ARC: optional cache-line isolation of the reference count#403
Draft
DTW-Thalion wants to merge 4 commits into
Draft
ARC: optional cache-line isolation of the reference count#403DTW-Thalion wants to merge 4 commits into
DTW-Thalion wants to merge 4 commits into
Conversation
Per issue gnustep#398. The reference-count fast paths in arc.mm seeded their CAS loops with __sync_fetch_and_add(refCount, 0), a full read-modify-write used only to read the count word, and updated it with sequentially consistent __sync_val_compare_and_swap. The seed only needs a plain load, and the exchanges do not need seq_cst. View the count word through std::atomic<uintptr_t> and operate on it directly. A relaxed load seeds each loop; compare_exchange_weak is acquire-release on the retain and weak-flag paths and release on the decrement; the final release takes an acquire fence before running -dealloc. No functional or ABI change. Isolating just the atomic pattern with no Objective-C involved, on a 32-core x86-64 host with clang 18.1.3: __sync seed + CAS, seq_cst 27.6 ns relaxed load seed + acq/rel CAS 14.4 ns and a retain/release pair on a real object drops from about 31 ns to about 16 ns. All 194 tests pass, along with an 8-thread retain/release balance stress where the count returns to its exact starting value and an 8-thread weak load/store/dealloc race stress.
| // retaining/releasing adjacent small objects ping-pong a shared line | ||
| // (measured: distinct-object retain/release at 4 threads 127ns -> 24ns). | ||
| // | ||
| // The cost is memory: cache-line alignment rounds every small allocation up to |
Member
There was a problem hiding this comment.
Note that this is specific to the malloc implementation used. Handling aligned allocations of things that are not a multiple of the alignment size is painful, so snmalloc rounds up the size.
I am also curious how much of the false-sharing overhead is an artefact of the allocation policy. Do you know how snmalloc affects your workload with and without this relative to the baseline?
Contributor
Author
There was a problem hiding this comment.
Good question - so yesterday I wrote a test to try to answer that and I sent you an e-mail on the results. Short answer: snmalloc does not remove the cliff. When the objects are allocated together it packs the 16-byte instances onto shared cache lines the same as glibc, and is in fact worse at high thread counts. Its size-class rounding does not spread these apart.
… CAS loop The strong-retain and release fast paths spun a compare-exchange loop that re-tried on every lost race, so under contention they wasted work that a single read-modify-write instruction avoids. A strong retain runs while the caller still owns a reference, so the object cannot be at (or reach) the deallocating sentinel; its increment is therefore a single fetch_add. Release becomes a single fetch_sub, handling the last-reference and saturation edges after the fact. The weak-to-strong retain keeps the compare-exchange loop, because it can race a concurrent final release and has to check-and-increment atomically to avoid resurrecting a dying object. Reserve the bit below the weak flag as a guard, so an optimistic increment can never carry a saturating count into the weak flag. FastRefCount.m mirrors the reference-count layout and is updated for the guard bit; the saturation and weak-at-saturation cases it exercises still pass. Measured on a 32-core machine: retain/release falls from 16.1 to 11.3 ns with no contention, and a single shared object under 24 threads from 2143 to 1124 ns.
DTW-Thalion
force-pushed
the
arc-refcount-isolation
branch
2 times, most recently
from
July 24, 2026 13:12
70cb22a to
79b1c79
Compare
DTW-Thalion
force-pushed
the
arc-refcount-isolation
branch
from
July 24, 2026 13:24
79b1c79 to
e4a58ca
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #399 and depends on it; the diff includes #399's commit until it merges.
Reference counts of distinct objects that share a cache line ping-pong between cores when those objects are retained and released concurrently, so unrelated objects contend even though they never share a reference.
This adds an opt-in -DOBJC_ALLOC_ALIGN=64 build that rounds each allocation up to a cache line, removing that false sharing at the cost of memory on the smallest objects. The default alignment is unchanged.
Opening as a draft while #399 is in review.