version
Your coding agent can read these notes before it upgrades. Set up the MCP server →