Emitted record and class constructors called scr_rt_calloc/scr_cyc_alloc
with a constant size, so the size class, the zeroing length and the
collector header setup were recomputed per object behind a call (plus an
empty scr_obj_alloc_note call and a variable-length memset). llvm/alloc.ts
now emits the fast paths inline: pop the class free list (clearing only
the bytes the helper does not store itself) or bump, with the old calls as
the cold slow path. Releases inline the weak-dispose hook check, the live
count, and the free-list push; blocks outside the reservation still go to
scr_rt_free.
The emitted IR does not know which runtime it links against (the sanitized
lane only differs at link time), so the runtime now defines scr_sa in
every build. Where the allocator is compiled out (ASan, SCR_RC_AUDIT,
wasm32, Windows, 32-bit, thread-instanced libraries) it stays all-zero and
the inline paths always take the slow path, which keeps the RC-audit
object notes. ScrSaState's layout is pinned with static asserts. Inline
emission is additionally gated off for 32-bit, WASI, Windows triples and
thread-instanced libraries.
Runtime callers with hot constant-size allocations (union boxes, arrays,
boxes, closures, dyn nodes) use new static inline twins
scr_cyc_alloc_inline/scr_cyc_free_inline; scr_cyc_live is exported.
The ABI guard now recognizes function-pointer extern data
(scr_weak_dispose_hook). New corpus program 4410 pins recycled-block
clearing, size classes, objects above the largest class, cycles and weak
observers.
On the x86_64 lane 41% of records-sort samples sat on the first load of
each record header (the comparator argument retain): a merge step touches
one element it has not seen before per comparison. Prefetching the
element four positions ahead on the advancing side (and during run
detection) overlaps that miss with the current comparator call: ~4%
faster on Linux (156 vs 162 ms median), neutral on macOS.
records-sort compares interned literals (`key === "department"`, equal
departments) and names that differ in their first bytes, yet every
comparison was an out-of-line call (and memcmp through a stub).
- sc_str_eq: identity, then byte length, then scr_str_eq.
- sc_str_cmp / sc_str_cmp_u16: identity; then the first differing byte
of one 8-byte word when both strings are that long, otherwise of the
first bytes; the UTF-16 order is decided inline only for ASCII bytes.
Everything else calls the runtime.
The helpers are internal module functions that LLVM inlines; they are
emitted whenever the runtime entry is declared.
The comparator sort kernel also holds one undefined-`this` window for the
whole sort (each emitted call restores the receiver stack it pushes) and
resolves the element retain function once.
After the union fix, records-sort spent most of its time in the IR merge
sort's element moves: every src[i] read retained the swapped `src` array
and the element (scr_arr_get_ref), every dst[k] write released the old
slot, and `dst` grew from an empty array.
scr_arr_sort_values runs the IR helper's exact algorithm (same comparator
call order) over borrowed slot copies of the private, compacted snapshot:
- comparator calls use the callValue ABI (owned element arguments, an
undefined `this` window, pending-exception check);
- the array keeps owning every element and is untouched while user code
runs, so the cycle collector and a throwing comparator see the original
snapshot; the sorted permutation is published only after success;
- f64 and reference-counted element types qualify (isSortValuesElement);
unions with an undefined arm and checked/island values keep the IR.
The new internal arrIntrinsic `sortValues` is validated, may throw, and
is not reference-preserving.
Move the allocator into scr_alloc.c with inline fast paths in scr_runtime.h;
strings, array headers/slot buffers, and emitted acyclic records/classes
(scr_rt_calloc/scr_rt_free) now use it. Adds scr_alloc.test.{c,ts}.
* Remove the large-array cliff past 2^20 elements
Arrays longer than 2^20 kept every element beyond the cutoff in a sorted
side store, and each indexed write to an existing entry deleted and
reinserted it, so fill + indexed read/write loops went quadratic
(n=1.2M: 10.4s on macOS arm64, 12.6s on Linux x64).
- Rewrite existing sparse entries in place.
- Let dense storage grow past the cutoff for contiguous extension of a
populated prefix, bulk range writes, and sparse tails that reach 3/16
density, migrating side-store entries so index < cap stays dense.
- Make the side store a deque with front slack so descending first-touch
writes and prefix migration are amortized O(1).
- Bound range-derived slice/splice capacity hints for hole runs.
scr_arr_new now reserves the live count, so a 2^20+3 element Set
converts into dense storage instead of a sparse tail; the white-box
test_lib check that pinned the old policy now asserts dense storage.
* Share immortal boolean union boxes
Maybe-undefined reads of boolean[] elements (`if (composite[i])` in a
sieve) box the element in a boolean|undefined union and free it at once:
one calloc/free pair per read, about a quarter of numeric-kernels samples
at large scales. Unions are immutable and boolean arms own no payload, so
low tags now return headerless immortal instances, the same convention as
the compiler's static unit-arm constants.
* Record preflight/order baselines for corpus 3138
- Preserve callback identity, receiver binding, and class references across unions and collections.
- Track record field presence independently of undefined values and improve native runtime behavior.
- Expand differential coverage and streamline CI bootstrap and compiler rebuild checks.
- Sort primitive arrays with stable native ordering and pure conversion keys.
- Preserve sparse semantics and argument effects while avoiding callback overhead.
- Verify Unicode ordering and ownership across native and WASI paths.
- Share dense array storage and move owned ranges without rebuilding containers.
- Batch array construction and insertion while preserving evaluation order and bounded temporary storage.
- Lower indexed mutations natively and preserve typed-array reversal across numeric element types.
- Combine bounded string concatenations while preserving evaluation and ownership.
- Build array joins and raw strings directly in their final storage.
- Support literal null and undefined string conversions with preserved effects.
- Reduce allocation and reference-management overhead in native array loops.
- Enable vectorization while preserving numeric behavior.
- Preserve live tuple iteration and cover reference lifetimes and missing values.
* Expand static compilation for common Effect 4 patterns
* Run every Effect 4 fixture slice in CI
* Preserve existing corpus behavior across Effect 4 lowering
* Preserve Effect platform callbacks and native stream state
* Record new corpus baselines and retain child stdin refusals
* Retire the obsolete stateful regex diagnostic
* Preserve promise reference count layout for cycle collection
- Store checked values in native collections while preserving supported object identities and JavaScript key equality.
- Support collection iteration and reject reference conversions that would copy or adapt stored objects.
Represent ordinary JavaScript arrays with hole, value, and explicit-undefined states; add sparse side storage and carry missing values through the frontend, native backends, runtime, and Node-facing consumers.
* feat: support thread-instanced library state
- Add the abi.instance_per_thread profile field: the archive's TUs compile
with -DSCR_THREAD_INSTANCES, and the SCR_TL qualifier in scr_runtime.h
moves every runtime unit's mutable statics into thread-local storage;
both backends emit the program TU's module globals, run-once guards, and
lazily-compiled regex literal caches thread-local to match
- ONE linked archive then serves one independent instance per embedder
thread through the unchanged entry family: a thread registers its sink
and calls the init entry, and owns its own collector, result arena,
poison flag, and program state — a trap poisons only the instance it
fired in while sibling threads' instances keep answering
- Document the contract beside the profile spec: one instance per thread,
selected implicitly by the calling thread; instance lifetime is the
thread's lifetime; the never-entered-from-two-threads rule is unchanged;
independent of and composable with abi.localize_runtime
- Immutable interned data (string literals, unit arms, template arrays,
vtables) stays shared — Darwin ASan's image-registration common
included, keeping the one-registration-per-image discipline; the
exception cell's current pointer resolves a NULL sentinel in this mode
because a thread-local address is not a constant initializer
- Add the four-thread acceptance probe (distinct per-thread workloads,
concurrent instance-local inits, per-instance collects, a trap delivered
to its own thread's sink exactly once), the composition probe pairing a
thread-instanced localized archive with a second different-prefix
localized archive, an explicit ASan rerun beside the suite-flavor
sanitized builds, and profile-shape coverage; schedule the new lane
contracts beside M1/M2
- Non-opted builds are byte-for-byte unchanged: the qualifier expands to
nothing outside -DSCR_LIB -DSCR_THREAD_INSTANCES (verified object-level
over every touched runtime TU in both default and library flavors)
* fix: keep sanitized runtime-localized archives linkable on ELF
ASan's instrumented globals ride ELF section groups; archives built from
shared runtime objects carry groups with REPEATED signatures, so a process
linking two runtime-localized sanitized archives kept one archive's group
and discarded the other's — whose now-local references then dangled at the
embedder's link. Resolving the groups into the combined relocatable member
(ld -r --force-group-allocation) keeps every archive's copies; the
localization step then demotes them per archive exactly like unsanitized
state. Plain builds carry no section groups, so the flag is inert there.
With the groups resolved, ELF surfaces the same deliberate exception
Mach-O already documented: the image-wide registration guard COMMON stays
shared so the final image registers its ASan globals exactly once — M1/M7
now pin that spelling on both platforms. The explicit ASan pairing (M8)
additionally points Linux LSan away from contractually thread-lifetime
instance state, exactly as the sanitized suite lanes do.
* fix: isolate inspect state per thread
* fix(runtime): share uptime anchor across threads
- Lower unshift and reverse through the C and LLVM backends.
- Preserve spread, mutation, evaluation-order, and reference-counting semantics.
- Add runtime, differential, and surface-manifest coverage.
- executable expansion in scr_console.c is byte-identical fputs/vfprintf + abort; core expansion in the new scr_core.c poisons, delivers to the registered sink exactly once, aborts pre-registration or if the sink returns
- 89 fputs/fprintf+abort pairs and 10 bare OOM aborts across the 20 core-linkable TUs now funnel; lazy atexit teardowns (json errdyn, regex bytecodes, symbol registry) reroute to the core reset registry via scr_atexit
- scr_core.c owns sink/poison/result-arena/reset-registry/marshalling helpers; scr_exception.c renders escaped exceptions as "Uncaught ..." into the funnel; scr_lib exposes its cleanup to the reset seam
Malte Ubl